cogito/calculet-npu-research-archive
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
main
calculet-npu-research-archive/history/llama.cpp/llama.cpp-commits-all.tsv
T

7104 lines
1.7 MiB
Plaintext
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
fd9bd632f346085920d523dd307a3e4c11cc0a05 e08e473cf292ba73789769b993bb35bcdfc31689 wangbomeng wangbomeng@calculet.tech 2026-06-01T14:15:35+08:00 wangbomeng wangbomeng@calculet.tech 2026-06-01T14:15:35+08:00 HEAD -> dev, origin/dev llama-bench: fix device-info
e08e473cf292ba73789769b993bb35bcdfc31689 f86112d6758298241ab2fa84ad4f0f7b1ace35c6 wangbomeng wangbomeng@calculet.tech 2026-06-01T11:47:01+08:00 wangbomeng wangbomeng@calculet.tech 2026-06-01T11:47:01+08:00 llama-bench:add device-info
f86112d6758298241ab2fa84ad4f0f7b1ace35c6 d1bc743c10080fed1253686ce01749b42611ded4 wangbomeng wangbomeng@calculet.tech 2026-05-29T16:40:38+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-29T16:40:38+08:00 add dump_pos,fix pos_tensor of pre_by_embeds
d1bc743c10080fed1253686ce01749b42611ded4 2f201160b1960fd7be18b70d5770599d7082dd2f wangbomeng wangbomeng@calculet.tech 2026-05-28T14:23:06+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-28T14:23:06+08:00 set u_batch = max_seq_len
2f201160b1960fd7be18b70d5770599d7082dd2f f131bf1908b8c5dd1f49d7ae7f616506fc471666 wangbomeng wangbomeng@calculet.tech 2026-05-27T09:58:31+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-27T09:58:31+08:00 delete useless code
f131bf1908b8c5dd1f49d7ae7f616506fc471666 33d96dae28e17208ef61a71dba40cda3c04f135a wangbomeng wangbomeng@calculet.tech 2026-05-27T09:38:51+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-27T09:38:51+08:00 fix patch of qwen2.5vl,n_patch of minicpm = 64,fix n_ctx_slot,comment out M-ROPE
ebce9e1a548c2329aa97f53b8b2ad50d968088dc e850ff4de25ad1221abc9bf3bb30df252ccd9c36 Yunzhe Jia yunzhe@calculet.tech 2026-05-27T09:38:00+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-27T09:38:00+08:00 origin/runtime_replace Refactor cal-llm integration by removing runtime capacity loading and adjusting request handling
e850ff4de25ad1221abc9bf3bb30df252ccd9c36 9985688f7fa96f755882f96a2cd8a18114fdc42d Yunzhe Jia yunzhe@calculet.tech 2026-05-26T10:59:45+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-26T10:59:45+08:00 Enhance cal-llm integration by adding runtime checks for CALRT_LIBRARY and updating link settings for Linux and Windows
9985688f7fa96f755882f96a2cd8a18114fdc42d 38ca6895815671a87ceec86b4aa4eb976580ce76 Yunzhe Jia yunzhe@calculet.tech 2026-05-25T09:05:49+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-25T09:05:49+08:00 Refactor CalrtEngineConfig initialization by removing redundant parameters
33d96dae28e17208ef61a71dba40cda3c04f135a 37f068d87cc2d0a0b40ce978031fc9932fc89077 wangbomeng wangbomeng@calculet.tech 2026-05-22T14:33:45+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-22T14:33:45+08:00 add llama-build.sh
1921024604db640f09ade83c8a437ebf15bde360 d028722a12c5a463cc97207787cfd03d7a2590b0 5af229fe518ee2d761ab8ae2a69f7a9aaa9057f4 Bomeng Wang wangbomeng@calculet.tech 2026-05-21T14:08:14+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-05-21T14:08:14+08:00 refs/stash WIP on dev: d028722a replace cparam.n_ubatch with n_ubatch()
5af229fe518ee2d761ab8ae2a69f7a9aaa9057f4 d028722a12c5a463cc97207787cfd03d7a2590b0 Bomeng Wang wangbomeng@calculet.tech 2026-05-21T14:08:14+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-05-21T14:08:14+08:00 index on dev: d028722a replace cparam.n_ubatch with n_ubatch()
37f068d87cc2d0a0b40ce978031fc9932fc89077 13508e5e18bbb1202ae60c2ee5ecbdda1cca817b wangbomeng wangbomeng@calculet.tech 2026-05-21T14:07:00+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-21T14:07:00+08:00 release
13508e5e18bbb1202ae60c2ee5ecbdda1cca817b b6b7919468126d5ba30a10941825154789e9082f wangbomeng wangbomeng@calculet.tech 2026-05-21T11:43:59+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-21T11:43:59+08:00 reduce cmake log
b6b7919468126d5ba30a10941825154789e9082f 46847c24fc5c52dd57bb4fb17fc4d684149bb4fa wangbomeng wangbomeng@calculet.tech 2026-05-20T16:40:33+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-20T16:40:33+08:00 reduce cmake log
38ca6895815671a87ceec86b4aa4eb976580ce76 ddd60c928224b52eb4e1e60e01e8a1ecbb80ee45 Yunzhe Jia yunzhe@calculet.tech 2026-05-20T16:37:13+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-20T16:37:13+08:00 Add support for cal-llm profile dumping to JSONL file
46847c24fc5c52dd57bb4fb17fc4d684149bb4fa d028722a12c5a463cc97207787cfd03d7a2590b0 wangbomeng wangbomeng@calculet.tech 2026-05-20T16:02:12+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-20T16:02:12+08:00 Simplify CMakeLists.txt
d028722a12c5a463cc97207787cfd03d7a2590b0 63442fde8dc285817f725bd6c5fae80f478a3813 wangbomeng wangbomeng@calculet.tech 2026-05-20T14:41:05+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-20T14:41:05+08:00 replace cparam.n_ubatch with n_ubatch()
63442fde8dc285817f725bd6c5fae80f478a3813 fc3f4a9fab88e98eff8eeca8cbc6617333edba2c wangbomeng wangbomeng@calculet.tech 2026-05-20T11:47:42+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-20T11:47:42+08:00 fix ubatch and cmakelist
ddd60c928224b52eb4e1e60e01e8a1ecbb80ee45 2716130010c58f1cdd9d94f0e2f8e047bc92dc66 Yunzhe Jia yunzhe@calculet.tech 2026-05-20T09:22:40+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-20T09:27:06+08:00 Add cal-llm backend profiling support with command-line option
fc3f4a9fab88e98eff8eeca8cbc6617333edba2c c931ef3163940227c9b6291e04216245cce21429 wangbomeng wangbomeng@calculet.tech 2026-05-19T16:54:16+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-19T16:54:16+08:00 add dump and rt case generation
c931ef3163940227c9b6291e04216245cce21429 a43741244a646d3f8fa3ff01bb9b7d059223d0a5 wangbomeng wangbomeng@calculet.tech 2026-05-18T17:34:37+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-18T17:34:37+08:00 calrt: fix prefill_by_ids.fix create_buf
2716130010c58f1cdd9d94f0e2f8e047bc92dc66 b35bda124ccd4ed581dd6088d3ffa4931c2809b6 Yunzhe Jia yunzhe@calculet.tech 2026-05-18T08:40:54+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-18T08:40:54+08:00 fix n_ctx_slot error by adding runtime capacity loading for cal-llm
a43741244a646d3f8fa3ff01bb9b7d059223d0a5 b2003d6c90ef0fa26ab9e930ba3e22b355eb3d39 wangbomeng wangbomeng@calculet.tech 2026-05-15T14:39:53+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-15T14:39:53+08:00 run llama-passkey successfully, fix llama-bench, copy 3 times input to infer qwen2.5vl
b35bda124ccd4ed581dd6088d3ffa4931c2809b6 81a2cbd4cef5d28334ff513d6b69e593d9d1cb49 Yunzhe Jia yunzhe@calculet.tech 2026-05-15T09:19:31+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-15T09:19:31+08:00 Implement vocab-only model initialization and enhance cal-llm backend error handling
81a2cbd4cef5d28334ff513d6b69e593d9d1cb49 3b14ed556c15529fe1f94b649f49db7156dfaf67 Yunzhe Jia yunzhe@calculet.tech 2026-05-14T16:06:37+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-14T16:06:37+08:00 Enhance CMake configuration for cal-llm integration by updating paths and adding library properties
3b14ed556c15529fe1f94b649f49db7156dfaf67 dde8b8228081769c8d8f1e0751d47fa7875f8423 Yunzhe Jia yunzhe@calculet.tech 2026-05-14T14:34:25+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-05-14T14:34:25+08:00 - Removed references to CALRT from clip.h, mtmd-helper.cpp, mtmd-helper.h, mtmd.cpp, and mtmd.h. - Updated server.cpp to conditionally include CAL-LLM support. - Introduced option LLAMA_SERVER_USE_CAL_LLM in CMakeLists.txt to enable CAL-LLM integration. - Refactored server context to handle CAL-LLM requests and responses. - Updated image processing logic to remove CALRT-specific implementations. - Ensured compatibility with existing llama functionality while integrating new CAL-LLM features.
b2003d6c90ef0fa26ab9e930ba3e22b355eb3d39 7800c772d44a7415640ec1669a691232939927cf wangbomeng wangbomeng@calculet.tech 2026-05-13T15:24:24+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-13T15:24:24+08:00 try passkey and embedding
7800c772d44a7415640ec1669a691232939927cf dde8b8228081769c8d8f1e0751d47fa7875f8423 wangbomeng wangbomeng@calculet.tech 2026-05-13T15:23:54+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-13T15:23:54+08:00 try passkey and embedding
dde8b8228081769c8d8f1e0751d47fa7875f8423 fabcc7dac4b34d83fe430980bb364b81087c7ee9 wangbomeng wangbomeng@calculet.tech 2026-05-12T10:46:08+08:00 wangbomeng wangbomeng@calculet.tech 2026-05-12T10:46:08+08:00 clip:add minicpm patch;
fabcc7dac4b34d83fe430980bb364b81087c7ee9 1e9787962f20a7e82c71589ac1dd0585454e4bfe wangbomeng wangbomeng@calculet.tech 2026-04-30T16:17:17+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-30T16:17:17+08:00 calrt: fix encode dump
1e9787962f20a7e82c71589ac1dd0585454e4bfe f584311c716dd078c4b737eebbfda97fe12b7274 1bec0db5a675ddc60fb793be2a746e8f3c8855dc wangbomeng wangbomeng@calculet.tech 2026-04-30T16:06:23+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-30T16:06:23+08:00 Merge remote-tracking branch 'origin/dev-ben' into dev
f584311c716dd078c4b737eebbfda97fe12b7274 a339954cc2a145a80c5ea349e7896211916cbed9 wangbomeng wangbomeng@calculet.tech 2026-04-30T16:05:15+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-30T16:05:15+08:00 calrt: fix encode dump
1bec0db5a675ddc60fb793be2a746e8f3c8855dc 465df20c3e1f3f58d0f5531c796e0941a49c32b0 wangbomeng wangbomeng@calculet.tech 2026-04-30T16:02:59+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-30T16:02:59+08:00 origin/dev-ben bench: support 2 calbins
465df20c3e1f3f58d0f5531c796e0941a49c32b0 a339954cc2a145a80c5ea349e7896211916cbed9 wangbomeng wangbomeng@calculet.tech 2026-04-30T09:08:12+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-30T09:08:12+08:00 support one calbin
a339954cc2a145a80c5ea349e7896211916cbed9 8e76fc330ad624d1768b412d7894e272dbe696f4 eeeef3f6d6a5aad9296323ca15d9d929745b8932 wangbomeng wangbomeng@calculet.tech 2026-04-29T14:57:55+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-29T14:57:55+08:00 tag: v0.3.0 Merge branch 'dev' into dev-cli
8e76fc330ad624d1768b412d7894e272dbe696f4 431ad6f1c787f1b47e9e7be14b2e1a7d8565485c wangbomeng wangbomeng@calculet.tech 2026-04-29T14:56:59+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-29T14:56:59+08:00 finish dev of llama-cli
eeeef3f6d6a5aad9296323ca15d9d929745b8932 431ad6f1c787f1b47e9e7be14b2e1a7d8565485c wangbomeng wangbomeng@calculet.tech 2026-04-28T16:16:19+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-28T16:16:19+08:00 add unknown time info: prefill and decode
431ad6f1c787f1b47e9e7be14b2e1a7d8565485c efa1876bb8b2a449a55054a8c3745892f2c0f262 wangbomeng wangbomeng@calculet.tech 2026-04-27T16:52:15+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-27T16:52:15+08:00 update calrt_infer
efa1876bb8b2a449a55054a8c3745892f2c0f262 4a2a3181f1cea170105c771e6ba69d851b82b937 wangbomeng wangbomeng@calculet.tech 2026-04-27T11:35:58+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-27T11:35:58+08:00 tag: v0.2.2 update version print
4a2a3181f1cea170105c771e6ba69d851b82b937 e3ea5541259f35871c3e995331770aaca297e9ec wangbomeng wangbomeng@calculet.tech 2026-04-26T10:56:02+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-26T10:56:02+08:00 add calrt-version
e3ea5541259f35871c3e995331770aaca297e9ec 0b8e43943846a1cd05c26c9186b9ef778ebae6e2 wangbomeng wangbomeng@calculet.tech 2026-04-26T10:55:48+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-26T10:55:48+08:00 add calrt-version
0b8e43943846a1cd05c26c9186b9ef778ebae6e2 5eb46fffbd2e260c540fb112314ad67bbb8f56b6 wangbomeng wangbomeng@calculet.tech 2026-04-26T08:43:43+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-26T08:43:43+08:00 to debug
5eb46fffbd2e260c540fb112314ad67bbb8f56b6 005112c51e00d998b457337ca3dad53fc46e2d0b wangbomeng wangbomeng@calculet.tech 2026-04-24T16:24:51+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-24T16:24:51+08:00 clip: add image_to_patches; calrt: vision copy data from bf16 to fp32
005112c51e00d998b457337ca3dad53fc46e2d0b 509670642e8090a1713f5d73590b549be827aaef wangbomeng wangbomeng@calculet.tech 2026-04-24T16:24:31+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-24T16:24:31+08:00 clip: add image_to_patches; calrt: vision copy data from bf16 to fp32
509670642e8090a1713f5d73590b549be827aaef 5d219bfcef239c1e3ddd1baecc0514a9462ff6a8 wangbomeng wangbomeng@calculet.tech 2026-04-24T09:40:34+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-24T09:40:34+08:00 rm debug code
5d219bfcef239c1e3ddd1baecc0514a9462ff6a8 6787a53ab9b4897c5a7b51a4062d7b90e6697d35 wangbomeng wangbomeng@calculet.tech 2026-04-24T09:31:32+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-24T09:31:32+08:00 calrt: recover base = batch_index * dict_len
6787a53ab9b4897c5a7b51a4062d7b90e6697d35 e1a76abf66178ea85656479a5e19aea7785c09f4 wangbomeng wangbomeng@calculet.tech 2026-04-24T09:22:02+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-24T09:22:02+08:00 calrt:input of multimodel from fp32 to bf16
e1a76abf66178ea85656479a5e19aea7785c09f4 99dd308adb8488fa869bb669e3335e646a938eff wangbomeng wangbomeng@calculet.tech 2026-04-22T16:29:22+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-22T16:29:22+08:00 calrt:add CalcoreRT version print
99dd308adb8488fa869bb669e3335e646a938eff 06c7852e99e34c71e24e3ef6001cb54c72970e4d wangbomeng wangbomeng@calculet.tech 2026-04-21T15:50:01+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-21T15:50:01+08:00 tag: v0.2.1 server: fix max_seq_len
06c7852e99e34c71e24e3ef6001cb54c72970e4d 96b4e267e562f44e76ec6c965ee0fa361f0439bf ed62eb33447f993d578e156a3c7e38c91b420ece wangbomeng wangbomeng@calculet.tech 2026-04-21T15:06:01+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-21T15:06:01+08:00 Merge branch 'dev-ot' into dev
ed62eb33447f993d578e156a3c7e38c91b420ece 015e988a271573a7c11e2a4b45084e1219d69d50 wangbomeng wangbomeng@calculet.tech 2026-04-21T15:05:22+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-21T15:05:22+08:00 try server-test:not success
015e988a271573a7c11e2a4b45084e1219d69d50 99962021f2dd1994dbbf90d5712f0ca2633f20f8 wangbomeng wangbomeng@calculet.tech 2026-04-17T16:56:08+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-17T16:56:08+08:00 little change
96b4e267e562f44e76ec6c965ee0fa361f0439bf 99962021f2dd1994dbbf90d5712f0ca2633f20f8 wangbomeng wangbomeng@calculet.tech 2026-04-17T15:33:01+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-17T15:33:01+08:00 merge dev-ot
99962021f2dd1994dbbf90d5712f0ca2633f20f8 9af6682ef101c6d006c743c703cf150ac25f466f wangbomeng wangbomeng@calculet.tech 2026-04-17T14:48:33+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-17T14:48:33+08:00 origin/dev-ot calrt: add onetoken slice and untile_prefill
9af6682ef101c6d006c743c703cf150ac25f466f e24f775ee26ea5249d5ff6cd23b8a0a17fe1438d wangbomeng wangbomeng@calculet.tech 2026-04-16T10:09:44+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-16T10:09:44+08:00 args: add -ca
e24f775ee26ea5249d5ff6cd23b8a0a17fe1438d a206460e610aef995e36e50b33e8bd41fe477541 wangbomeng wangbomeng@calculet.tech 2026-04-15T10:01:09+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-15T10:01:09+08:00 calrt: add copy_data_to_ibuf for vision model
795179beb5bba728e0ccf0a7e80a5ec5ef36e92a 6d55eae7e76b768acfcec045b8e84cded8ae3b55 Yunzhe Jia yunzhe@calculet.tech 2026-04-14T16:03:22+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-14T16:03:22+08:00 origin/test-cicd add ci yaml
a206460e610aef995e36e50b33e8bd41fe477541 d8b2aaa00049978cebac15041f960d75b552f97b wangbomeng wangbomeng@calculet.tech 2026-04-14T10:23:43+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-14T10:23:43+08:00 calrt: support mixture of prefill_by_ids and prefill_by_embeds
d8b2aaa00049978cebac15041f960d75b552f97b b6f29ec8140028619efaa50aa6e73146cecf631d wangbomeng wangbomeng@calculet.tech 2026-04-13T11:23:44+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-13T11:23:44+08:00 server: fix context-shift
b6f29ec8140028619efaa50aa6e73146cecf631d 6d55eae7e76b768acfcec045b8e84cded8ae3b55 wangbomeng wangbomeng@calculet.tech 2026-04-13T11:23:31+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-13T11:23:31+08:00 server: fix context-shift
6d55eae7e76b768acfcec045b8e84cded8ae3b55 af8055f5f033a730e7e1fe80185b18b7e9473966 wangbomeng wangbomeng@calculet.tech 2026-04-10T14:58:56+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-10T14:58:56+08:00 tag: v0.2.0 add annotations
af8055f5f033a730e7e1fe80185b18b7e9473966 6f54682df62b8ec6bb4154b99c47b5becda3291f wangbomeng wangbomeng@calculet.tech 2026-04-10T14:58:19+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-10T14:58:19+08:00 add annotations
6f54682df62b8ec6bb4154b99c47b5becda3291f 42a181674565411efa5c8b318bc77d6fd084df8a a2d5bf362d6c9a546f9deadf4f8975605a915d33 Yunzhe Jia yunzhe@calculet.tech 2026-04-10T09:07:20+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-10T09:07:20+08:00 Merge branch 'dev' of http://192.168.10.26:1000/yunzhe/llama.cpp into dev
42a181674565411efa5c8b318bc77d6fd084df8a bd3a57a11aa008e0b5dc2a30f82fcd3a02587208 Yunzhe Jia yunzhe@calculet.tech 2026-04-10T09:06:55+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-10T09:06:55+08:00 fix prompt_cache issue
a2d5bf362d6c9a546f9deadf4f8975605a915d33 bd3a57a11aa008e0b5dc2a30f82fcd3a02587208 Bomeng Wang wangbomeng@calculet.tech 2026-04-09T21:01:46+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-04-09T21:01:46+08:00 server: comment fixed time
bd3a57a11aa008e0b5dc2a30f82fcd3a02587208 fcfcdf80b01664a08fb12df879a05df3cc0eba70 74d437ab1456831573def7ca385a74d0ca460c37 Yunzhe Jia yunzhe@calculet.tech 2026-04-09T19:42:57+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-09T19:42:57+08:00 Merge branch 'dev' of http://192.168.10.26:1000/yunzhe/llama.cpp into dev
fcfcdf80b01664a08fb12df879a05df3cc0eba70 0cd44929b442860dc0e5f88976108be82350f9dc Yunzhe Jia yunzhe@calculet.tech 2026-04-09T19:41:52+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-09T19:41:52+08:00 tmp fix one run >4K problem
74d437ab1456831573def7ca385a74d0ca460c37 5813c943005d1ede551e67ee98e35aefd210ff22 0cd44929b442860dc0e5f88976108be82350f9dc wangbomeng wangbomeng@calculet.tech 2026-04-09T17:44:26+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-09T17:44:26+08:00 Merge branch 'dev' of http://192.168.10.26:1000/yunzhe/llama.cpp into dev
5813c943005d1ede551e67ee98e35aefd210ff22 1f0d5bb0453f93ffa51faf68237c89251c2e0c40 wangbomeng wangbomeng@calculet.tech 2026-04-09T17:38:57+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-09T17:38:57+08:00 add --device-info
1f0d5bb0453f93ffa51faf68237c89251c2e0c40 b8153b6b8f244b223c9f4c2940ae1f212634519e wangbomeng wangbomeng@calculet.tech 2026-04-09T17:38:37+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-09T17:38:37+08:00 add --device-info
0cd44929b442860dc0e5f88976108be82350f9dc b8153b6b8f244b223c9f4c2940ae1f212634519e Bomeng Wang wangbomeng@calculet.tech 2026-04-09T15:08:50+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-04-09T15:08:50+08:00 rm calculet0.txt
b8153b6b8f244b223c9f4c2940ae1f212634519e 82842cfc3fa3d5c004c4540aeff8a81f38509bee wangbomeng wangbomeng@calculet.tech 2026-04-09T15:07:02+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-09T15:07:02+08:00 server: fix limit tokens to max_seq_len
82842cfc3fa3d5c004c4540aeff8a81f38509bee 622a22991089c6debd5f4b57738f3df3a0bda8b5 wangbomeng wangbomeng@calculet.tech 2026-04-09T15:06:26+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-09T15:06:26+08:00 server: fix limit tokens to max_seq_len
622a22991089c6debd5f4b57738f3df3a0bda8b5 5f47a738ba56a37d0ff281a16212f41979568e35 wangbomeng wangbomeng@calculet.tech 2026-04-09T10:36:31+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-09T10:36:31+08:00 server: comment circle print
5f47a738ba56a37d0ff281a16212f41979568e35 ecf01a17651cd7b1e8b85fd6075e83831463ee23 207ca38c751f7e6c251833eb11e89a2a0bb0fd07 Yunzhe Jia yunzhe@calculet.tech 2026-04-09T09:48:20+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-09T09:48:20+08:00 Merge branch 'dev' of http://192.168.10.26:1000/yunzhe/llama.cpp into dev
ecf01a17651cd7b1e8b85fd6075e83831463ee23 99d2078e89305035d87d966cee7987c7d50b3925 Yunzhe Jia yunzhe@calculet.tech 2026-04-09T09:47:55+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-09T09:47:55+08:00 fix n_parallel problem
207ca38c751f7e6c251833eb11e89a2a0bb0fd07 99d2078e89305035d87d966cee7987c7d50b3925 wangbomeng wangbomeng@calculet.tech 2026-04-08T19:14:37+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-08T19:14:37+08:00 delet extra printf
99d2078e89305035d87d966cee7987c7d50b3925 9626239f5a9e568c9975d67dd6745fef90279a87 Yunzhe Jia yunzhe@calculet.tech 2026-04-08T19:01:16+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-04-08T19:01:16+08:00 fix buffer resize problem
9626239f5a9e568c9975d67dd6745fef90279a87 da724f3e2a8af2c2537e3edcff41e9415ac66fef wangbomeng wangbomeng@calculet.tech 2026-04-08T17:47:04+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-08T17:47:04+08:00 calrt: fix MapBuf
da724f3e2a8af2c2537e3edcff41e9415ac66fef 18814f9bf1a423ebe8f32d6043c9cbc5b21b0218 Bomeng Wang wangbomeng@calculet.tech 2026-04-08T16:58:32+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-04-08T16:58:32+08:00 calrt: add MapBuf
18814f9bf1a423ebe8f32d6043c9cbc5b21b0218 bbee06d6d851e3f84f75b578047967d7e0e4a8dc wangbomeng wangbomeng@calculet.tech 2026-04-08T13:55:52+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-08T13:55:52+08:00 server : reset inference time
bbee06d6d851e3f84f75b578047967d7e0e4a8dc e126c21ed7b793b648d63533ea7ee30885f5a82a wangbomeng wangbomeng@calculet.tech 2026-04-08T11:29:36+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-08T11:29:36+08:00 calrt: add test time print
39edc6ec2e9560ccb20a2b8e547f58102bb268c1 2006831fd6ea8a8e16518b81bcee29fb674676ea Bomeng Wang wangbomeng@calculet.tech 2026-04-01T14:06:50+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-04-01T14:06:50+08:00 tag: v0.1.1 add t_incopy t_outcopy
2006831fd6ea8a8e16518b81bcee29fb674676ea 41e6636ea1fb3884e41fb2023f92b0e7262ef09b wangbomeng wangbomeng@calculet.tech 2026-04-01T10:42:27+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-01T10:42:27+08:00 tag: v0.1.0 v0.1.0
e126c21ed7b793b648d63533ea7ee30885f5a82a 5603e050af07173589f3b8850121263d86da01fd wangbomeng wangbomeng@calculet.tech 2026-04-01T10:07:12+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-01T10:07:12+08:00 sever: fix commit id
5603e050af07173589f3b8850121263d86da01fd 1e8a16cd0e3ad3c9a0f8746852a0a38c6e52322e wangbomeng wangbomeng@calculet.tech 2026-04-01T05:57:55+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-01T05:57:55+08:00 calrt: add t_incopy and t_outcopy
1e8a16cd0e3ad3c9a0f8746852a0a38c6e52322e 893e52bda0fee99bbda1521ea733a760bc4201f1 wangbomeng wangbomeng@calculet.tech 2026-04-01T10:14:22+08:00 wangbomeng wangbomeng@calculet.tech 2026-04-01T10:14:22+08:00 calrt: fix slice and add infer/copy time
893e52bda0fee99bbda1521ea733a760bc4201f1 398ecf71f5f574037ce33f84dd1576fa164909ad Bomeng Wang wangbomeng@calculet.tech 2026-03-30T11:35:40+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-03-30T11:35:40+08:00 delete unnecessary commit id
398ecf71f5f574037ce33f84dd1576fa164909ad 6d5e83902bdd3142a8ff0d3db19f02a8619e95a0 wangbomeng wangbomeng@calculet.tech 2026-03-27T02:01:23+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-27T02:01:23+08:00 rm llama-server_old.cpp
6d5e83902bdd3142a8ff0d3db19f02a8619e95a0 e4eaab700333490e80ebefc8d023f6c1adfcc312 wangbomeng wangbomeng@calculet.tech 2026-03-27T00:57:11+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-27T00:57:11+08:00 server: printf calrt commit ID
e4eaab700333490e80ebefc8d023f6c1adfcc312 aeacc23a883400db0ebffb50e23355ada054da66 wangbomeng wangbomeng@calculet.tech 2026-03-25T09:03:55+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-25T09:03:55+08:00 calrt: prefil to prefill
aeacc23a883400db0ebffb50e23355ada054da66 cbaa7492663630132adeddccd9fec5e44ffa3336 wangbomeng wangbomeng@calculet.tech 2026-03-25T08:49:22+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-25T08:49:22+08:00 calrt: add slice only on decode
cbaa7492663630132adeddccd9fec5e44ffa3336 3c08ebf0e442cd730ddffd959ec676f9caf31c82 wangbomeng wangbomeng@calculet.tech 2026-03-25T03:21:02+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-25T03:21:02+08:00 reduce warning
3c08ebf0e442cd730ddffd959ec676f9caf31c82 613bc2b9a9d7af92efbe42d876ffcdb244f7a00e wangbomeng wangbomeng@calculet.tech 2026-03-24T02:43:03+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-24T02:43:03+08:00 calrt: comment LOG_ERROR in untile_decode_cpy
613bc2b9a9d7af92efbe42d876ffcdb244f7a00e 4e194ecc6c4b8ebe4c72d93f6055e3605dbfcefa wangbomeng wangbomeng@calculet.tech 2026-03-24T02:30:12+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-24T02:30:12+08:00 calrt: fix max_seq_len judgment
4e194ecc6c4b8ebe4c72d93f6055e3605dbfcefa 16cbf81a731f5839a2f6b0eb9d8539826353748b wangbomeng wangbomeng@calculet.tech 2026-03-24T01:59:03+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-24T01:59:03+08:00 calrt: fix ubatch.embd cpy
16cbf81a731f5839a2f6b0eb9d8539826353748b 8b4e17d990c23025148c4abadefe1010a0dd0734 wangbomeng wangbomeng@calculet.tech 2026-03-23T09:38:21+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-23T09:38:21+08:00 calrt: fix model_name of untile_cpy
8b4e17d990c23025148c4abadefe1010a0dd0734 6dacdd2cf9574214ca272a86674e7bd7ba21c4f0 41e6636ea1fb3884e41fb2023f92b0e7262ef09b wangbomeng wangbomeng@calculet.tech 2026-03-23T14:50:27+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-23T14:50:27+08:00 merge origin dev
6dacdd2cf9574214ca272a86674e7bd7ba21c4f0 3e6ad64467771224f072e5bad90e270951ba4cdf wangbomeng wangbomeng@calculet.tech 2026-03-23T14:42:28+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-23T14:42:28+08:00 calrt: add multimodal
3e6ad64467771224f072e5bad90e270951ba4cdf b37b7d9756feb3dac2db526eeab38b572a095d47 wangbomeng wangbomeng@calculet.tech 2026-03-23T14:36:41+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-23T14:36:41+08:00 calrt: add multimodal
41e6636ea1fb3884e41fb2023f92b0e7262ef09b d0a1b2c758440720d42fa108b3444f54593daa73 Bomeng Wang wangbomeng@calculet.tech 2026-03-12T16:36:07+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-03-12T16:36:07+08:00 tag: v0.0.1 server: correct the max_seq_len judgment
d0a1b2c758440720d42fa108b3444f54593daa73 b37b7d9756feb3dac2db526eeab38b572a095d47 Bomeng Wang wangbomeng@calculet.tech 2026-03-12T16:10:28+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-03-12T16:10:28+08:00 calrt: comment out dump
b37b7d9756feb3dac2db526eeab38b572a095d47 07817cefc14fa359dcecad0630defd3610f8f5c8 7ddafbdcb9426ae3af016925b8b596f11e8742d0 Yunzhe Jia yunzhe@calculet.tech 2026-03-10T16:32:31+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-03-10T16:32:31+08:00 Merge branch 'dev' of http://192.168.10.26:1000/yunzhe/llama.cpp into dev
07817cefc14fa359dcecad0630defd3610f8f5c8 b2c45d8ddefb7ec09435231c8ec5b736fabf9c91 Yunzhe Jia yunzhe@calculet.tech 2026-03-10T16:32:07+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-03-10T16:32:07+08:00 fix kv issue
7ddafbdcb9426ae3af016925b8b596f11e8742d0 059009eeaf20a569abfbac5eb37dad3bb9376428 wangbomeng wangbomeng@calculet.tech 2026-03-10T07:13:04+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-10T07:13:04+08:00 calrt: commented out cal_ctx->encode(batch)
059009eeaf20a569abfbac5eb37dad3bb9376428 1b073661d1311b5300173993b9f31ee7241d8606 wangbomeng wangbomeng@calculet.tech 2026-03-10T07:04:22+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-10T07:04:22+08:00 calrt: add multimodel
1b073661d1311b5300173993b9f31ee7241d8606 6295273c5866b9249bacdedbfc3a31877e6e6a85 b2c45d8ddefb7ec09435231c8ec5b736fabf9c91 wangbomeng wangbomeng@calculet.tech 2026-03-10T06:58:35+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-10T06:58:35+08:00 Merge branch 'dev' of http://192.168.20.229:1000/yunzhe/llama.cpp into dev
6295273c5866b9249bacdedbfc3a31877e6e6a85 c6ea3f0d2468e3b8eab5de80bacee7fe4c9dd3d7 wangbomeng wangbomeng@calculet.tech 2026-03-10T06:54:41+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-10T06:54:41+08:00 arg: update hf_repo url to https://download.calculet.tech:9443
b2c45d8ddefb7ec09435231c8ec5b736fabf9c91 c6ea3f0d2468e3b8eab5de80bacee7fe4c9dd3d7 Bomeng Wang wangbomeng@calculet.tech 2026-03-09T15:03:26+08:00 Bomeng Wang wangbomeng@calculet.tech 2026-03-09T15:03:26+08:00 clart: past_kv_len = n_tokens
c6ea3f0d2468e3b8eab5de80bacee7fe4c9dd3d7 f8fba66b657c51a1df1efc7267bfea9f6140cebf wangbomeng wangbomeng@calculet.tech 2026-03-09T03:26:47+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-09T03:26:47+08:00 CMakeLists: STATIC to SHARED
f8fba66b657c51a1df1efc7267bfea9f6140cebf db4a61d2234c2f457b42ecc3f7219a1e1a35508a Yunzhe Jia yunzhe@calculet.tech 2026-03-06T15:06:36+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-03-06T07:05:39+08:00 fix calrt_install include path
db4a61d2234c2f457b42ecc3f7219a1e1a35508a e672e02ca1a582f9e3a9d011d9e1b4fea4942a25 Yunzhe Jia yunzhe@calculet.tech 2026-03-06T06:59:58+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-03-06T06:59:58+08:00 adapt to host-rt install structure
e672e02ca1a582f9e3a9d011d9e1b4fea4942a25 27b82f64252779ad160a3fa76bd378c39421de36 7a3a9a0e76bb5f530beafcfe52e87e388e93117a Yunzhe Jia yunzhe@calculet.tech 2026-03-06T11:47:09+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-03-06T11:47:09+08:00 Merge branch 'dev' of http://192.168.10.26:1000/yunzhe/llama.cpp into dev
27b82f64252779ad160a3fa76bd378c39421de36 0bd95719fed2f0247c8d0199dd897d24157dfae6 Yunzhe Jia yunzhe@calculet.tech 2026-03-06T11:42:45+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-03-06T11:45:03+08:00 adapt to calrt_objs target
7a3a9a0e76bb5f530beafcfe52e87e388e93117a 583c387efaf7bc030574e554088de79d09a27fbd wangbomeng wangbomeng@calculet.tech 2026-03-05T07:24:59+08:00 wangbomeng wangbomeng@calculet.tech 2026-03-05T07:24:59+08:00 calrt_infer: fix past_kv_len
583c387efaf7bc030574e554088de79d09a27fbd 0edc14a60a2d5596f71d728c655e3523da924d89 wangbomeng wangbomeng@calculet.tech 2026-02-27T02:24:56+08:00 wangbomeng wangbomeng@calculet.tech 2026-02-27T02:24:56+08:00 caserver: lim generated tokens to n_ctx_per_seq
0edc14a60a2d5596f71d728c655e3523da924d89 aa0a17d719326a63e0d6fea60aa0ced44e7abc33 wangbomeng wangbomeng@calculet.tech 2026-02-27T01:57:39+08:00 wangbomeng wangbomeng@calculet.tech 2026-02-27T01:57:39+08:00 caserver: limit generated tokens to n_ctx_per_seq
aa0a17d719326a63e0d6fea60aa0ced44e7abc33 fa332e773da681f3e77d6ffe83a87edb557f758b wangbomeng wangbomeng@calculet.tech 2026-02-27T01:40:11+08:00 wangbomeng wangbomeng@calculet.tech 2026-02-27T01:40:11+08:00 calrt: add bf16_to_fp32
fa332e773da681f3e77d6ffe83a87edb557f758b aa54a7c637f2c92d924a5f35386975b1c27de898 wangbomeng wangbomeng@calculet.tech 2026-02-27T01:31:42+08:00 wangbomeng wangbomeng@calculet.tech 2026-02-27T01:31:42+08:00 Add bf16_to_fp32
aa54a7c637f2c92d924a5f35386975b1c27de898 84f2c0b5f334a7ab77347c6779b6027a951d0ff0 wangbomeng wangbomeng@calculet.tech 2026-02-27T01:30:36+08:00 wangbomeng wangbomeng@calculet.tech 2026-02-27T01:30:36+08:00 Add bf16_to_fp32
84f2c0b5f334a7ab77347c6779b6027a951d0ff0 0bd95719fed2f0247c8d0199dd897d24157dfae6 wangbomeng wangbomeng@calculet.tech 2026-02-09T01:46:36+08:00 wangbomeng wangbomeng@calculet.tech 2026-02-09T01:46:36+08:00 add bf16_to_fp32 in untile_decode/prefill_cpy
0bd95719fed2f0247c8d0199dd897d24157dfae6 365eb0b719e30b0f9e348d854f11c9c3f954a96e Yunzhe Jia yunzhe@calculet.tech 2026-02-07T10:25:06+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-02-07T10:25:06+08:00 fix get_model_by_name
365eb0b719e30b0f9e348d854f11c9c3f954a96e 886cd1fb3b217efcf51c46209fcb694022346797 Yunzhe Jia yunzhe@calculet.tech 2026-02-07T09:05:17+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-02-07T09:05:17+08:00 comment out pos buffer file
886cd1fb3b217efcf51c46209fcb694022346797 9328f21378275f2354a19c3af907332b216ab492 Yunzhe Jia yunzhe@calculet.tech 2026-02-06T17:36:54+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-02-06T17:36:54+08:00 fix compile problem
9328f21378275f2354a19c3af907332b216ab492 49fea70a20f550bfdb59d0ffe041b54919c22447 Yunzhe Jia yunzhe@calculet.tech 2026-01-28T15:32:36+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-01-28T15:32:36+08:00 add prefill-only mode
49fea70a20f550bfdb59d0ffe041b54919c22447 74a118df31b6a4acc8876b2a2db90d2395cc6ec3 Yunzhe Jia yunzhe@calculet.tech 2026-01-28T09:11:16+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-01-28T09:11:16+08:00 adapt to new calbin test_case
74a118df31b6a4acc8876b2a2db90d2395cc6ec3 98e8f9acfe87cd93646482cb2d942b8a132f0852 Yunzhe Jia yunzhe@calculet.tech 2026-01-13T14:32:47+08:00 Yunzhe Jia yunzhe@calculet.tech 2026-01-13T14:32:47+08:00 sync with calrt update
98e8f9acfe87cd93646482cb2d942b8a132f0852 8c8152a04036bb0d4756a4e08e54f514dae4e64d Yunzhe Jia yunzhe@calculet.tech 2025-12-10T00:20:57+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-12-10T00:20:57+08:00 add timestamp
8c8152a04036bb0d4756a4e08e54f514dae4e64d 5d2387931528fbe04aa8d66de935147efb65ca4d Yunzhe Jia yunzhe@calculet.tech 2025-12-09T14:40:04+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-12-09T14:40:28+08:00 adapt to calrt csr
5d2387931528fbe04aa8d66de935147efb65ca4d e6285a748657add8d974b78ca920c5b41753ee12 Yunzhe Jia yunzhe@calculet.tech 2025-12-08T16:40:34+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-12-08T16:40:54+08:00 add timestamp for one decode process
e6285a748657add8d974b78ca920c5b41753ee12 6b1ecdbbe3cc72ed8888be38f9e934cfd79b9fbb Yunzhe Jia yunzhe@calculet.tech 2025-11-27T16:58:35+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-27T16:58:35+08:00 adapt to calrt
6b1ecdbbe3cc72ed8888be38f9e934cfd79b9fbb 240d9bb75241c91896afe125f2fc74698a7395f3 Yunzhe Jia yunzhe@calculet.tech 2025-11-25T17:06:14+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-25T17:06:14+08:00 add kv_tensor for pld test
240d9bb75241c91896afe125f2fc74698a7395f3 7eeb241aad429a68fb8cfa3ba85d000fb2348cd2 Yunzhe Jia yunzhe@calculet.tech 2025-11-24T12:03:00+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-24T12:03:00+08:00 adapt to calrt modification
7eeb241aad429a68fb8cfa3ba85d000fb2348cd2 dfb6250031a25058dbc3757e8a7ec75d90dcb48a Yunzhe Jia yunzhe@calculet.tech 2025-11-21T09:20:01+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-21T09:20:01+08:00 fix byte_size error
dfb6250031a25058dbc3757e8a7ec75d90dcb48a e31bce25e85dc7fdf0a0e69fd0e58ba4a2ffaf3e Yunzhe Jia yunzhe@calculet.tech 2025-11-18T10:28:00+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-18T10:43:01+08:00 add elf in CMakelist
e31bce25e85dc7fdf0a0e69fd0e58ba4a2ffaf3e 6444e227c36df16199c40d8cead2244fe014f377 Yunzhe Jia yunzhe@calculet.tech 2025-11-07T11:52:17+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-07T11:52:17+08:00 dump src tensor
6444e227c36df16199c40d8cead2244fe014f377 d81d7b190ff5f21325167936496dfe5ab48b922e Yunzhe Jia yunzhe@calculet.tech 2025-11-07T09:19:11+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-07T09:19:11+08:00 test golden case
d81d7b190ff5f21325167936496dfe5ab48b922e e492f5dc8cec17529f53ef2bbdf7b502107f8148 1f5accb8d0056e6099cd5b772b1cb787dd590a13 Yunzhe Jia yunzhe@calculet.tech 2025-11-04T14:43:32+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-04T15:04:07+08:00 Merge branch 'master' into dev
e492f5dc8cec17529f53ef2bbdf7b502107f8148 2e05afd22b3d77c7013b11eee88e88b17188f21a Yunzhe Jia yunzhe@calculet.tech 2025-11-04T14:09:59+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-11-04T14:09:59+08:00 redo untile logic
1f5accb8d0056e6099cd5b772b1cb787dd590a13 2759ccdb4adc8568add4316780d5e675519b0775 Noah 99681487+NoahOksuz@users.noreply.github.com 2025-11-04T05:04:59Z GitHub noreply@github.com 2025-11-03T21:04:59-08:00 Fix garbled output with REPACK at high thread counts (#16956)
2759ccdb4adc8568add4316780d5e675519b0775 c5023daf607c578d6344c628eb7da18ac3d92d32 Aman Gupta amangupta052@gmail.com 2025-11-04T10:53:48+08:00 GitHub noreply@github.com 2025-11-04T10:53:48+08:00 CUDA: avoid mul + bias fusion when doing fusion (#16935)
c5023daf607c578d6344c628eb7da18ac3d92d32 e7da30b584dc1f2ee0414c4a1298ce64eef97e8d lhez lih@qti.qualcomm.com 2025-11-03T11:47:57-08:00 GitHub noreply@github.com 2025-11-03T11:47:57-08:00 opencl: support imrope (#16914)
e7da30b584dc1f2ee0414c4a1298ce64eef97e8d ed8aa63320393512bdcfe4b05b5ae01ba91888e1 Aleksander Grygier aleksander.grygier@gmail.com 2025-11-03T18:53:26+01:00 GitHub noreply@github.com 2025-11-03T18:53:26+01:00 fix: Viewing multiple PDF attachments (#16974)
ed8aa63320393512bdcfe4b05b5ae01ba91888e1 48bd26501b08a3f0bff1249db47f313641f7bebb Daniel Bevenius daniel.bevenius@gmail.com 2025-11-03T18:01:59+01:00 GitHub noreply@github.com 2025-11-03T18:01:59+01:00 model-conversion : pass config to from_pretrained (#16963)
48bd26501b08a3f0bff1249db47f313641f7bebb 622cd010ff4a65bef67edbf2f9bf4707c01f98f7 Georgi Gerganov ggerganov@gmail.com 2025-11-03T15:38:23+02:00 GitHub noreply@github.com 2025-11-03T14:38:23+01:00 server : add props.model_alias (#16943)
622cd010ff4a65bef67edbf2f9bf4707c01f98f7 070ff4d5356083d60b807bb34d36b31c3653a29e theo77186 theo77186@users.noreply.github.com 2025-11-03T14:29:11+01:00 GitHub noreply@github.com 2025-11-03T14:29:11+01:00 ggml: CUDA: add head size 72 for flash-attn (#16962)
070ff4d5356083d60b807bb34d36b31c3653a29e bf7b0c9725ed0c406c5debaea022ec7258430634 Xuan-Son Nguyen son@huggingface.co 2025-11-03T11:11:18+01:00 GitHub noreply@github.com 2025-11-03T11:11:18+01:00 mtmd: add --image-min/max-tokens (#16921)
bf7b0c9725ed0c406c5debaea022ec7258430634 fcfce040e816c542f374ad51461b3561d73c4bc9 Xuan-Son Nguyen son@huggingface.co 2025-11-03T10:25:55+01:00 GitHub noreply@github.com 2025-11-03T10:25:55+01:00 mtmd: pad mask for qwen2.5vl (#16954)
fcfce040e816c542f374ad51461b3561d73c4bc9 ee3a5a10adf9e83722d1914dddc56a0623ececaf Jinyang He hejinyang@loongson.cn 2025-11-03T14:40:02+08:00 GitHub noreply@github.com 2025-11-03T08:40:02+02:00 ggml : LoongArch fixes (#16958)
ee3a5a10adf9e83722d1914dddc56a0623ececaf 7e994168b1ccc12337ba8de939c4fd466107c1fb Olivier Chafik olivier.chafik@gmail.com 2025-11-03T05:33:56Z GitHub noreply@github.com 2025-11-03T07:33:56+02:00 sync: minja (glm 4.6 & minmax m2 templates) (#16949)
7e994168b1ccc12337ba8de939c4fd466107c1fb bcfa87622ae46be6345a8e3dfdbdc5ba5414042b shani-f s0556787439@gmail.com 2025-11-03T03:35:33+02:00 GitHub noreply@github.com 2025-11-03T09:35:33+08:00 SYCL: optimized repeat_back kernel (3× fewer asm instructions, 2× faster)Feature/sycl repeat back opt (#16869)
bcfa87622ae46be6345a8e3dfdbdc5ba5414042b a2054e3a8ff0da3978a4acc18c349ff58554d336 Sascha Rogmann 59577610+srogmann@users.noreply.github.com 2025-11-03T00:41:08+01:00 GitHub noreply@github.com 2025-11-03T00:41:08+01:00 feat(webui): improve LaTeX rendering with currency detection (#16508)
a2054e3a8ff0da3978a4acc18c349ff58554d336 dd5286805004db1f9ac3176a1cbbfe373bdda0f8 Shagun Bera 141054835+notV3NOM@users.noreply.github.com 2025-11-03T04:40:30+05:30 GitHub noreply@github.com 2025-11-03T00:10:30+01:00 test-backend-ops : fix segfault in moe-expert-reduce test in support mode and coverage (#16936)
dd5286805004db1f9ac3176a1cbbfe373bdda0f8 6b9a52422bac0f50dd8f1f8386744fa3ce9783bf Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-11-02T23:11:21+01:00 GitHub noreply@github.com 2025-11-02T23:11:21+01:00 ci : disable failing riscv cross build (#16952)
6b9a52422bac0f50dd8f1f8386744fa3ce9783bf 2f966b8ed87514e74bb96592217226cb6a6974dd Zhiyong Wang 85110830+ravenouse@users.noreply.github.com 2025-11-02T13:08:04-08:00 GitHub noreply@github.com 2025-11-02T22:08:04+01:00 model: add Janus Pro for image understanding (#16906)
2f966b8ed87514e74bb96592217226cb6a6974dd cd5e3b57541ecc52421130742f4d89acbcf77cd4 Georgi Gerganov ggerganov@gmail.com 2025-11-02T22:21:48+02:00 GitHub noreply@github.com 2025-11-02T21:21:48+01:00 clip : use FA (#16837)
cd5e3b57541ecc52421130742f4d89acbcf77cd4 87c9efc3b297b8a498716b1db3d061842e6fc85b Georgi Gerganov ggerganov@gmail.com 2025-11-02T18:14:04+02:00 GitHub noreply@github.com 2025-11-02T18:14:04+02:00 server : support unified cache across slots (#16736)
87c9efc3b297b8a498716b1db3d061842e6fc85b 76af40aaaad78c42faecd8016a88362c788b84b0 Aldehir Rojas hello@alde.dev 2025-11-02T08:56:28-06:00 GitHub noreply@github.com 2025-11-02T16:56:28+02:00 common : move gpt-oss reasoning processing to init params (#16937)
76af40aaaad78c42faecd8016a88362c788b84b0 7db35a7958a943be1693879f42d166f152613979 Adrian Lundberg 47256989+alundb@users.noreply.github.com 2025-11-02T10:28:37+01:00 GitHub noreply@github.com 2025-11-02T11:28:37+02:00 docs: remove llama_sampler_accept reference in sampling sample usage (#16920)
7db35a7958a943be1693879f42d166f152613979 a864132ba546b6385922226480ee4e392aaa065c mnehete32 33429707+mnehete32@users.noreply.github.com 2025-11-02T08:42:57+05:30 GitHub noreply@github.com 2025-11-02T11:12:57+08:00 CUDA: add FLOOR, CEIL, ROUND, TRUNC unary ops (#16917)
a864132ba546b6385922226480ee4e392aaa065c d38d9f0877a5872daa3c5f06fb9a86376bf15d50 Aaron Teo aaron.teo1@ibm.com 2025-11-02T08:48:46+08:00 GitHub noreply@github.com 2025-11-02T08:48:46+08:00 devops: fix failing s390x docker build (#16918)
d38d9f0877a5872daa3c5f06fb9a86376bf15d50 7fd205a8e8832ead273f62c5fd81d2b8aa910535 Aaron Teo aaron.teo1@ibm.com 2025-11-02T08:48:23+08:00 GitHub noreply@github.com 2025-11-02T08:48:23+08:00 ggml: add s390x cpu-feats (#16774)
7fd205a8e8832ead273f62c5fd81d2b8aa910535 2f68ce7cfd20e9e7098514bf730e5389b7bba908 Georgi Gerganov ggerganov@gmail.com 2025-11-02T00:15:31+02:00 GitHub noreply@github.com 2025-11-02T00:15:31+02:00 scripts : add script to bench models (#16894)
2f68ce7cfd20e9e7098514bf730e5389b7bba908 e4a71599e5846110159955dec0008eb4aa24222b Pascal admin@serveurperso.com 2025-11-01T19:49:51+01:00 GitHub noreply@github.com 2025-11-01T19:49:51+01:00 webui: auto-refresh /props on inference start to resync model metadata (#16784)
e4a71599e5846110159955dec0008eb4aa24222b dd5e8cab512c7752392b6e51a6f118f348fb3f16 Pascal admin@serveurperso.com 2025-11-01T17:14:54+01:00 GitHub noreply@github.com 2025-11-01T17:14:54+01:00 webui: add HTML/JS preview support to MarkdownContent with sandboxed iframe (#16757)
dd5e8cab512c7752392b6e51a6f118f348fb3f16 cf659bbb8ef9eb048e5153a27cf787fd83c05560 Adrien Gallouët angt@huggingface.co 2025-11-01T16:52:17+01:00 GitHub noreply@github.com 2025-11-01T16:52:17+01:00 vendor : update cpp-httplib to 0.27.0 (#16846)
cf659bbb8ef9eb048e5153a27cf787fd83c05560 d8b860a219c2415faac8cc0e50b48b4aa11e3b64 Xuan-Son Nguyen son@huggingface.co 2025-11-01T15:51:36+01:00 GitHub noreply@github.com 2025-11-01T15:51:36+01:00 mtmd: refactor preprocessing + support max/min pixels (#16878)
d8b860a219c2415faac8cc0e50b48b4aa11e3b64 1ae74882f8f6755e44dff8f23f3abdc5b53ab7c1 Aleksander Grygier aleksander.grygier@gmail.com 2025-11-01T15:35:57+01:00 GitHub noreply@github.com 2025-11-01T15:35:57+01:00 Add a setting to display message generation statistics (#16901)
1ae74882f8f6755e44dff8f23f3abdc5b53ab7c1 961660b8c395afb8903f61ace69396a158ea0cb4 Jaromír Hradílek jhradilek@gmail.com 2025-11-01T15:02:57+01:00 GitHub noreply@github.com 2025-11-01T15:02:57+01:00 webui: recognize AsciiDoc files as valid text files (#16850)
961660b8c395afb8903f61ace69396a158ea0cb4 74fef4129fa58f6277c243777a4894371479dbad Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-11-01T11:01:42+01:00 GitHub noreply@github.com 2025-11-01T11:01:42+01:00 common : allow --system-prompt-file for diffusion-cli (#16903)
74fef4129fa58f6277c243777a4894371479dbad 5d8bb900bc7daa84bfa7bb1d25ab7e32394919f3 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-11-01T08:55:25+01:00 GitHub noreply@github.com 2025-11-01T09:55:25+02:00 codeowners : update after refactor (#16905)
5d8bb900bc7daa84bfa7bb1d25ab7e32394919f3 2e76e013600cb0d51ccf158571ca1d0502952a07 Jeff Bolz jbolz@nvidia.com 2025-11-01T00:52:14-05:00 GitHub noreply@github.com 2025-11-01T06:52:14+01:00 vulkan: Fix multi_add invalid descriptor usage (#16899)
2e76e013600cb0d51ccf158571ca1d0502952a07 d3dc9dd898be805c23a408cc36daed5b3bf29221 Jeff Bolz jbolz@nvidia.com 2025-11-01T00:45:28-05:00 GitHub noreply@github.com 2025-11-01T06:45:28+01:00 vulkan: fuse mul_mat+add and mul_mat_id+add_id (#16868)
d3dc9dd898be805c23a408cc36daed5b3bf29221 bea04522ff1a0d8559ccfd353aa018dcfbb608cc Oliver Simons osimons@nvidia.com 2025-11-01T06:13:26+01:00 GitHub noreply@github.com 2025-11-01T13:13:26+08:00 CUDA: Remove unneded bias/gate dims in fused mmvq (#16858)
bea04522ff1a0d8559ccfd353aa018dcfbb608cc 0de0a01576772032008a689afc4d7c80685074c4 Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-10-31T23:40:23+01:00 GitHub noreply@github.com 2025-10-31T23:40:23+01:00 refactor : llama-model.cpp (#16252)
0de0a01576772032008a689afc4d7c80685074c4 e58d585604bbbbbc24f8149effe444621b5191c6 Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-10-31T21:20:47+01:00 GitHub noreply@github.com 2025-10-31T21:20:47+01:00 model : Minimax M2 (#16831)
e58d585604bbbbbc24f8149effe444621b5191c6 31c511a968348281e11d590446bb815048a1e912 Giuseppe Scrivano gscrivan@redhat.com 2025-10-31T21:20:07+01:00 GitHub noreply@github.com 2025-10-31T21:20:07+01:00 model : add Granite Hybrid nano types (#16896)
31c511a968348281e11d590446bb815048a1e912 6d39015a744b47bb7643ea73f412e6d64677f305 Johannes Gäßler johannesg@5d6.de 2025-10-31T15:57:19+01:00 GitHub noreply@github.com 2025-10-31T15:57:19+01:00 CUDA: Volta tensor core support for MMF (#16843)
6d39015a744b47bb7643ea73f412e6d64677f305 4146d6a1a6228711a487a1e3e9ddd120f8d027d7 Georgi Gerganov ggerganov@gmail.com 2025-10-31T16:25:50+02:00 Georgi Gerganov ggerganov@gmail.com 2025-10-31T16:26:28+02:00 sync : ggml
4146d6a1a6228711a487a1e3e9ddd120f8d027d7 8da3c0e200a586f768ada6f38745acb01380174c Aman Gupta amangupta052@gmail.com 2025-10-31T20:05:07+08:00 GitHub noreply@github.com 2025-10-31T20:05:07+08:00 CUDA: add expert reduce kernel (#16857)
8da3c0e200a586f768ada6f38745acb01380174c c22473b580807929fd9e3a3344a48e8cfbe6c88f Georgi Gerganov ggerganov@gmail.com 2025-10-31T13:50:33+02:00 GitHub noreply@github.com 2025-10-31T13:50:33+02:00 batch : fix consistency checks for the input positions (#16890)
c22473b580807929fd9e3a3344a48e8cfbe6c88f 0f715b4e759acceccb9f437cfd2a988fff85514a Georgi Gerganov ggerganov@gmail.com 2025-10-31T10:54:19+02:00 GitHub noreply@github.com 2025-10-31T10:54:19+02:00 server : don't print user inputs to console (#16871)
0f715b4e759acceccb9f437cfd2a988fff85514a d2d931f173b8a736b08999436e9259aafddec718 Daniel Bevenius daniel.bevenius@gmail.com 2025-10-31T09:51:26+01:00 GitHub noreply@github.com 2025-10-31T09:51:26+01:00 server : fix typos in server.cpp comments [no ci] (#16883)
d2d931f173b8a736b08999436e9259aafddec718 2976b0374d36609b0429dd6ce48807e2ad39a7c2 Jeff Bolz jbolz@nvidia.com 2025-10-31T02:34:47-05:00 GitHub noreply@github.com 2025-10-31T08:34:47+01:00 vulkan: disable spirv-opt for rope shaders (#16872)
2976b0374d36609b0429dd6ce48807e2ad39a7c2 d2a2673dd1c86a01ab010e18b13b8bb959968c48 Masato Nakasaka masato.nakasaka@intel.com 2025-10-31T16:18:59+09:00 GitHub noreply@github.com 2025-10-31T08:18:59+01:00 vulkan: Fix crash when FP16 mul_mat accumulation is not supported (#16796)
d2a2673dd1c86a01ab010e18b13b8bb959968c48 13002a08960e51a76c4d696165b5d7638d2f2b99 Ruben Ortlam picard12@live.de 2025-10-31T08:14:49+01:00 GitHub noreply@github.com 2025-10-31T08:14:49+01:00 vulkan: fix shmem overrun in mmq id shader (#16873)
2e05afd22b3d77c7013b11eee88e88b17188f21a c20c0a5c0c44179f5267a262546624854b1203d1 Yunzhe Jia yunzhe@calculet.tech 2025-10-31T14:06:06+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-10-31T14:06:06+08:00 add pcie support
13002a08960e51a76c4d696165b5d7638d2f2b99 6eb208d17ea29bb60295d9a2b5e7122dfb8f4b55 l3utterfly gc.pthzfoldr@gmail.com 2025-10-31T12:46:31+08:00 GitHub noreply@github.com 2025-10-30T21:46:31-07:00 ggml-hexagon: respect input size when getting/setting tensor data (#16836)
6eb208d17ea29bb60295d9a2b5e7122dfb8f4b55 9984cbb61d12491d604484fb38b678fa15064061 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-31T00:34:27+01:00 GitHub noreply@github.com 2025-10-31T00:34:27+01:00 ci : enable free-disk-space on cuda docker build (#16877)
9984cbb61d12491d604484fb38b678fa15064061 ce18efeaf1234bd08f25bc08f88ef95cf5d9e51f lhez lih@qti.qualcomm.com 2025-10-30T16:00:20-07:00 GitHub noreply@github.com 2025-10-30T16:00:20-07:00 opencl: fix boundary handling for mul_mm (#16875)
ce18efeaf1234bd08f25bc08f88ef95cf5d9e51f 16724b5b6836a2d4b8936a5824d2ff27c52b4517 RodriMora bullerwins@gmail.com 2025-10-30T23:15:03+01:00 GitHub noreply@github.com 2025-10-30T23:15:03+01:00 convert : update transformers requirements (#16866)
16724b5b6836a2d4b8936a5824d2ff27c52b4517 b52edd25586fabb70f0c21b274473b307cf14499 chansikpark chansik.park@gmail.com 2025-10-30T14:22:23-04:00 GitHub noreply@github.com 2025-10-30T20:22:23+02:00 server : bump request URI max length to 32768 (#16862)
b52edd25586fabb70f0c21b274473b307cf14499 517b7170e1a4d733583c4b07c5b7a49acc05911c Georgi Gerganov ggerganov@gmail.com 2025-10-30T18:42:57+02:00 GitHub noreply@github.com 2025-10-30T18:42:57+02:00 server : remove n_past (#16818)
517b7170e1a4d733583c4b07c5b7a49acc05911c 835e918d8428f5119927d7150bf5a26176dedda0 Max Krasnyansky maxk@qti.qualcomm.com 2025-10-30T09:06:13-07:00 GitHub noreply@github.com 2025-10-30T09:06:13-07:00 cpu: introduce chunking for repack matmuls and enable matmul-id chunking on ARM64 (#16833)
835e918d8428f5119927d7150bf5a26176dedda0 d261223d24e97f2df50220e4a5b7f0adb69bba81 Shagun Bera 141054835+notV3NOM@users.noreply.github.com 2025-10-30T21:17:31+05:30 GitHub noreply@github.com 2025-10-30T17:47:31+02:00 common: fix typo in cli help text (#16864)
d261223d24e97f2df50220e4a5b7f0adb69bba81 dcca0d3ab840ebe9b2ccd4719033d408eeb758d7 JJJYmmm 92386084+JJJYmmm@users.noreply.github.com 2025-10-30T23:19:14+08:00 GitHub noreply@github.com 2025-10-30T16:19:14+01:00 model: add support for qwen3vl series (#16780)
dcca0d3ab840ebe9b2ccd4719033d408eeb758d7 bacddc049a00786df44e682262f6e298742bfbc3 Max Krasnyansky maxk@qti.qualcomm.com 2025-10-30T05:26:05-07:00 GitHub noreply@github.com 2025-10-30T14:26:05+02:00 cpu: introduce chunking for flash attention (#16829)
bacddc049a00786df44e682262f6e298742bfbc3 229bf686287d18f82c44e89888cc662145ecfdb4 Tianyue-Zhao zhaotianyue@outlook.com 2025-10-30T07:18:50-04:00 GitHub noreply@github.com 2025-10-30T12:18:50+01:00 model: Add support for CogVLM model (#15002)
229bf686287d18f82c44e89888cc662145ecfdb4 d7395115baf395b75a73a17b0b796e746e468da9 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-30T08:56:28+01:00 GitHub noreply@github.com 2025-10-30T08:56:28+01:00 cuda : fix argsort with 64k+ rows (#16849)
d7395115baf395b75a73a17b0b796e746e468da9 052df28b0e3ca0398e5928c61c7de40254317894 Jan Boon jan.boon@kaetemi.be 2025-10-30T14:30:58+08:00 GitHub noreply@github.com 2025-10-30T08:30:58+02:00 llama : use std::abs instead of abs (#16853)
052df28b0e3ca0398e5928c61c7de40254317894 8b11deea4663f29d3e042ce1056ba643264cd5f1 Jeff Bolz jbolz@nvidia.com 2025-10-30T01:27:41-05:00 GitHub noreply@github.com 2025-10-30T07:27:41+01:00 vulkan: Handle argsort with a large number of rows (#16851)
8b11deea4663f29d3e042ce1056ba643264cd5f1 b9ce94017729465895402cbcfffb51fa926c15e3 Oliver Simons osimons@nvidia.com 2025-10-30T04:34:15+01:00 GitHub noreply@github.com 2025-10-30T11:34:15+08:00 Hide latency of bias and gate-loading (#16847)
b9ce94017729465895402cbcfffb51fa926c15e3 3464bdac37027c5e9661621fc75ffcef3c19c6ef Jeff Bolz jbolz@nvidia.com 2025-10-29T15:13:10-05:00 GitHub noreply@github.com 2025-10-29T15:13:10-05:00 vulkan: Fuse rope+set_rows (#16769)
3464bdac37027c5e9661621fc75ffcef3c19c6ef e3af5563bd049141e036b50f843196db33d23e97 Xuan-Son Nguyen son@huggingface.co 2025-10-29T20:11:39+01:00 GitHub noreply@github.com 2025-10-29T20:11:39+01:00 llama: fix ASAN error with M-RoPE (#16848)
e3af5563bd049141e036b50f843196db33d23e97 10fcc41290e233788f5a4215314156e8e023eb92 Xuan-Son Nguyen son@huggingface.co 2025-10-29T18:09:18+01:00 GitHub noreply@github.com 2025-10-29T18:09:18+01:00 llama: store mrope data in KV cell (#16825)
10fcc41290e233788f5a4215314156e8e023eb92 bcf5bda6f5df559565d11d7c8e8295c1159a85ec Jeff Bolz jbolz@nvidia.com 2025-10-29T08:44:29-05:00 GitHub noreply@github.com 2025-10-29T14:44:29+01:00 vulkan: Update topk_moe fusion to handle gpt's late softmax (#16656)
bcf5bda6f5df559565d11d7c8e8295c1159a85ec 3eb2be1ca5f37480aeb16102970d9e65f43347fe Ruben Ortlam picard12@live.de 2025-10-29T14:39:03+01:00 GitHub noreply@github.com 2025-10-29T14:39:03+01:00 Vulkan MMQ Integer Dot Refactor and K-Quant support (#16536)
3eb2be1ca5f37480aeb16102970d9e65f43347fe e41bcce8f0b53032a1fed275cd253e931c041cf6 Max Krasnyansky maxk@qti.qualcomm.com 2025-10-29T06:29:12-07:00 GitHub noreply@github.com 2025-10-29T06:29:12-07:00 Hexagon Op queue & dispatch optimizations (#16820)
e41bcce8f0b53032a1fed275cd253e931c041cf6 144a4ce824b6bd0e48d62009d10cae1daf5308db Aman Gupta amangupta052@gmail.com 2025-10-29T21:11:53+08:00 GitHub noreply@github.com 2025-10-29T21:11:53+08:00 CUDA: use fastdiv in set-rows (#16834)
144a4ce824b6bd0e48d62009d10cae1daf5308db f549b0007dbdd683215820f7229ce180a12b191d Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-29T14:09:50+01:00 GitHub noreply@github.com 2025-10-29T14:09:50+01:00 vendor : sync minja (#16500)
f549b0007dbdd683215820f7229ce180a12b191d 9a3ea685b937c0f0cbfda2e50004ea54bf187512 Jeff Bolz jbolz@nvidia.com 2025-10-29T03:53:04-05:00 GitHub noreply@github.com 2025-10-29T09:53:04+01:00 vulkan: Call ggml_vk_buffer_write_2d from ggml_vk_buffer_copy (#16793)
9a3ea685b937c0f0cbfda2e50004ea54bf187512 338074c383c81366320d176d83b94b0a567ee0c2 Aman Gupta amangupta052@gmail.com 2025-10-29T15:55:06+08:00 GitHub noreply@github.com 2025-10-29T15:55:06+08:00 CUDA: Fix bug in topk-moe for gpt-oss (#16821)
338074c383c81366320d176d83b94b0a567ee0c2 851553ea6b24cb39fd5fd188b437d777cb411de8 YaelLogic y0548591250@gmail.com 2025-10-29T08:14:39+02:00 GitHub noreply@github.com 2025-10-29T14:14:39+08:00 sycl: add RMS_NORM_BACK operation support (#16808)
851553ea6b24cb39fd5fd188b437d777cb411de8 85a7d8677bf2200981e52f744a21d5267964ffcf YaelGitAccount 38328157276@mby.co.il 2025-10-28T21:10:28+02:00 GitHub noreply@github.com 2025-10-28T20:10:28+01:00 cuda: add SET operation support (#16804)
85a7d8677bf2200981e52f744a21d5267964ffcf a8ca18b4b815a2abdbecb958ee5f4c542d69aac7 Georgi Gerganov ggerganov@gmail.com 2025-10-28T20:19:44+02:00 GitHub noreply@github.com 2025-10-28T20:19:44+02:00 memory : remove KV cache size padding (#16812)
a8ca18b4b815a2abdbecb958ee5f4c542d69aac7 8284efc35c909217ee1b9a06903245d808ac2283 Georgi Gerganov ggerganov@gmail.com 2025-10-28T19:41:43+02:00 GitHub noreply@github.com 2025-10-28T19:41:43+02:00 llama-bench : clarify benchmarked parts of the computation (#16823)
8284efc35c909217ee1b9a06903245d808ac2283 1c1409e13169d0ced4b1b4d39fb5b268b7525091 l3utterfly gc.pthzfoldr@gmail.com 2025-10-28T23:16:20+08:00 GitHub noreply@github.com 2025-10-28T08:16:20-07:00 initialise buffer.device in ggml_hexagon_session (#16816)
1c1409e13169d0ced4b1b4d39fb5b268b7525091 7a0e900e3615fa46c074a7fdf900b47d3c0a1c7e Sam Malayek 12037535+SamMalayek@users.noreply.github.com 2025-10-28T03:51:41-07:00 GitHub noreply@github.com 2025-10-28T12:51:41+02:00 embedding: add raw option for --embd-output-format (#16541)
7a0e900e3615fa46c074a7fdf900b47d3c0a1c7e 280d97be9660e7a5feaa28a6e7a299bc73dd83fc Johannes Gäßler johannesg@5d6.de 2025-10-28T11:23:54+01:00 GitHub noreply@github.com 2025-10-28T11:23:54+01:00 llama: consistent ctx <-> buf order for KV cache (#16746)
280d97be9660e7a5feaa28a6e7a299bc73dd83fc 3479efd112b3910d2d008ad08fe4cb4895605903 Aldehir Rojas hello@alde.dev 2025-10-28T03:37:52-05:00 GitHub noreply@github.com 2025-10-28T09:37:52+01:00 grammar : support array references in json schema (#16792)
3479efd112b3910d2d008ad08fe4cb4895605903 463bbf20bfbe0da10ac58984d4d62bed5a49362a Chenguang Li 757486878@qq.com 2025-10-28T10:54:53+08:00 GitHub noreply@github.com 2025-10-28T10:54:53+08:00 CANN: Improve device ID handling and aclnnArange checks (#16752)
463bbf20bfbe0da10ac58984d4d62bed5a49362a ad8d36beffd791db10c94eb9e964afb891e3ca55 Aman Gupta amangupta052@gmail.com 2025-10-28T10:31:21+08:00 GitHub noreply@github.com 2025-10-28T10:31:21+08:00 CUDA: add unused vars to mmvf and mmvq (#16807)
ad8d36beffd791db10c94eb9e964afb891e3ca55 c053e18a66dd95dc340aa61317877c2a41d4e3cf tamarPal tamarp3385@gmail.com 2025-10-28T03:50:33+02:00 GitHub noreply@github.com 2025-10-28T09:50:33+08:00 sycl: add SSM_CONV operation support (#16800)
c053e18a66dd95dc340aa61317877c2a41d4e3cf e1ab0848037c9f9bfe68b3e1cee9ee375e1018a3 Yuri Khrustalev ykhrustalev@users.noreply.github.com 2025-10-27T18:54:01-04:00 GitHub noreply@github.com 2025-10-27T23:54:01+01:00 chat: Add LFM2 tool handling (#16763)
e1ab0848037c9f9bfe68b3e1cee9ee375e1018a3 5a4ff43e7dd049e35942bc3d12361dab2f155544 Xuan-Son Nguyen son@huggingface.co 2025-10-27T23:12:16+01:00 GitHub noreply@github.com 2025-10-27T23:12:16+01:00 mtmd : fix idefics3 preprocessing (#16806)
5a4ff43e7dd049e35942bc3d12361dab2f155544 10640e31aab0819f31c1e1f2d008b019ee737232 Diego Devesa slarengh@gmail.com 2025-10-27T13:51:28-07:00 GitHub noreply@github.com 2025-10-27T21:51:28+01:00 llama : disable pipeline parallelism if compute buffer allocation fails (#16748)
10640e31aab0819f31c1e1f2d008b019ee737232 80d28f104c0c3e61c11d6af073642b246a9fc19c Acly aclysia@gmail.com 2025-10-27T21:50:22+01:00 GitHub noreply@github.com 2025-10-27T21:50:22+01:00 ggml : fix interpolate with align-corners and ne=1 (#16700)
80d28f104c0c3e61c11d6af073642b246a9fc19c c55d53acec864f64afa1ba92972203dce1bf88f5 Johannes Gäßler johannesg@5d6.de 2025-10-27T21:39:49+01:00 GitHub noreply@github.com 2025-10-27T21:39:49+01:00 HIP: fix AMDGPU_TARGETS, update documentation (#16803)
c55d53acec864f64afa1ba92972203dce1bf88f5 945501f5ea4b8ca56b181ecb035e9ee3fb31f432 Xuan-Son Nguyen son@huggingface.co 2025-10-27T16:02:58+01:00 GitHub noreply@github.com 2025-10-27T16:02:58+01:00 model : add LightOnOCR-1B model (#16764)
945501f5ea4b8ca56b181ecb035e9ee3fb31f432 75cbdd3fce38ea12d50cd19e73a069aa5dbbd5fa Johannes Gäßler johannesg@5d6.de 2025-10-27T09:17:31+01:00 GitHub noreply@github.com 2025-10-27T09:17:31+01:00 llama: fix leaked buffers for mmap + split files (#16765)
75cbdd3fce38ea12d50cd19e73a069aa5dbbd5fa 2b9bd9bf4e759c05db629ec1c391dc8aeaa71887 Aman Gupta amangupta052@gmail.com 2025-10-27T09:25:10+08:00 GitHub noreply@github.com 2025-10-27T09:25:10+08:00 test-backend-ops: print failed tests at the end (#16785)
2b9bd9bf4e759c05db629ec1c391dc8aeaa71887 59fc1ec8e83b14354c1a3a8acf8c5c2cbf9af42f tamarPal tamarp3385@gmail.com 2025-10-27T03:20:24+02:00 GitHub noreply@github.com 2025-10-27T09:20:24+08:00 sycl: add ROLL operation support (#16665)
59fc1ec8e83b14354c1a3a8acf8c5c2cbf9af42f 75d33b9302f84a5b89f82205d2bcd8def5a64e0a shani-f s0556787439@gmail.com 2025-10-27T03:19:50+02:00 GitHub noreply@github.com 2025-10-27T09:19:50+08:00 sycl: add REPEAT_BACK operation support (#16734)
75d33b9302f84a5b89f82205d2bcd8def5a64e0a 3470a5c891dcc94363e492a3760af92b6b07241c Aman Gupta amangupta052@gmail.com 2025-10-27T09:06:16+08:00 GitHub noreply@github.com 2025-10-27T09:06:16+08:00 CUDA: support for weight clamp in top-k norm (#16702)
3470a5c891dcc94363e492a3760af92b6b07241c bd562fe4f7bd55625511d5f9d639c4fb1db1d440 Acly aclysia@gmail.com 2025-10-26T23:19:03+01:00 GitHub noreply@github.com 2025-10-26T23:19:03+01:00 ggml-alloc : make gallocr prefer chunks that allow memory reuse (#16788)
bd562fe4f7bd55625511d5f9d639c4fb1db1d440 bbac6a26b2bd7f7c1f0831cb1e7b52734c66673b Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-26T21:31:41+01:00 GitHub noreply@github.com 2025-10-26T21:31:41+01:00 cuda : use fast copy when src and dst are of different type and contiguous (#16789)
bbac6a26b2bd7f7c1f0831cb1e7b52734c66673b 73a48c9790d320476b3e5ef75bda09f2f8269e6e leejet leejet714@gmail.com 2025-10-27T02:13:31+08:00 GitHub noreply@github.com 2025-10-26T19:13:31+01:00 ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to support large batch (#16744)
73a48c9790d320476b3e5ef75bda09f2f8269e6e f696428ce8e4d16c17acbffeaa7feac3b0fb9061 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-26T17:21:23+01:00 GitHub noreply@github.com 2025-10-26T17:21:23+01:00 convert : enable expert group selection for all models with it (#16691)
f696428ce8e4d16c17acbffeaa7feac3b0fb9061 7cce4f8158f0c4c88d8dadd4c23d33938127b897 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-26T17:20:32+01:00 GitHub noreply@github.com 2025-10-26T17:20:32+01:00 graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (#16655)
7cce4f8158f0c4c88d8dadd4c23d33938127b897 8d8862829cd770fc9c0c7a726ed162de824ff5ea Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-26T16:08:52+01:00 GitHub noreply@github.com 2025-10-26T16:08:52+01:00 model : set res->t_embd in SmallThinker models (#16782)
8d8862829cd770fc9c0c7a726ed162de824ff5ea f77c13b91f4d25754b6a0b857f98a6bc922a0aa7 amirai21 89905406+amirai21@users.noreply.github.com 2025-10-26T14:01:20+02:00 GitHub noreply@github.com 2025-10-26T13:01:20+01:00 docs : add Jamba to Text-only models list (#16778)
f77c13b91f4d25754b6a0b857f98a6bc922a0aa7 3cfa9c3f125763305b4226bc032f1954f08990dc Aman Gupta amangupta052@gmail.com 2025-10-26T19:28:04+08:00 GitHub noreply@github.com 2025-10-26T19:28:04+08:00 CUDA: General GEMV fusion (#16715)
3cfa9c3f125763305b4226bc032f1954f08990dc 5d195f17bc60eacc15cfb929f9403cf29ccdf419 Gilad S. 7817232+giladgd@users.noreply.github.com 2025-10-26T06:37:38+02:00 GitHub noreply@github.com 2025-10-26T05:37:38+01:00 vulkan: deduplicate Microsoft Direct3D12 devices (#16689)
5d195f17bc60eacc15cfb929f9403cf29ccdf419 226f295f4dd92ad714533adc5497afed5fa88bb8 Galunid karolek1231456@gmail.com 2025-10-25T20:41:36+02:00 GitHub noreply@github.com 2025-10-25T20:41:36+02:00 convert : handle mmproj filename/path properly (#16760)
226f295f4dd92ad714533adc5497afed5fa88bb8 f90b4a8efe4466215c51155a5c22c1ae6207da23 Shunta Saito shunta.saito@gmail.com 2025-10-25T19:26:27+09:00 GitHub noreply@github.com 2025-10-25T12:26:27+02:00 model : set res->t_embd in PLaMo2 models (#16766)
f90b4a8efe4466215c51155a5c22c1ae6207da23 8423d019318b446640bb620e4ce80066d8530f05 Giuseppe Scrivano gscrivan@redhat.com 2025-10-25T10:59:54+02:00 GitHub noreply@github.com 2025-10-25T10:59:54+02:00 vulkan: delete dead code (#16732)
8423d019318b446640bb620e4ce80066d8530f05 5cca2542ac3f3f86831d32bce744d08fc2b353b0 Jeff Bolz jbolz@nvidia.com 2025-10-25T00:04:12-05:00 GitHub noreply@github.com 2025-10-25T07:04:12+02:00 vulkan: Optimize SSM_SCAN (#16645)
5cca2542ac3f3f86831d32bce744d08fc2b353b0 55945d2ef51b93821d4b6f4a9b994393344a90db compilade git@compilade.net 2025-10-24T20:52:00-04:00 GitHub noreply@github.com 2025-10-24T20:52:00-04:00 convert : avoid dequantizing mxfp4 for GPT-OSS (#16756)
55945d2ef51b93821d4b6f4a9b994393344a90db 0bcb40b48c6fc6f17ba9672625e526ab2574344b leejet leejet714@gmail.com 2025-10-25T03:39:37+08:00 GitHub noreply@github.com 2025-10-24T21:39:37+02:00 ggml: fix CUDA grid launch condition for large block_nums.y in binbcast (#16742)
0bcb40b48c6fc6f17ba9672625e526ab2574344b 69e9ff010309a1155d704cf9320bdb3aaf4160ca Aman Gupta amangupta052@gmail.com 2025-10-24T20:46:19+08:00 GitHub noreply@github.com 2025-10-24T20:46:19+08:00 CUDA: use CUB for arbitary size argsort (#16754)
69e9ff010309a1155d704cf9320bdb3aaf4160ca 5a91109a5d7dab5d7adc40bedb397ede99a705b1 Florian Badie florianbadie@odrling.xyz 2025-10-24T14:10:29+02:00 GitHub noreply@github.com 2025-10-24T14:10:29+02:00 webui: support q URL parameter (#16728)
5a91109a5d7dab5d7adc40bedb397ede99a705b1 f8f071faddf32ea09f4234edb6e809b380a9ee26 Daniel Bevenius daniel.bevenius@gmail.com 2025-10-24T12:02:02+02:00 GitHub noreply@github.com 2025-10-24T12:02:02+02:00 model-conversion : add trust_remote_code for orig model run [no ci] (#16751)
f8f071faddf32ea09f4234edb6e809b380a9ee26 0bf47a1dbba4d36f2aff4e8c34b06210ba34e688 compilade git@compilade.net 2025-10-23T16:31:41-04:00 GitHub noreply@github.com 2025-10-23T16:31:41-04:00 convert : handle pre-quantized models (#14810)
0bf47a1dbba4d36f2aff4e8c34b06210ba34e688 dd62dcfab97e420949519fd0eac9fca7bf97e635 Johannes Gäßler johannesg@5d6.de 2025-10-23T21:30:17+02:00 GitHub noreply@github.com 2025-10-23T21:30:17+02:00 server: add memory breakdown print (#16740)
dd62dcfab97e420949519fd0eac9fca7bf97e635 d0660f237a5c31771a3d6d1030ebe3e0c409ba92 Julien Denize 40604584+juliendenize@users.noreply.github.com 2025-10-23T15:54:46+02:00 GitHub noreply@github.com 2025-10-23T15:54:46+02:00 convert : Make mistral-common dependency optional (#16738)
d0660f237a5c31771a3d6d1030ebe3e0c409ba92 fe6a9882acf5c02f96624ed8f80144100d7006cb Xuan-Son Nguyen son@huggingface.co 2025-10-23T15:00:49+02:00 GitHub noreply@github.com 2025-10-23T15:00:49+02:00 mtmd-cli : allow using --jinja (#16718)
fe6a9882acf5c02f96624ed8f80144100d7006cb 061f0eff02dc9a82f7bd850db3bd70b8a0b5e87a Prajwal B Mehendarkar prajwal.b.mehendarkar@ibm.com 2025-10-23T17:07:31+05:30 GitHub noreply@github.com 2025-10-23T19:37:31+08:00 Manually link -lbsd to resolve flock symbol on AIX (#16610)
061f0eff02dc9a82f7bd850db3bd70b8a0b5e87a 8cf6b42d467d05fa7d9776d2bcc69974ecce6900 Aman Gupta amangupta052@gmail.com 2025-10-23T19:14:06+08:00 GitHub noreply@github.com 2025-10-23T19:14:06+08:00 ggml-cuda: use passed ops instead of hardcoded ops (#16712)
8cf6b42d467d05fa7d9776d2bcc69974ecce6900 9de9672adb0f4ca4e39483ac3ffed52b3f70a55d matteo matteo.serva@gmail.com 2025-10-23T11:32:24+02:00 GitHub noreply@github.com 2025-10-23T12:32:24+03:00 server : send partial stop string when <EOG> is reached (#15007)
9de9672adb0f4ca4e39483ac3ffed52b3f70a55d 63d2fc46e17a06be5b4b5823a5ada088317f1f0a Matthew Michel matthew.michel@intel.com 2025-10-22T20:05:15-05:00 GitHub noreply@github.com 2025-10-23T09:05:15+08:00 sycl: use async memory allocation to fix crashes during graph recording (#16644)
63d2fc46e17a06be5b4b5823a5ada088317f1f0a a2e0088d9242bd9e57f8b852b05a6e47843b5a45 Max Krasnyansky maxk@qti.qualcomm.com 2025-10-22T13:47:09-07:00 GitHub noreply@github.com 2025-10-22T13:47:09-07:00 Add experimental ggml-hexagon backend for the Hexagon NPU (#16547)
a2e0088d9242bd9e57f8b852b05a6e47843b5a45 9b9201f65a22c02cee8e300f58f480a588591227 Diego Devesa slarengh@gmail.com 2025-10-22T11:20:55-07:00 GitHub noreply@github.com 2025-10-22T20:20:55+02:00 Revert "ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_v…" (#16723)
9b9201f65a22c02cee8e300f58f480a588591227 19a5a3edfd306516cc419679d69d6435943b6816 Pascal admin@serveurperso.com 2025-10-22T16:58:23+02:00 GitHub noreply@github.com 2025-10-22T16:58:23+02:00 webui: introduce OpenAI-compatible model selector in JSON payload (#16562)
19a5a3edfd306516cc419679d69d6435943b6816 d8eaa26e4d9228df3aa46a930db60c8eaab67c1b sirus20x6 sirus20x6@users.noreply.github.com 2025-10-22T05:14:14-05:00 GitHub noreply@github.com 2025-10-22T12:14:14+02:00 ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_vec_set_f32 for faster fills (#16522)
d8eaa26e4d9228df3aa46a930db60c8eaab67c1b 9285325ce0174631c3cd6121d56084adc4ef2d8f Acly aclysia@gmail.com 2025-10-22T12:01:22+02:00 GitHub noreply@github.com 2025-10-22T12:01:22+02:00 tests : fix test-thread-safety when compiling with multiple backends (#16699)
9285325ce0174631c3cd6121d56084adc4ef2d8f 03792ad93609fc67e41041c6347d9aa14e5e0d74 Aman Gupta amangupta052@gmail.com 2025-10-22T12:33:08+08:00 GitHub noreply@github.com 2025-10-22T12:33:08+08:00 CUDA: fix bug in topk-moe softmax (#16711)
03792ad93609fc67e41041c6347d9aa14e5e0d74 51d1a8c997bd2629ef211a30208058ea87a30982 Aman Gupta amangupta052@gmail.com 2025-10-21T22:40:38+08:00 GitHub noreply@github.com 2025-10-21T22:40:38+08:00 CUDA: topk-moe: add optional parameter for gpt-oss (#16649)
51d1a8c997bd2629ef211a30208058ea87a30982 4926419c4d74a1cf724e7163d937eb72f36e7b26 Johannes Gäßler johannesg@5d6.de 2025-10-21T15:27:53+02:00 GitHub noreply@github.com 2025-10-21T15:27:53+02:00 CUDA: better error for FA kernel with 0 occupancy (#16643)
4926419c4d74a1cf724e7163d937eb72f36e7b26 6ea37f57391d27736c35cd3c20c1f990b7952b74 Aman Gupta amangupta052@gmail.com 2025-10-21T16:43:14+08:00 GitHub noreply@github.com 2025-10-21T16:43:14+08:00 ggml: add ggml_can_fuse_subgraph (#16662)
6ea37f57391d27736c35cd3c20c1f990b7952b74 fb349848f387f355450c3187556e71e6d32c145f lhez lih@qti.qualcomm.com 2025-10-20T22:26:17-07:00 GitHub noreply@github.com 2025-10-20T22:26:17-07:00 opencl: fix warnings and clean up profiling (#16688)
fb349848f387f355450c3187556e71e6d32c145f 6de8ed75196c7cd98c1f34bbf3a7452451ba8ac2 Jeff Bolz jbolz@nvidia.com 2025-10-20T22:16:08-05:00 GitHub noreply@github.com 2025-10-20T22:16:08-05:00 vulkan: Handle FA with all -inf mask values (#16447)
6de8ed75196c7cd98c1f34bbf3a7452451ba8ac2 84bf3c677857279037adf67cdcfd89eaa4ca9281 YehuditE y8703470@gmail.com 2025-10-21T01:21:12+03:00 GitHub noreply@github.com 2025-10-21T00:21:12+02:00 sycl : add PAD_REFLECT_D1 operator support (#16145)
84bf3c677857279037adf67cdcfd89eaa4ca9281 c9c1972e2c2cc6a771fcc145bfa138700179f961 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-20T21:38:20+02:00 GitHub noreply@github.com 2025-10-20T21:38:20+02:00 model : add BailingMoeV2 support (#16063)
c9c1972e2c2cc6a771fcc145bfa138700179f961 b617cfd2896edd592a36ebbc041817eb030a1005 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-20T19:49:02+02:00 GitHub noreply@github.com 2025-10-20T19:49:02+02:00 Handle legacy 'context' attachments (#16687)
b617cfd2896edd592a36ebbc041817eb030a1005 79068501fac9a74cca7129a8e5a8281b410a853e Diego Devesa slarengh@gmail.com 2025-10-20T05:53:50-07:00 GitHub noreply@github.com 2025-10-20T14:53:50+02:00 ggml-alloc : fix leak when reusing a tensor with a larger size (#16679)
79068501fac9a74cca7129a8e5a8281b410a853e 0e4a0cf2fae667d3efcf52f2f52398779d986b1d Aleksander Grygier aleksander.grygier@gmail.com 2025-10-20T14:21:12+02:00 GitHub noreply@github.com 2025-10-20T14:21:12+02:00 Prevent premature submission on IME input (#16673)
0e4a0cf2fae667d3efcf52f2f52398779d986b1d 13f2cfad4170c096c51a02c24a6a158cb47f1480 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-20T13:29:14+02:00 GitHub noreply@github.com 2025-10-20T13:29:14+02:00 Import/Export UX improvements (#16619)
13f2cfad4170c096c51a02c24a6a158cb47f1480 06332e28672356b964d6dfc2ba4657e20581cd43 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-20T12:41:13+02:00 GitHub noreply@github.com 2025-10-20T12:41:13+02:00 Enable per-conversation loading states to allow having parallel conversations (#16327)
06332e28672356b964d6dfc2ba4657e20581cd43 72d53e6c4decee8b339e49aed8cc0e234b9639dc takuya kodama otegami@clear-code.com 2025-10-20T16:27:09+08:00 GitHub noreply@github.com 2025-10-20T11:27:09+03:00 llama-batch: fix build fails with `-Werror=missing-braces` (#16614)
72d53e6c4decee8b339e49aed8cc0e234b9639dc 2330de7b847ca84eac766df372c604c26db72747 Ron Evans ron@hybridgroup.com 2025-10-20T10:20:04+02:00 GitHub noreply@github.com 2025-10-20T11:20:04+03:00 readme: update bindings (#16651)
2330de7b847ca84eac766df372c604c26db72747 7062dd8460685d6700ed7621e50a22c6f3400ca3 safranowith bsh155762@gmail.com 2025-10-20T11:08:32+03:00 GitHub noreply@github.com 2025-10-20T11:08:32+03:00 SYCL: Add support for FLOOR,CEIL,ROUND and TRUNC unary operators (#16613)
7062dd8460685d6700ed7621e50a22c6f3400ca3 0398752dd450dfabdd1b9e289f6364c2600f6ab5 takuya kodama otegami@clear-code.com 2025-10-20T15:44:21+08:00 GitHub noreply@github.com 2025-10-20T10:44:21+03:00 llama-context: only warn on pooling_type when user specified (#16674)
c20c0a5c0c44179f5267a262546624854b1203d1 62425b3a7e8b169a3bbe6ce8c7324513a7fcee8c Yunzhe Jia yunzhe@calculet.tech 2025-10-20T14:41:08+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-10-20T14:41:08+08:00 kv: turn on common-prefix mtmd: fix output size bug
0398752dd450dfabdd1b9e289f6364c2600f6ab5 4f73d0a95120687e2c527739f771330a5271259a Giuseppe Scrivano gscrivan@redhat.com 2025-10-19T23:54:31+02:00 GitHub noreply@github.com 2025-10-19T23:54:31+02:00 model : add Granite Hybrid types (#16635)
4f73d0a95120687e2c527739f771330a5271259a cec5edbcaec69bbf6d5851cabce4ac148be41701 Aaron Teo aaron.teo1@ibm.com 2025-10-20T05:06:39+08:00 GitHub noreply@github.com 2025-10-19T23:06:39+02:00 ci : fix binaries release failure for s390x (binaries may not work yet) (#16664)
cec5edbcaec69bbf6d5851cabce4ac148be41701 fcb235b46618921cbd826acd49b553b5302233aa Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-19T14:03:25+02:00 GitHub noreply@github.com 2025-10-19T14:03:25+02:00 ci : avoid manual updates of docs/ops.md (#16663)
fcb235b46618921cbd826acd49b553b5302233aa 55754bebd5d570960cde9c0ba991a0b2991f6b1e Aaron Teo aaron.teo1@ibm.com 2025-10-19T18:37:47+08:00 GitHub noreply@github.com 2025-10-19T18:37:47+08:00 ci: include s390x release binaries (#16648)
55754bebd5d570960cde9c0ba991a0b2991f6b1e ee09828cb057460b369576410601a3a09279e23c Aman Gupta amangupta052@gmail.com 2025-10-19T15:37:12+08:00 GitHub noreply@github.com 2025-10-19T10:37:12+03:00 CODEOWNERS: update for ggml-cuda/mmf (#16660)
ee09828cb057460b369576410601a3a09279e23c e56abd2098dd2e2b0804691b93c13b48ae421627 Johannes Gäßler johannesg@5d6.de 2025-10-18T14:47:32+02:00 GitHub noreply@github.com 2025-10-18T14:47:32+02:00 HIP: fix GPU_TARGETS (#16642)
e56abd2098dd2e2b0804691b93c13b48ae421627 38355c6c8e43204e11a22daa7483082c0ff01e71 Jeff Bolz jbolz@nvidia.com 2025-10-18T05:22:57-05:00 GitHub noreply@github.com 2025-10-18T12:22:57+02:00 vulkan: Implement topk_moe fused shader, ported from CUDA (#16641)
38355c6c8e43204e11a22daa7483082c0ff01e71 81387858f1fbcc1acedbd308486e1016618ca8f8 Aman Gupta amangupta052@gmail.com 2025-10-18T17:52:53+08:00 GitHub noreply@github.com 2025-10-18T11:52:53+02:00 CUDA: use registers instead of smem in topk-moe (#16647)
81387858f1fbcc1acedbd308486e1016618ca8f8 66b0dbcb2d462e7b70ba5a69ee8c3899ac2efb1c Shawn Gu shawngu@qti.qualcomm.com 2025-10-17T17:55:32-07:00 GitHub noreply@github.com 2025-10-17T17:55:32-07:00 opencl: transposed gemm/gemv moe kernel with mxfp4,f32 (#16602)
66b0dbcb2d462e7b70ba5a69ee8c3899ac2efb1c 41386cf365d894134ee0813d15e2f5d76f6a4d8e Johannes Gäßler johannesg@5d6.de 2025-10-17T17:41:09+02:00 GitHub noreply@github.com 2025-10-17T17:41:09+02:00 llama-model: fix insonsistent ctxs <-> bufs order (#16581)
41386cf365d894134ee0813d15e2f5d76f6a4d8e 3d4e86bbeb15f487d6da6174ba6191b7c212cc25 Radoslav Gerganov rgerganov@gmail.com 2025-10-17T18:02:52+03:00 GitHub noreply@github.com 2025-10-17T18:02:52+03:00 rpc : report actual free memory (#16616)
3d4e86bbeb15f487d6da6174ba6191b7c212cc25 342c728d031d50673feded797520a44127d73379 Giuseppe Scrivano gscrivan@redhat.com 2025-10-17T14:23:47+02:00 GitHub noreply@github.com 2025-10-17T14:23:47+02:00 vulkan: Add State Space Model (SSM) Operations Support (#16463)
342c728d031d50673feded797520a44127d73379 ababae7e1ec3e9cfdab0322ee55ea3389e82a4d5 muggle-stack promuggle@qq.com 2025-10-17T18:01:23+08:00 GitHub noreply@github.com 2025-10-17T13:01:23+03:00 ggml : fix SpaceMit IME array out-of-bounds in task assignment (#16629)
ababae7e1ec3e9cfdab0322ee55ea3389e82a4d5 b19491599d4d42c606601d75e95c1d1de3291f8e Pascal admin@serveurperso.com 2025-10-17T10:35:03+02:00 GitHub noreply@github.com 2025-10-17T10:35:03+02:00 webui: reorganize settings layout (#16607)
b19491599d4d42c606601d75e95c1d1de3291f8e 9ad4f1931ee0f3b41d9355245ef744786aaae0aa Jeff Bolz jbolz@nvidia.com 2025-10-17T02:31:04-05:00 GitHub noreply@github.com 2025-10-17T09:31:04+02:00 vulkan: fix debug build (add_rms_len/data not found) (#16624)
9ad4f1931ee0f3b41d9355245ef744786aaae0aa 79967ec596c0dacfd2251b085a57e79df292b1cc Ilia Ilmer iliailmer@users.noreply.github.com 2025-10-17T02:33:58-04:00 GitHub noreply@github.com 2025-10-17T09:33:58+03:00 metal : add `CONV_TRANSPOSE_2D` (#16542)
79967ec596c0dacfd2251b085a57e79df292b1cc ceff6bb253dd306f5404d7ccb3f11fadafe71b52 Olivier Chafik olivier.chafik@gmail.com 2025-10-17T06:59:31+01:00 GitHub noreply@github.com 2025-10-17T08:59:31+03:00 grammar : use int64_t to avoid int overflows in int schema to grammar conversion logic (#16626)
ceff6bb253dd306f5404d7ccb3f11fadafe71b52 1bb4f43380944e94c9a86e305789ba103f5e62bd GittyBurstein g0534163997@gmail.com 2025-10-17T05:36:40+03:00 GitHub noreply@github.com 2025-10-17T10:36:40+08:00 SYCL SET operator optimized for F32 tensors (#16350)
1bb4f43380944e94c9a86e305789ba103f5e62bd 683fa6ba4ed3a23b939d3e11e6dc860bd47a0ccf Xuan-Son Nguyen son@huggingface.co 2025-10-16T19:00:31+02:00 GitHub noreply@github.com 2025-10-16T19:00:31+02:00 mtmd : support home-cooked Mistral Small Omni (#14928)
683fa6ba4ed3a23b939d3e11e6dc860bd47a0ccf b22572e97dc51757d3ebe917a5a283385010ec68 Pascal admin@serveurperso.com 2025-10-16T16:28:41+02:00 GitHub noreply@github.com 2025-10-16T16:28:41+02:00 fix: added a normalization step for MathJax-style \[\] and \(\) delimiters (#16599)
b22572e97dc51757d3ebe917a5a283385010ec68 7a50cf388a530127cbbdd2a507ef81a451c9d819 GittyBurstein g0534163997@gmail.com 2025-10-16T16:26:21+03:00 GitHub noreply@github.com 2025-10-16T15:26:21+02:00 sycl : add ARANGE operator (#16362)
62425b3a7e8b169a3bbe6ce8c7324513a7fcee8c 76a63a2998286c1649de2496bccb7d8a46549895 Yunzhe Jia yunzhe@calculet.tech 2025-10-16T17:34:39+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-10-16T17:34:39+08:00 kv: let seq_rm free whole seq since current hw-op cannot support common-prefix
7a50cf388a530127cbbdd2a507ef81a451c9d819 6f5d924637a15abedb111cbbffd7da5f31c81855 Chenguang Li 757486878@qq.com 2025-10-16T16:41:11+08:00 GitHub noreply@github.com 2025-10-16T16:41:11+08:00 CANN: format code using .clang-format (#15863)
76a63a2998286c1649de2496bccb7d8a46549895 b67217078aa748b06e1b2269d88dd18c0aca2c26 Yunzhe Jia yunzhe@calculet.tech 2025-10-16T16:31:15+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-10-16T16:31:15+08:00 kv: fix common-prefix bug by implementing llama_memory_seq_rm
6f5d924637a15abedb111cbbffd7da5f31c81855 adc9b60f190c1016a09f439862fa1cbb302262ac takasurazeem takasurazeem@gmail.com 2025-10-16T01:11:33-04:00 GitHub noreply@github.com 2025-10-16T08:11:33+03:00 common : Update the docs on -t --threads (#16236)
adc9b60f190c1016a09f439862fa1cbb302262ac ee50ee1eadff58777ae746827b04de7ba0befc55 takuya kodama a.s.takuya1026@gmail.com 2025-10-16T13:10:32+08:00 GitHub noreply@github.com 2025-10-16T08:10:32+03:00 ggml-cpu: replace putenv with setenv for const-correctness (#16573)
ee50ee1eadff58777ae746827b04de7ba0befc55 7adc79c03234de9a20661fd6dbf2d02c32ca7acb yael-works 106673277+yael-works@users.noreply.github.com 2025-10-16T07:21:28+03:00 GitHub noreply@github.com 2025-10-16T12:21:28+08:00 SYCL: Add GGML_OP_MEAN operator support (#16009)
7adc79c03234de9a20661fd6dbf2d02c32ca7acb 466c1911ab736f0b7366127edee99f8ee5687417 Aleksei Nikiforov 103434461+AlekseiNikiforovIBM@users.noreply.github.com 2025-10-15T22:43:08+02:00 GitHub noreply@github.com 2025-10-15T22:43:08+02:00 gguf-py : add support for endian conversion of BF16 data (#16594)
466c1911ab736f0b7366127edee99f8ee5687417 0cb7a0683b0529172472d74d21f05470a607f297 safranowith bsh155762@gmail.com 2025-10-15T22:24:51+03:00 GitHub noreply@github.com 2025-10-15T21:24:51+02:00 cpu : add FLOOR, CEIL, ROUND and TRUNC unary operators (#16083)
0cb7a0683b0529172472d74d21f05470a607f297 d93f8439b08c4f35e13a41a7366901fdbe770fc8 lhez lih@qti.qualcomm.com 2025-10-15T10:51:04-07:00 GitHub noreply@github.com 2025-10-15T10:51:04-07:00 opencl: add q8_0 mm support (#16469)
d93f8439b08c4f35e13a41a7366901fdbe770fc8 f9fb33f2630b4b4ba9081ce9c0c921f8cd8ba4eb lhez lih@qti.qualcomm.com 2025-10-15T10:48:28-07:00 GitHub noreply@github.com 2025-10-15T10:48:28-07:00 opencl: fix FA for f32 (#16584)
f9fb33f2630b4b4ba9081ce9c0c921f8cd8ba4eb f4ce81c45e7bd910e36bf44c253fc5255c49b1e4 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-15T16:22:20+02:00 GitHub noreply@github.com 2025-10-15T16:22:20+02:00 Add server-driven parameter defaults and syncing (#16515)
f4ce81c45e7bd910e36bf44c253fc5255c49b1e4 17304cbcc1dd24de7741cbe57925d58e90a98ac1 Sam/Samuel 57896620+cern1710@users.noreply.github.com 2025-10-15T23:05:56+09:00 GitHub noreply@github.com 2025-10-15T17:05:56+03:00 metal: optimise `GGML_OP_SUM` (#16559)
17304cbcc1dd24de7741cbe57925d58e90a98ac1 3e3cb19f6449f9168a128eaeae01f8f41b049acc Georgi Gerganov ggerganov@gmail.com 2025-10-15T16:53:12+03:00 GitHub noreply@github.com 2025-10-15T16:53:12+03:00 server : fix img token logs (#16595)
3e3cb19f6449f9168a128eaeae01f8f41b049acc 5acd455460f457942d8dd02e3dd9b1eebfce99fe Xuan-Son Nguyen son@huggingface.co 2025-10-15T14:48:08+02:00 GitHub noreply@github.com 2025-10-15T14:48:08+02:00 llama-quant: add support for mmproj (#16592)
5acd455460f457942d8dd02e3dd9b1eebfce99fe 554fd578a5ed78a10f371f5850c6a69aa83df15a Julius Tischbein ju.tischbein@gmail.com 2025-10-15T13:54:15+02:00 GitHub noreply@github.com 2025-10-15T14:54:15+03:00 CUDA: Changing the CUDA scheduling strategy to spin (#16585)
554fd578a5ed78a10f371f5850c6a69aa83df15a fa882fd2b1bcb663de23af06fdc391489d05b007 Georgi Gerganov ggerganov@gmail.com 2025-10-15T12:51:27+03:00 GitHub noreply@github.com 2025-10-15T11:51:27+02:00 server : fix mtmd checkpoints (#16591)
fa882fd2b1bcb663de23af06fdc391489d05b007 ffa059034c1e41f3e58363ef3c46d2fcf854bc4d Georgi Gerganov ggerganov@gmail.com 2025-10-14T20:33:05+03:00 GitHub noreply@github.com 2025-10-14T20:33:05+03:00 metal : avoid using Metal's gpuAddress property (#16576)
ffa059034c1e41f3e58363ef3c46d2fcf854bc4d 120bf7046d85a893f064c12abe58bfeebd735f84 SavicStefan 50296686+SavicStefan@users.noreply.github.com 2025-10-14T19:18:05+02:00 GitHub noreply@github.com 2025-10-14T19:18:05+02:00 vulkan: Add ACC_TYPE_VEC2 implementation (#16203)
120bf7046d85a893f064c12abe58bfeebd735f84 4258e0cfe72c72ad2787be1e9f93253e89219455 Aman Gupta amangupta052@gmail.com 2025-10-14T22:48:08+08:00 GitHub noreply@github.com 2025-10-14T07:48:08-07:00 CUDA + openCL: fix bug in accessing rms_norm->src while doing fusion (#16577)
4258e0cfe72c72ad2787be1e9f93253e89219455 7ea15bb64c81e3813eb0babf9a57e1bc5697f569 Jeff Bolz jbolz@nvidia.com 2025-10-14T08:53:37-05:00 GitHub noreply@github.com 2025-10-14T15:53:37+02:00 vulkan: Support FA with K/V in F32 (#16543)
7ea15bb64c81e3813eb0babf9a57e1bc5697f569 9c7185dd28416cf67f5e3b268381f311b5e3da56 Jeff Bolz jbolz@nvidia.com 2025-10-14T07:51:36-05:00 GitHub noreply@github.com 2025-10-14T14:51:36+02:00 vulkan: Improve build time for MSVC (#16545)
9c7185dd28416cf67f5e3b268381f311b5e3da56 1ee9d0b415cdf5240418c110a18b419f4002b154 Johannes Gäßler johannesg@5d6.de 2025-10-14T14:22:47+02:00 GitHub noreply@github.com 2025-10-14T14:22:47+02:00 CUDA: enable FA for FP32 KV cache (#16546)
1ee9d0b415cdf5240418c110a18b419f4002b154 48e2fa9fb7c2de1e53808fdb65ec33f916020fc4 Aman Gupta amangupta052@gmail.com 2025-10-14T19:16:21+08:00 GitHub noreply@github.com 2025-10-14T13:16:21+02:00 CUDA: use fastdiv + ggml_cuda_mad for mmvf (#16557)
48e2fa9fb7c2de1e53808fdb65ec33f916020fc4 5b6913c47b6bc71a6f927805a45387d5657d8b89 Aman Gupta amangupta052@gmail.com 2025-10-14T19:15:15+08:00 GitHub noreply@github.com 2025-10-14T13:15:15+02:00 CUDA: add fp kernel for larger batch size MoE (#16512)
5b6913c47b6bc71a6f927805a45387d5657d8b89 bc07349a7f87ba6eb31ed4b0ea9d9a7352185213 Anav Prasad anavp@nvidia.com 2025-10-14T09:53:49Z GitHub noreply@github.com 2025-10-14T11:53:49+02:00 cuda : remove legacy copy-op pointer indirection code (#16485)
bc07349a7f87ba6eb31ed4b0ea9d9a7352185213 e60f241eacec42d3bd7c9edd37d236ebf35132a8 Georgi Gerganov ggerganov@gmail.com 2025-10-14T08:48:50+03:00 GitHub noreply@github.com 2025-10-14T08:48:50+03:00 server : dynamic token limit for prompt cache (#16560)
e60f241eacec42d3bd7c9edd37d236ebf35132a8 e38b7c6e9e4453e3b3e96d76e38bc2ccb6bce458 Georgi Gerganov ggerganov@gmail.com 2025-10-13T23:07:57+03:00 GitHub noreply@github.com 2025-10-13T23:07:57+03:00 metal : FA support F32 K and V and head size = 32 (#16531)
e38b7c6e9e4453e3b3e96d76e38bc2ccb6bce458 5016b7286240d29f8f640039989b84ea3a854344 Georgi Gerganov ggerganov@gmail.com 2025-10-13T22:42:37+03:00 GitHub noreply@github.com 2025-10-13T22:42:37+03:00 graph : support cacheless embeddings with FA and iSWA (#16528)
5016b7286240d29f8f640039989b84ea3a854344 7049736b2dd9011bf819e298b844ebbc4b5afdc9 lhez lih@qti.qualcomm.com 2025-10-13T11:50:37-07:00 GitHub noreply@github.com 2025-10-13T11:50:37-07:00 opencl: fix build targeting CL 2 (#16554)
7049736b2dd9011bf819e298b844ebbc4b5afdc9 01d2bdc2bc61b676706830305b286b08b9885a41 Johannes Gäßler johannesg@5d6.de 2025-10-13T16:29:45+02:00 GitHub noreply@github.com 2025-10-13T17:29:45+03:00 CUDA: fix numerical issues in tile FA kernel (#16540)
01d2bdc2bc61b676706830305b286b08b9885a41 56fc38b9655fbe1869d8bd6cfb269418196cea69 Jie Fu (傅杰) jiefu@tencent.com 2025-10-13T20:48:47+08:00 GitHub noreply@github.com 2025-10-13T15:48:47+03:00 ggml : fix build broken with -march=armv9-a on MacOS (#16520)
56fc38b9655fbe1869d8bd6cfb269418196cea69 1fb9504eb744969a990bfe4cfcf1d3d7a479541c Chenguang Li 757486878@qq.com 2025-10-13T17:01:24+08:00 GitHub noreply@github.com 2025-10-13T17:01:24+08:00 CANN: fix CPU memory leak in CANN backend (#16549)
1fb9504eb744969a990bfe4cfcf1d3d7a479541c 3f750f8d760ab5a61491e6a9409072dfeee4b4d7 Pascal admin@serveurperso.com 2025-10-13T10:55:32+02:00 GitHub noreply@github.com 2025-10-13T10:55:32+02:00 fix: add remark plugin to render raw HTML as literal text (#16505)
b67217078aa748b06e1b2269d88dd18c0aca2c26 98c8269abefdfa8fde47dfb2769ceebc8b704e86 Yunzhe Jia yunzhe@calculet.tech 2025-10-13T11:48:14+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-10-13T16:52:48+08:00 add different path for unified and seperate compilation
3f750f8d760ab5a61491e6a9409072dfeee4b4d7 c515fc577166042234241c6bd0da9b08dcbe2bb9 Sam/Samuel 57896620+cern1710@users.noreply.github.com 2025-10-13T16:25:02+08:00 GitHub noreply@github.com 2025-10-13T11:25:02+03:00 metal: add support for opt_step_sgd (#16539)
c515fc577166042234241c6bd0da9b08dcbe2bb9 f9bc66c3ebcfddb5f09e4b21253623caeb8e414a Georgi Gerganov ggerganov@gmail.com 2025-10-13T11:22:27+03:00 GitHub noreply@github.com 2025-10-13T11:22:27+03:00 ggml : fix scalar path for computing norm (#16558)
f9bc66c3ebcfddb5f09e4b21253623caeb8e414a a31cf36ad946a13b3a646bf0dadf2a481e89f944 hipudding huafengchun@gmail.com 2025-10-13T08:52:22+08:00 GitHub noreply@github.com 2025-10-13T08:52:22+08:00 CANN: Update several operators to support FP16 data format (#16251)
a31cf36ad946a13b3a646bf0dadf2a481e89f944 81d54bbfd599811b354c39f04550888168be7780 Sam/Samuel 57896620+cern1710@users.noreply.github.com 2025-10-13T02:43:14+08:00 GitHub noreply@github.com 2025-10-12T21:43:14+03:00 metal : add opt_step_adamw and op_sum (#16529)
81d54bbfd599811b354c39f04550888168be7780 c7be9febcbafa9af7d1b9443f86475c59c9c5f87 Pascal admin@serveurperso.com 2025-10-12T18:06:41+02:00 GitHub noreply@github.com 2025-10-12T18:06:41+02:00 webui: remove client-side context pre-check and rely on backend for limits (#16506)
c7be9febcbafa9af7d1b9443f86475c59c9c5f87 8415f61e23d04427cd0d912fbb9d33b85f849456 Neo Zhang Jianyu jianyu.zhang@intel.com 2025-10-12T21:53:35+08:00 GitHub noreply@github.com 2025-10-12T21:53:35+08:00 [SYCL] fix UT fault cases: count-equal, argsort, pad OPs (#16521)
8415f61e23d04427cd0d912fbb9d33b85f849456 2c301e91abb92d03c1a682b4b540ba835562a74b Mathieu Baudier mbaudier@argeo.org 2025-10-12T15:48:03+02:00 GitHub noreply@github.com 2025-10-12T15:48:03+02:00 ci : add Vulkan on Ubuntu with default packages build (#16532)
2c301e91abb92d03c1a682b4b540ba835562a74b 4b2dae383df708e2afc49c4859a81cd074f5ac10 Aldehir Rojas hello@alde.dev 2025-10-12T08:18:47-05:00 GitHub noreply@github.com 2025-10-12T16:18:47+03:00 common : handle unicode during partial json parsing (#16526)
4b2dae383df708e2afc49c4859a81cd074f5ac10 41aac5c69b5fb281bc1f486afb053f78101bb39e Georgi Gerganov ggerganov@gmail.com 2025-10-12T09:29:13+03:00 GitHub noreply@github.com 2025-10-12T09:29:13+03:00 common : update presets (#16504)
41aac5c69b5fb281bc1f486afb053f78101bb39e a2fba89a426ff8005d303c73f0436e7e67368b70 sirus20x6 sirus20x6@users.noreply.github.com 2025-10-12T00:25:37-05:00 GitHub noreply@github.com 2025-10-12T08:25:37+03:00 ggml : Fix FP16 ELU positive branch (#16519)
a2fba89a426ff8005d303c73f0436e7e67368b70 20cc625edc2264aae2779e71bef1593e6a4e8c43 Daniel Bevenius daniel.bevenius@gmail.com 2025-10-12T07:19:06+02:00 GitHub noreply@github.com 2025-10-12T07:19:06+02:00 hparams : add check for layer index in is_recurrent (#16511)
20cc625edc2264aae2779e71bef1593e6a4e8c43 11f0af5504252e453d57406a935480c909e3ff37 sirus20x6 sirus20x6@users.noreply.github.com 2025-10-12T00:15:00-05:00 GitHub noreply@github.com 2025-10-12T08:15:00+03:00 ggml: Correct SVE implementation in ggml_vec_dot_f16_unroll (#16518)
11f0af5504252e453d57406a935480c909e3ff37 a3cb04744fb5c591985f53b749fef5407d07a145 Johannes Gäßler johannesg@5d6.de 2025-10-11T20:54:32+02:00 GitHub noreply@github.com 2025-10-11T20:54:32+02:00 CUDA: faster tile FA, add oob checks, more HSs (#16492)
a3cb04744fb5c591985f53b749fef5407d07a145 4a8fbe0a5eb75f339782ebcf29c17848122184d3 Georgi Gerganov ggerganov@gmail.com 2025-10-11T16:54:10+03:00 GitHub noreply@github.com 2025-10-11T16:54:10+03:00 metal : fix mul-mm condition + fix mul-mv permuted kernels (#16494)
4a8fbe0a5eb75f339782ebcf29c17848122184d3 31d0ff1869aa2ea31f4e96d5877d0343e9a2171b Pascal admin@serveurperso.com 2025-10-11T15:50:49+02:00 GitHub noreply@github.com 2025-10-11T15:50:49+02:00 feat: render user content as markdown option (#16358)
31d0ff1869aa2ea31f4e96d5877d0343e9a2171b 97870e64975b26c5e06a3540a8dc0ff601351e86 Yann Follet 131855179+YannFollet@users.noreply.github.com 2025-10-11T21:39:04+08:00 GitHub noreply@github.com 2025-10-11T16:39:04+03:00 server / ranking : add sorting and management of top_n (#16403)
97870e64975b26c5e06a3540a8dc0ff601351e86 477a66b03501cf3bd067f8968b77ca4d053ff1bd Diego Devesa slarengh@gmail.com 2025-10-11T04:02:26-07:00 GitHub noreply@github.com 2025-10-11T13:02:26+02:00 cuda : avoid initializing unused devices (#16510)
477a66b03501cf3bd067f8968b77ca4d053ff1bd e60f01d941bc5b7fae62dd57fee4cec76ec0ea6e amirai21 89905406+amirai21@users.noreply.github.com 2025-10-11T11:33:41+03:00 GitHub noreply@github.com 2025-10-11T10:33:41+02:00 convert : correctly handle LLaMA tokenizer for Jamba (#16470)
e60f01d941bc5b7fae62dd57fee4cec76ec0ea6e 81086cd6a3ca1252f0dc0f938171648399179c53 Georgi Gerganov ggerganov@gmail.com 2025-10-10T22:15:05+03:00 GitHub noreply@github.com 2025-10-10T22:15:05+03:00 server : fix division by zero when reporting stats (#16501)
81086cd6a3ca1252f0dc0f938171648399179c53 68ee98ae181a5c83a5cc6261daeee69a1f588c15 Georgi Gerganov ggerganov@gmail.com 2025-10-10T17:17:31+03:00 GitHub noreply@github.com 2025-10-10T17:17:31+03:00 vocab : mark EOT token for Granite models (#16499)
68ee98ae181a5c83a5cc6261daeee69a1f588c15 cdb6da468cc33323955a523738d2e1675aeb5e9a Radoslav Gerganov rgerganov@gmail.com 2025-10-10T17:11:07+03:00 GitHub noreply@github.com 2025-10-10T16:11:07+02:00 server : return HTTP 400 if prompt exceeds context length (#16486)
cdb6da468cc33323955a523738d2e1675aeb5e9a 6d69ab3f262eca3f0c6ad0ce075d2c80b9924d0e Radoslav Gerganov rgerganov@gmail.com 2025-10-10T13:22:27+03:00 GitHub noreply@github.com 2025-10-10T13:22:27+03:00 server : log requests to /v1/completions (#16495)
6d69ab3f262eca3f0c6ad0ce075d2c80b9924d0e 1faa13a1187051af66b0fd9f0d6effe4c77f0b3e Prajwal B Mehendarkar prajwal.b.mehendarkar@ibm.com 2025-10-10T13:45:46+05:30 GitHub noreply@github.com 2025-10-10T11:15:46+03:00 cmake : Dont define XOPENSOURCE on AIX (#16481)
1faa13a1187051af66b0fd9f0d6effe4c77f0b3e 1deee0f8d494981c32597dca8b5f8696d399b0f2 Pascal admin@serveurperso.com 2025-10-09T22:54:57+02:00 GitHub noreply@github.com 2025-10-09T22:54:57+02:00 webui: updated the chat service to only include max_tokens in the req… (#16489)
1deee0f8d494981c32597dca8b5f8696d399b0f2 d00cbea63c671cd85a57adaa50abf60b3b87d86f duduta simona.gherman@gmail.com 2025-10-09T22:11:15+03:00 GitHub noreply@github.com 2025-10-09T21:11:15+02:00 cpu : optimize the ggml NORM operation (#15953)
d00cbea63c671cd85a57adaa50abf60b3b87d86f 8328fd4bae76fc027f8ca0e9a05acd3788dabe3f Georgi Gerganov ggerganov@gmail.com 2025-10-09T18:54:51+03:00 GitHub noreply@github.com 2025-10-09T18:54:51+03:00 server : host-memory prompt caching (#16391)
8328fd4bae76fc027f8ca0e9a05acd3788dabe3f 56b4795842d852152222ca7a2d4304008facf1b9 Pascal admin@serveurperso.com 2025-10-09T17:36:29+02:00 GitHub noreply@github.com 2025-10-09T17:36:29+02:00 No markdown in cot (#16483)
56b4795842d852152222ca7a2d4304008facf1b9 2c0d875ae6c6043fbbafd2831e160955e9ca6af1 Daniel Bevenius daniel.bevenius@gmail.com 2025-10-09T14:35:22+02:00 GitHub noreply@github.com 2025-10-09T14:35:22+02:00 model-conversion : add support for SentenceTransformers (#16387)
2c0d875ae6c6043fbbafd2831e160955e9ca6af1 aa4711d369b0d4adb6802b98ad6ea362f838710e sudhiarm sudhi.sathyavathy@arm.com 2025-10-09T09:13:18+01:00 GitHub noreply@github.com 2025-10-09T10:13:18+02:00 ci: add ARM64 Kleidiai build and test support (#16462)
aa4711d369b0d4adb6802b98ad6ea362f838710e d80d6d2400b8faa80654c723cb5bdf9fc8f4db06 Chenguang Li 757486878@qq.com 2025-10-09T15:50:25+08:00 GitHub noreply@github.com 2025-10-09T15:50:25+08:00 CANN: Improve ACL graph matching (#16166)
d80d6d2400b8faa80654c723cb5bdf9fc8f4db06 b2602137557b2b28a39e03612717d85ead9a6f5a Charles Xu charles.xu@arm.com 2025-10-09T09:29:17+02:00 GitHub noreply@github.com 2025-10-09T10:29:17+03:00 kleidiai: kernel interface refactoring (#16460)
b2602137557b2b28a39e03612717d85ead9a6f5a e08db4259521de493b7aeb49dadf29ebd1ee966a Neo Zhang Jianyu jianyu.zhang@intel.com 2025-10-09T15:25:11+08:00 GitHub noreply@github.com 2025-10-09T10:25:11+03:00 [SYCL] refactor soft_max, add soft_max_back (#16472)
e08db4259521de493b7aeb49dadf29ebd1ee966a 12bbc3fa50b6df03318a4451c9a2210200a0b28d Saba Fallah 10401143+sfallah@users.noreply.github.com 2025-10-09T08:39:18+02:00 GitHub noreply@github.com 2025-10-09T09:39:18+03:00 model: EmbeddingGemma Adding Support for SentenceTransformers Dense Modules (#16367)
12bbc3fa50b6df03318a4451c9a2210200a0b28d 9d0882840e6c3fb62965d03af0e22880ea90e012 Pascal admin@serveurperso.com 2025-10-08T22:18:41+02:00 GitHub noreply@github.com 2025-10-08T23:18:41+03:00 refactor: centralize CoT parsing in backend for streaming mode (#16394)
9d0882840e6c3fb62965d03af0e22880ea90e012 d2ee056e1df0fdf646270fb5621e9f92084b59a7 ai-fonsi length-amiss-7k@icloud.com 2025-10-08T20:21:46+02:00 GitHub noreply@github.com 2025-10-08T20:21:46+02:00 Disable CUDA host buffers on integrated GPUs (#16308)
d2ee056e1df0fdf646270fb5621e9f92084b59a7 b2c08c9ec4cd59ed88c1ebdd94f109af5ccc978e issixx 46835150+issixx@users.noreply.github.com 2025-10-08T17:20:18+09:00 GitHub noreply@github.com 2025-10-08T11:20:18+03:00 server : fix cancel pending task (#16467)
b2c08c9ec4cd59ed88c1ebdd94f109af5ccc978e 7fdd16b432d247121fe5fe7b21f8805f85266c85 Georgi Gerganov ggerganov@gmail.com 2025-10-08T10:57:53+03:00 GitHub noreply@github.com 2025-10-08T10:57:53+03:00 metal : mark FA blocks (#16372)
7fdd16b432d247121fe5fe7b21f8805f85266c85 74b8fc17f92ada295a648e3c5eb28f46bca7d892 Georgi Gerganov ggerganov@gmail.com 2025-10-08T10:57:29+03:00 GitHub noreply@github.com 2025-10-08T10:57:29+03:00 server : improve context checkpoint logic (#16440)
74b8fc17f92ada295a648e3c5eb28f46bca7d892 aeaf8a36f06b5810f5ae4bbefe26edb33925cf5e Reese Levine reeselevine1@gmail.com 2025-10-07T13:48:56-07:00 GitHub noreply@github.com 2025-10-07T13:48:56-07:00 ggml webgpu: profiling, CI updates, reworking of command submission (#16452)
aeaf8a36f06b5810f5ae4bbefe26edb33925cf5e df1b612e29ba97a2e67db339b1e8c7465702b7e8 Tarek Dakhran tarek@liquid.ai 2025-10-07T20:03:35+02:00 GitHub noreply@github.com 2025-10-07T20:03:35+02:00 llama : support LiquidAI LFM2-MoE hybrid model (#16464)
df1b612e29ba97a2e67db339b1e8c7465702b7e8 4e0388aa8a1f4ef1065701dc9c3947aea0a1b9a5 Georgi Gerganov ggerganov@gmail.com 2025-10-07T15:57:14+03:00 GitHub noreply@github.com 2025-10-07T15:57:14+03:00 server : add `/v1/health` endpoint (#16461)
4e0388aa8a1f4ef1065701dc9c3947aea0a1b9a5 ef4c5b87ea2556ff8ca99cca3abdf48bdbca22f2 Sascha Rogmann 59577610+srogmann@users.noreply.github.com 2025-10-07T11:11:08+02:00 GitHub noreply@github.com 2025-10-07T11:11:08+02:00 webui : added download action (#13552) (#16282)
ef4c5b87ea2556ff8ca99cca3abdf48bdbca22f2 c61ae20d05bd4fdd8551311325a2845336449426 Georgi Gerganov ggerganov@gmail.com 2025-10-07T10:32:32+03:00 GitHub noreply@github.com 2025-10-07T10:32:32+03:00 presets : fix pooling param for embedding models (#16455)
c61ae20d05bd4fdd8551311325a2845336449426 0123ff38f53d34752f29239a29d0e40a6dc4110f Radoslav Gerganov rgerganov@gmail.com 2025-10-07T09:59:13+03:00 GitHub noreply@github.com 2025-10-07T06:59:13Z rpc : update documentation (#16441)
0123ff38f53d34752f29239a29d0e40a6dc4110f 0a319bb75ed29d968e2a9b544011b09ccb932915 Georgi Gerganov ggerganov@gmail.com 2025-10-07T08:24:17+03:00 GitHub noreply@github.com 2025-10-07T08:24:17+03:00 memory : use sequential equal splits for recurrent modules (#16442)
0a319bb75ed29d968e2a9b544011b09ccb932915 1d6092fc72f4d10f4486ac95edfd414bc08b62b8 Georgi Gerganov ggerganov@gmail.com 2025-10-07T08:23:30+03:00 GitHub noreply@github.com 2025-10-07T08:23:30+03:00 metal : add support for non-padded FA KV (#16148)
1d6092fc72f4d10f4486ac95edfd414bc08b62b8 8ae32dc9ecd7aeebf3a5b43557e0552a0a04cd4f Georgi Gerganov ggerganov@gmail.com 2025-10-07T08:22:35+03:00 GitHub noreply@github.com 2025-10-07T08:22:35+03:00 tests : add -INF blocks to the KQ mask in the FA tests (#16380)
8ae32dc9ecd7aeebf3a5b43557e0552a0a04cd4f 3df2244df40c67dfd6ad548b40ccc507a066af2b Georgi Gerganov ggerganov@gmail.com 2025-10-07T08:21:40+03:00 GitHub noreply@github.com 2025-10-07T08:21:40+03:00 metal : various optimizations + refactoring (#16446)
3df2244df40c67dfd6ad548b40ccc507a066af2b c08002a1988348403a5fd59b1fa3de3a10a6f92f Gadflyii 34758915+Gadflyii@users.noreply.github.com 2025-10-06T12:55:53-05:00 GitHub noreply@github.com 2025-10-06T19:55:53+02:00 llama : add --no-host to disable host buffers (#16310)
c08002a1988348403a5fd59b1fa3de3a10a6f92f 3a002afafa8e06555f11aed88b02d055d8a166f3 Gabe Goodhart ghart@us.ibm.com 2025-10-06T10:59:40-06:00 GitHub noreply@github.com 2025-10-06T18:59:40+02:00 chat : Granite Docling stopping (#16438)
3a002afafa8e06555f11aed88b02d055d8a166f3 a23b9bdbd3b64ce172f9962249f432d01aea7437 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-06T17:40:21+02:00 GitHub noreply@github.com 2025-10-06T17:40:21+02:00 ci : refactor sdk caching to minimize storage (#16414)
a23b9bdbd3b64ce172f9962249f432d01aea7437 04e632a4aab8e6bfff0f8bc216b36ceb1e199ff9 Georgi Gerganov ggerganov@gmail.com 2025-10-06T16:05:27+03:00 GitHub noreply@github.com 2025-10-06T16:05:27+03:00 ggml : fix unaligned access in AMX code (#16315)
04e632a4aab8e6bfff0f8bc216b36ceb1e199ff9 a80ff183abe4e5a76316257ffa597da41b3b6fa0 Daniel Bevenius daniel.bevenius@gmail.com 2025-10-06T14:56:59+02:00 GitHub noreply@github.com 2025-10-06T14:56:59+02:00 ci : remove missing reranker model files (#16444)
a80ff183abe4e5a76316257ffa597da41b3b6fa0 1d49ca37594fb49db6aa9518ba7c512e5ccd0108 Daniel Bevenius daniel.bevenius@gmail.com 2025-10-06T14:17:12+02:00 GitHub noreply@github.com 2025-10-06T14:17:12+02:00 ggml-cpu : fix leftover handling in ggml_vec_scale_f32 for SVE (#16443)
1d49ca37594fb49db6aa9518ba7c512e5ccd0108 c5fef0fcea3b978a2318b1af170209ecec7c37b4 Yuannan yuannan@users.noreply.github.com 2025-10-06T09:29:56Z GitHub noreply@github.com 2025-10-06T12:29:56+03:00 nix : removed metal for nix (#16118)
c5fef0fcea3b978a2318b1af170209ecec7c37b4 ca71fb9b368e3db96e028f80c4c9df6b6b370edd Oleksandr Kuvshynov 661042+okuvshynov@users.noreply.github.com 2025-10-06T03:53:31-04:00 GitHub noreply@github.com 2025-10-06T10:53:31+03:00 server: update readme to mention n_past_max metric (#16436)
ca71fb9b368e3db96e028f80c4c9df6b6b370edd 35266573b968e1c947b367782fb4b3eddbb4f3c0 Gabe Goodhart ghart@us.ibm.com 2025-10-05T06:57:47-06:00 GitHub noreply@github.com 2025-10-05T14:57:47+02:00 model : Granite docling + Idefics3 preprocessing (SmolVLM) (#16206)
35266573b968e1c947b367782fb4b3eddbb4f3c0 86df2c9ae4f2f1ee63d2558a9dc797b98524639b Reese Levine reeselevine1@gmail.com 2025-10-04T20:59:31-07:00 GitHub noreply@github.com 2025-10-04T20:59:31-07:00 ggml webgpu: actually add softmax, fix rms_norm offset (#16400)
86df2c9ae4f2f1ee63d2558a9dc797b98524639b f39283960b58a92ecc0c72567711318b20e22b55 Eve 139727413+netrunnereve@users.noreply.github.com 2025-10-04T20:04:27Z GitHub noreply@github.com 2025-10-04T22:04:27+02:00 vulkan: use a more appropriate amount of threads when generating shaders (#16418)
f39283960b58a92ecc0c72567711318b20e22b55 898acba6816ad23b6a9491347d30e7570bffadfd Radoslav Gerganov rgerganov@gmail.com 2025-10-04T16:22:45+03:00 GitHub noreply@github.com 2025-10-04T16:22:45+03:00 rpc : check src buffer when copying tensor (#16421)
898acba6816ad23b6a9491347d30e7570bffadfd e29acf74fea996014380d59d31aa504ae8964258 Radoslav Gerganov rgerganov@gmail.com 2025-10-04T12:49:16+03:00 GitHub noreply@github.com 2025-10-04T12:49:16+03:00 rpc : add support for multiple devices (#16276)
e29acf74fea996014380d59d31aa504ae8964258 128d522c04286e019666bd6ee4d18e3fbf8772e2 Acly aclysia@gmail.com 2025-10-04T11:42:56+02:00 GitHub noreply@github.com 2025-10-04T11:42:56+02:00 vulkan : incremental shader builds (#16341)
128d522c04286e019666bd6ee4d18e3fbf8772e2 f6dcda390004b627ef30af378d0c01ad2519289e Pascal admin@serveurperso.com 2025-10-03T20:51:48+02:00 GitHub noreply@github.com 2025-10-03T21:51:48+03:00 chat : support Magistral thinking (#16413)
f6dcda390004b627ef30af378d0c01ad2519289e 606a73f53175077429484b23dcf799f69a31d0bd ddh0 dylanhalladay02@icloud.com 2025-10-03T13:34:51-05:00 GitHub noreply@github.com 2025-10-03T21:34:51+03:00 server : context checkpointing for hybrid and recurrent models (#16382)
606a73f53175077429484b23dcf799f69a31d0bd 946f71ed9ade07e319859b5ce656144140e066fb Georgi Gerganov ggerganov@gmail.com 2025-10-03T19:18:56+03:00 GitHub noreply@github.com 2025-10-03T19:18:56+03:00 metal : fix loop bound in ggml_mem_ranges (#16412)
946f71ed9ade07e319859b5ce656144140e066fb 638d330246954e88dffc22ce01fec15e6894e544 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-03T14:40:25+02:00 GitHub noreply@github.com 2025-10-03T14:40:25+02:00 llama : fix shapes for bert/mpt q/k norm (#16409)
638d330246954e88dffc22ce01fec15e6894e544 84c8e305e8010a1a3d43bdd0a25f737ac67809a4 Acly aclysia@gmail.com 2025-10-03T13:49:08+02:00 GitHub noreply@github.com 2025-10-03T13:49:08+02:00 ggml : fix graph reallocation with multiple chunks (#16396)
84c8e305e8010a1a3d43bdd0a25f737ac67809a4 2aaf0a2a2056d75d0dd53ab8a181473760e6ab22 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-03T12:51:40+02:00 GitHub noreply@github.com 2025-10-03T12:51:40+02:00 Fix missing messages on sibling navigation (#16408)
2aaf0a2a2056d75d0dd53ab8a181473760e6ab22 0e1f83855609d73beaf05d818640b6cfd39d287b Jeff Bolz jbolz@nvidia.com 2025-10-03T05:50:46-05:00 GitHub noreply@github.com 2025-10-03T12:50:46+02:00 vulkan: Replace uses of maxMemoryAllocationSize and VK_WHOLE_SIZE (#16354)
0e1f83855609d73beaf05d818640b6cfd39d287b ad126479c25cf983a0f994a08ba0911cf49ed62b Jeff Bolz jbolz@nvidia.com 2025-10-03T04:52:46-05:00 GitHub noreply@github.com 2025-10-03T11:52:46+02:00 vulkan: Fix FA coopmat1 invalid array indexing (#16365)
ad126479c25cf983a0f994a08ba0911cf49ed62b 77233277c912d495d1069543ca540c21a2f9e5b4 Daniel Bevenius daniel.bevenius@gmail.com 2025-10-03T11:45:16+02:00 GitHub noreply@github.com 2025-10-03T11:45:16+02:00 ci : change macos-13 to macos-15-intel (#16401)
77233277c912d495d1069543ca540c21a2f9e5b4 e308efda8e54e3f07578937a45a5d83624eb7970 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-03T11:30:39+02:00 GitHub noreply@github.com 2025-10-03T11:30:39+02:00 Capture model name only after first token (streaming) or completed request (#16405)
e308efda8e54e3f07578937a45a5d83624eb7970 136bda78c5679b266362b39c58338fff46fd2592 Jeff Bolz jbolz@nvidia.com 2025-10-03T03:33:08-05:00 GitHub noreply@github.com 2025-10-03T10:33:08+02:00 vulkan: in flash attention, bounds check against nem1 (don't rely on GGML_KQ_MASK_PAD) (#16316)
136bda78c5679b266362b39c58338fff46fd2592 5113efd34ceda709292de26c72716c49e024fb32 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-03T09:11:34+02:00 GitHub noreply@github.com 2025-10-03T10:11:34+03:00 webui : Fix messages payload sent to chat completions (#16402)
5113efd34ceda709292de26c72716c49e024fb32 d64c8104f090b27b1f99e8da5995ffcfa6b726e2 Pascal admin@serveurperso.com 2025-10-03T08:01:31+02:00 GitHub noreply@github.com 2025-10-03T08:01:31+02:00 fix: track viewportHeight via window.innerHeight to avoid unwanted scrolling (#16356)
d64c8104f090b27b1f99e8da5995ffcfa6b726e2 ef07a4090672a3438d7f64f197795d7dc1c18957 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-02T20:10:12+02:00 GitHub noreply@github.com 2025-10-02T20:10:12+02:00 test-barrier : do not use more threads than physically available (#16389)
ef07a4090672a3438d7f64f197795d7dc1c18957 34fcc5a4ace8c69476ef2ea3857f39a60334acc4 Reese Levine reeselevine1@gmail.com 2025-10-02T11:00:31-07:00 GitHub noreply@github.com 2025-10-02T11:00:31-07:00 ggml webgpu: add support for soft_max, optimize rms_norm (#16357)
34fcc5a4ace8c69476ef2ea3857f39a60334acc4 91a2a5655658bb9ab77894716b82fae7ecb4b4d1 Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-10-02T19:43:22+02:00 GitHub noreply@github.com 2025-10-02T20:43:22+03:00 model : Apertus model implementation (#15852)
91a2a5655658bb9ab77894716b82fae7ecb4b4d1 72ee736c447be4ededb51ab97904b4d33c90c261 R0CKSTAR yeahdongcn@gmail.com 2025-10-02T21:29:56+08:00 GitHub noreply@github.com 2025-10-02T16:29:56+03:00 musa: update compile flags (#16265)
72ee736c447be4ededb51ab97904b4d33c90c261 f09aefaa84d7f4d5df3f400f67944b94fef5b795 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-02T13:51:36+02:00 GitHub noreply@github.com 2025-10-02T13:51:36+02:00 ci : fix ubuntu-latest-cmake-rpc (disable ccache) (#16388)
f09aefaa84d7f4d5df3f400f67944b94fef5b795 bbd32bc0384fbfcf07369617de58856b1e0e95a3 Eve 139727413+netrunnereve@users.noreply.github.com 2025-10-02T08:10:07Z GitHub noreply@github.com 2025-10-02T10:10:07+02:00 ci: update vulkan ci (#16294)
bbd32bc0384fbfcf07369617de58856b1e0e95a3 2be72c2b121ee99f33927149265ce6073ade9e59 Georgi Gerganov ggerganov@gmail.com 2025-10-02T10:35:43+03:00 GitHub noreply@github.com 2025-10-02T10:35:43+03:00 ci : fix clean-up of old logs (#16381)
2be72c2b121ee99f33927149265ce6073ade9e59 95ce0985449c52fb9d6849b95563f7cf933ea0e3 Neo Zhang Jianyu jianyu.zhang@intel.com 2025-10-02T15:16:25+08:00 GitHub noreply@github.com 2025-10-02T10:16:25+03:00 SYCL: Update to oneAPI 2025.2 (#16371)
95ce0985449c52fb9d6849b95563f7cf933ea0e3 c8dedc9999eccf7821a9fe5b29f10e8d075e2217 uvos carl@uvos.xyz 2025-10-02T05:52:59+02:00 GitHub noreply@github.com 2025-10-02T05:52:59+02:00 HIP: add IMbackK to codeowner (#16375)
c8dedc9999eccf7821a9fe5b29f10e8d075e2217 e95fec640f43623911a2cd5bda8b19b1898c530c uvos carl@uvos.xyz 2025-10-01T23:32:39+02:00 GitHub noreply@github.com 2025-10-01T23:32:39+02:00 CI: reenable cdna in rocm docker builds (#16376)
e95fec640f43623911a2cd5bda8b19b1898c530c ded67b94446ef4f7fd988dbde7a12deef9870c13 uvos carl@uvos.xyz 2025-10-01T23:09:25+02:00 GitHub noreply@github.com 2025-10-01T23:09:25+02:00 HIP: Disable ROCWMMA fattn on CDNA when compiled against ROCWMMA 2.0.0 (#16221)
ded67b94446ef4f7fd988dbde7a12deef9870c13 1fe4e38cc20af058ed320bd46cac934991190056 Shunta Saito shunta.saito@gmail.com 2025-10-02T06:08:15+09:00 GitHub noreply@github.com 2025-10-01T23:08:15+02:00 llama : parameter conversion and loading fixes for PLaMo2 variants (#16075)
1fe4e38cc20af058ed320bd46cac934991190056 4201deae9c2ae4732db8957b6ce0808d02ec597c uvos carl@uvos.xyz 2025-10-01T20:18:03+02:00 GitHub noreply@github.com 2025-10-01T20:18:03+02:00 ci: Properly install rocwmma for hip builds (#16305)
4201deae9c2ae4732db8957b6ce0808d02ec597c 764799279f801c696503bacca490b4358587d94c Adrien Gallouët adrien@gallouet.fr 2025-10-01T19:22:18+02:00 GitHub noreply@github.com 2025-10-01T20:22:18+03:00 common: introduce http.h for httplib-based client (#16373)
764799279f801c696503bacca490b4358587d94c 2a9b63383a448ec18c754dfdc6e95cb853940a52 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-01T18:18:10+02:00 GitHub noreply@github.com 2025-10-01T18:18:10+02:00 Conversation action dialogs as singletons from Chat Sidebar + apply conditional rendering for Actions Dropdown for Chat Conversation Items (#16369)
2a9b63383a448ec18c754dfdc6e95cb853940a52 1104ca1a1c926b832274694a2773a17066982ad0 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-01T15:54:42+02:00 GitHub noreply@github.com 2025-10-01T15:54:42+02:00 Improve code block color theming (#16325)
1104ca1a1c926b832274694a2773a17066982ad0 4f1575921cac9b489fe8f8bbf20aff985e7da45e Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-10-01T14:09:52+02:00 GitHub noreply@github.com 2025-10-01T14:09:52+02:00 ci : use registry cache for docker builds (#16366)
4f1575921cac9b489fe8f8bbf20aff985e7da45e 132d673554e65282e92505f4f77401ab971528bb Aleksander Grygier aleksander.grygier@gmail.com 2025-10-01T12:08:16+02:00 GitHub noreply@github.com 2025-10-01T12:08:16+02:00 Add optional setting for showing "Model used:" information (#16337)
132d673554e65282e92505f4f77401ab971528bb aa9538a63aef35f0224b2502f13b19d650a8ce04 Eve 139727413+netrunnereve@users.noreply.github.com 2025-10-01T07:56:36Z GitHub noreply@github.com 2025-10-01T09:56:36+02:00 vulkan: make ggml_vk_default_dispatcher support older vulkan headers (#16345)
aa9538a63aef35f0224b2502f13b19d650a8ce04 e74c92e84236b2bab3f3c77bee4ead94928be360 Aleksander Grygier aleksander.grygier@gmail.com 2025-10-01T07:40:26+02:00 GitHub noreply@github.com 2025-10-01T08:40:26+03:00 webui: Remove running `llama-server` within WebUI `dev.sh` script (#16363)
e74c92e84236b2bab3f3c77bee4ead94928be360 b2ba81dbe07b6dbea9c96b13346c66973dede32c Bartowski 3266127+bartowski1182@users.noreply.github.com 2025-09-30T16:24:36-04:00 GitHub noreply@github.com 2025-09-30T22:24:36+02:00 model : support GLM 4.6 (make a few NextN/MTP tensors not required) (#16359)
b2ba81dbe07b6dbea9c96b13346c66973dede32c bf6f3b3a1965d70e07ca94aab7b01268fe483e96 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-30T21:41:42+02:00 GitHub noreply@github.com 2025-09-30T21:41:42+02:00 ci : fix ccache key for ubuntu-cpu-cmake (#16355)
bf6f3b3a1965d70e07ca94aab7b01268fe483e96 7c156df4148eb524b22f7f363bfd2df00f264e0b Adrien Gallouët angt@huggingface.co 2025-09-30T19:52:41+02:00 GitHub noreply@github.com 2025-09-30T20:52:41+03:00 common : disable progress bar without a tty (#16352)
7c156df4148eb524b22f7f363bfd2df00f264e0b 16b0ca0d2e63fc1a0a43795f8552f9cb61d9f7a5 lhez lih@qti.qualcomm.com 2025-09-30T10:45:45-07:00 GitHub noreply@github.com 2025-09-30T10:45:45-07:00 opencl: support pad_ext (#15888)
16b0ca0d2e63fc1a0a43795f8552f9cb61d9f7a5 8d78cd2613ccdeb3cc86f59bc8f9ddd31cfbd3ed Pascal admin@serveurperso.com 2025-09-30T19:18:54+02:00 GitHub noreply@github.com 2025-09-30T19:18:54+02:00 Chatapi ignore empty sampling (#16330)
8d78cd2613ccdeb3cc86f59bc8f9ddd31cfbd3ed d1c84a662daa91be975863913a59975b11458141 Reese Levine reeselevine1@gmail.com 2025-09-30T09:57:51-07:00 GitHub noreply@github.com 2025-09-30T09:57:51-07:00 ggml webgpu: support for rope,div,sub,glu,scale,cont operators (#16187)
d1c84a662daa91be975863913a59975b11458141 364a7a6d4a786e98947c8a90430ea581213c0ba9 lhez lih@qti.qualcomm.com 2025-09-30T09:55:13-07:00 GitHub noreply@github.com 2025-09-30T09:55:13-07:00 opencl: support ne3 in get_rows (#15866)
364a7a6d4a786e98947c8a90430ea581213c0ba9 2df5bcf357dba0c49b4df1d684b2ffc9e88b7054 Adrien Gallouët angt@huggingface.co 2025-09-30T16:39:44+02:00 GitHub noreply@github.com 2025-09-30T17:39:44+03:00 common : remove common_has_curl() (#16351)
2df5bcf357dba0c49b4df1d684b2ffc9e88b7054 075c01567bb7e93965ae60b7e119182c3e68f3e9 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-30T15:38:01+02:00 GitHub noreply@github.com 2025-09-30T15:38:01+02:00 ci : disable ccache for android (#16348)
075c01567bb7e93965ae60b7e119182c3e68f3e9 a014310374a16f9204f2bcc1b458fc1eda67e469 Georgi Gerganov ggerganov@gmail.com 2025-09-30T13:42:39+03:00 Georgi Gerganov ggerganov@gmail.com 2025-09-30T13:53:55+03:00 ggml : bump version to 0.9.4 (ggml/1363)
98c8269abefdfa8fde47dfb2769ceebc8b704e86 2790ecc304e384d9075c55d0628d8bf440025dc4 Yunzhe Jia yunzhe@calculet.tech 2025-09-30T17:42:57+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-09-30T17:44:27+08:00 server: adapt to calrt
2790ecc304e384d9075c55d0628d8bf440025dc4 f528e97d894b4d93934a0711a00253eea980e799 Yunzhe Jia yunzhe@calculet.tech 2025-09-30T17:37:58+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-09-30T17:44:27+08:00 adapt to calrt
f528e97d894b4d93934a0711a00253eea980e799 dcbe998a6d8f13b6c9cb36bbe7b355ced2ac3842 Yunzhe Jia yunzhe@calculet.tech 2025-09-11T14:13:35+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-09-30T17:44:27+08:00 add calrt mtmd support
a014310374a16f9204f2bcc1b458fc1eda67e469 35fb82497ec6c5904b0adb7e1c881a76c1c692db anavp-nvidia anavp@nvidia.com 2025-09-30T08:13:22Z GitHub noreply@github.com 2025-09-30T11:13:22+03:00 cuda : Enable CUDA Graph usage for Nemotron Nano v2 (NemotronH) (#16328)
35fb82497ec6c5904b0adb7e1c881a76c1c692db 3c62aed89fbe3e15f6f82035e805a0a1add26306 Georgi Gerganov ggerganov@gmail.com 2025-09-30T11:03:23+03:00 GitHub noreply@github.com 2025-09-30T11:03:23+03:00 metal : dynamic simdgroups for MV kernels (#16340)
3c62aed89fbe3e15f6f82035e805a0a1add26306 f1eb1cb1eba042b2d583234e80f65e2e225d8995 Adrien Gallouët angt@huggingface.co 2025-09-30T09:36:33+02:00 GitHub noreply@github.com 2025-09-30T10:36:33+03:00 common : simplify etag tracking by removing json (#16342)
f1eb1cb1eba042b2d583234e80f65e2e225d8995 de41f2b7bffab7582944b213ce12b605f8cf6e0f Charles Xu charles.xu@arm.com 2025-09-30T09:07:20+02:00 GitHub noreply@github.com 2025-09-30T10:07:20+03:00 kleidiai : fix work size and threads sync for fp16 (#16246)
de41f2b7bffab7582944b213ce12b605f8cf6e0f a74a0d69f34f52fa10d4f0a7ce749fb3490d0774 lhez lih@qti.qualcomm.com 2025-09-29T22:30:16-07:00 GitHub noreply@github.com 2025-09-30T08:30:16+03:00 codeowners: add codeowners for opencl backend (#16344)
a74a0d69f34f52fa10d4f0a7ce749fb3490d0774 5f7e166cbf7b9ca928c7fad990098ef32358ac75 Jeff Bolz jbolz@nvidia.com 2025-09-29T19:26:34-05:00 GitHub noreply@github.com 2025-09-29T19:26:34-05:00 tests: override test_set_rows::max_nmse_err to allow for occasional rounding differences (#16295)
5f7e166cbf7b9ca928c7fad990098ef32358ac75 d72f5f7ba260b546190338b0b76f2f152581424f Pascal admin@serveurperso.com 2025-09-29T18:49:47+02:00 GitHub noreply@github.com 2025-09-29T18:49:47+02:00 Fix thinking blocks with quotes + add handling `[THINK]...[/THINK]` blocks (#16326)
d72f5f7ba260b546190338b0b76f2f152581424f b77e6c18e1a6fac5705ed95f03af5436d67484c1 Georgi Gerganov ggerganov@gmail.com 2025-09-29T17:51:48+03:00 GitHub noreply@github.com 2025-09-29T17:51:48+03:00 ci : add AMD runners and workflows (#16249)
b77e6c18e1a6fac5705ed95f03af5436d67484c1 2ddd3f2356c313385fafc19a9f0ce678c8fe03ee alex-spacemit jinghui.huang@spacemit.com 2025-09-29T22:50:44+08:00 GitHub noreply@github.com 2025-09-29T17:50:44+03:00 ggml: riscv: add riscv spacemit backend (#15288)
2ddd3f2356c313385fafc19a9f0ce678c8fe03ee 4d3d455d3c8719eb8f206de328d1b9ba191efd7c Georgi Gerganov ggerganov@gmail.com 2025-09-29T16:50:52+03:00 Georgi Gerganov ggerganov@gmail.com 2025-09-29T17:43:58+03:00 sync : ggml
4d3d455d3c8719eb8f206de328d1b9ba191efd7c c9b1c06467aca8d30fafcab51292bcb285df5b0d Georgi Gerganov ggerganov@gmail.com 2025-09-29T16:49:11+03:00 Georgi Gerganov ggerganov@gmail.com 2025-09-29T17:43:58+03:00 sync : whisper.cpp (ggml/1359)
c9b1c06467aca8d30fafcab51292bcb285df5b0d b6ae75afb49b23db3dfbc2b01e8aabfb047e6880 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-26T17:34:42+02:00 Georgi Gerganov ggerganov@gmail.com 2025-09-29T17:43:58+03:00 ggml : remove -dev suffix from release version (ggml/1355)
b6ae75afb49b23db3dfbc2b01e8aabfb047e6880 b6dff20e2fe352093039e2359370c6bb2b505e0b Daniel Bevenius daniel.bevenius@gmail.com 2025-09-25T14:39:05+02:00 Georgi Gerganov ggerganov@gmail.com 2025-09-29T17:43:58+03:00 ggml : bump version to 0.9.3 (ggml/1353)
b6dff20e2fe352093039e2359370c6bb2b505e0b 2db78c75e4ef7497313fda6dbbb3fcaf9ca0752d Georgi Gerganov ggerganov@gmail.com 2025-09-20T16:44:23+03:00 Georgi Gerganov ggerganov@gmail.com 2025-09-29T17:43:58+03:00 ggml : prepare for development of 0.9.2-dev
2db78c75e4ef7497313fda6dbbb3fcaf9ca0752d 02463ab27b1379ff5e3d936ce8b3bfd356872ea6 Georgi Gerganov ggerganov@gmail.com 2025-09-20T16:44:23+03:00 Georgi Gerganov ggerganov@gmail.com 2025-09-29T17:43:58+03:00 ggml : bump version to 0.9.1
02463ab27b1379ff5e3d936ce8b3bfd356872ea6 adc76347d73b3d915e946efa5de8d0ad9f3904c2 Rafal Lewczuk rafal.lewczuk@gmail.com 2025-09-29T13:17:09+02:00 GitHub noreply@github.com 2025-09-29T13:17:09+02:00 ggml-backend : add root cause in error message if loading backend library fails (#16172)
adc76347d73b3d915e946efa5de8d0ad9f3904c2 3a2bdcda0b31e51db954551e0cd49424133a84f3 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-29T11:09:00+02:00 GitHub noreply@github.com 2025-09-29T11:09:00+02:00 ggml : check cuda and metal argsort limits and add test (#16323)
3a2bdcda0b31e51db954551e0cd49424133a84f3 66bb7985c3fcefda7a74aae9f8321f70eb1646c4 Aleksander Grygier aleksander.grygier@gmail.com 2025-09-29T10:37:20+02:00 GitHub noreply@github.com 2025-09-29T10:37:20+02:00 Improve Mobile UI for dialogs and action dropdowns (#16222)
66bb7985c3fcefda7a74aae9f8321f70eb1646c4 2f61c0f5bf8a620ca4c3872408803ab38cfb9613 Pascal admin@serveurperso.com 2025-09-29T09:08:41+02:00 GitHub noreply@github.com 2025-09-29T09:08:41+02:00 fix: preserved zero values in chat settings inputs and textareas by switching to nullish coalescing for field values and default placeholders (#16312)
2f61c0f5bf8a620ca4c3872408803ab38cfb9613 3ffd0fae473c954bb3e67526b31262048fb508d4 Vinkal vinkal-chudgar@users.noreply.github.com 2025-09-29T12:33:12+05:30 GitHub noreply@github.com 2025-09-29T10:03:12+03:00 llama-cli: prevent spurious assistant token (#16202)
3ffd0fae473c954bb3e67526b31262048fb508d4 a4a0aa5ea2a88e4858996199fefd5439f17b481c ddh0 chemist-mulches-39@icloud.com 2025-09-29T01:30:45-05:00 GitHub noreply@github.com 2025-09-29T09:30:45+03:00 perplexity : show more kl-divergence data (#16321)
a4a0aa5ea2a88e4858996199fefd5439f17b481c 92cd103f627015a95f2107f1c27dc218c7b8ec64 Georgi Gerganov ggerganov@gmail.com 2025-09-29T08:41:28+03:00 GitHub noreply@github.com 2025-09-29T08:41:28+03:00 ggml : fix dependencies for ggml_set_rows (#16318)
92cd103f627015a95f2107f1c27dc218c7b8ec64 b887d2f3413ac231e3cb5925260c39902af4a70c Jeff Bolz jbolz@nvidia.com 2025-09-28T23:50:37-05:00 GitHub noreply@github.com 2025-09-29T06:50:37+02:00 vulkan: Fix validation failure in quantized flash attention (#16292)
b887d2f3413ac231e3cb5925260c39902af4a70c bd0af02fc96c2057726f33c0f0daf7bb8f3e462a Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-28T23:15:03+02:00 GitHub noreply@github.com 2025-09-28T23:15:03+02:00 ggml : fix GGML_F32_VEC_FMA argument order in ggml_vec_mad1_f32 (#16307)
bd0af02fc96c2057726f33c0f0daf7bb8f3e462a d9e0e7c8194dfd7d23bf3a86608c9ece68d77c93 crat0z 11581854+crat0z@users.noreply.github.com 2025-09-28T14:13:50-04:00 GitHub noreply@github.com 2025-09-28T21:13:50+03:00 common : fix reasoning before forced tool call via tool_choice = required (#16264)
d9e0e7c8194dfd7d23bf3a86608c9ece68d77c93 0124ac989f7e7bf08803788f66dbe4106bdcdd58 R0CKSTAR yeahdongcn@gmail.com 2025-09-28T22:38:15+08:00 GitHub noreply@github.com 2025-09-28T16:38:15+02:00 ci : fix musa docker build (#16306)
0124ac989f7e7bf08803788f66dbe4106bdcdd58 2811c65286ae954bec87049f75b86dc022006dcc Aaron Teo aaron.teo1@ibm.com 2025-09-28T19:25:58+08:00 GitHub noreply@github.com 2025-09-28T19:25:58+08:00 devops: switch to using ubuntu-22.04-s390x image (#16302)
2811c65286ae954bec87049f75b86dc022006dcc d8359f5fde480da030bf75c7711573c7c4d993ba Imad Saddik 79410781+ImadSaddik@users.noreply.github.com 2025-09-28T12:04:46+01:00 GitHub noreply@github.com 2025-09-28T13:04:46+02:00 Fixed a few typos in the README of the LLaMA.cpp HTTP Server [no ci] (#16297)
d8359f5fde480da030bf75c7711573c7c4d993ba 6a2c6145a0b91b40eb3c3dba7b20ccc4b270490f Jeff Bolz jbolz@nvidia.com 2025-09-28T01:38:37-05:00 GitHub noreply@github.com 2025-09-28T08:38:37+02:00 vulkan: 64-bit im2col (#16135)
6a2c6145a0b91b40eb3c3dba7b20ccc4b270490f 3b53634fe35771e2e318227aa81585726bae7234 Georgi Gerganov ggerganov@gmail.com 2025-09-28T09:34:44+03:00 GitHub noreply@github.com 2025-09-28T09:34:44+03:00 metal : extend mat-mat multiplication support (#16225)
3b53634fe35771e2e318227aa81585726bae7234 1384abf8b8d5894d32fada453ccf4d196ffba7de Georgi Gerganov ggerganov@gmail.com 2025-09-28T09:34:05+03:00 GitHub noreply@github.com 2025-09-28T09:34:05+03:00 metal : fuse non-sequential nodes (#16102)
1384abf8b8d5894d32fada453ccf4d196ffba7de e6d65fb02d553bd79cad94e517cdca18b687788d Jeff Bolz jbolz@nvidia.com 2025-09-27T20:36:34-05:00 GitHub noreply@github.com 2025-09-27T20:36:34-05:00 vulkan: handle mat_mul with A matrix > 4GB (#16176)
e6d65fb02d553bd79cad94e517cdca18b687788d 8656f5de688cddcaea1d6174535eb60ee23ef6a0 Jeff Bolz jbolz@nvidia.com 2025-09-27T16:43:39-04:00 GitHub noreply@github.com 2025-09-27T22:43:39+02:00 vulkan: support arbitrary KV dimension in flash attention (#16160)
8656f5de688cddcaea1d6174535eb60ee23ef6a0 4807e8f96a61b2adccebd5e57444c94d18de7264 Acly aclysia@gmail.com 2025-09-27T22:41:03+02:00 GitHub noreply@github.com 2025-09-27T22:41:03+02:00 vulkan : make the vulkan.hpp dynamic dispatcher instance private (#16224)
4807e8f96a61b2adccebd5e57444c94d18de7264 c0bfc57af421f8fd63c946c13b7666aed82560e2 Aleksander Grygier aleksander.grygier@gmail.com 2025-09-27T19:56:40+02:00 GitHub noreply@github.com 2025-09-27T19:56:40+02:00 Show message actions by default (#16289)
c0bfc57af421f8fd63c946c13b7666aed82560e2 75a3a6c2cd0002ba40e2dcc92007bc9fdbc69f1a Aman Gupta amangupta052@gmail.com 2025-09-28T00:49:32+08:00 GitHub noreply@github.com 2025-09-27T18:49:32+02:00 CUDA: mul_mat_id for mmf for bs <= 64 for f16 and bs <= 32 for f32 (#16277)
75a3a6c2cd0002ba40e2dcc92007bc9fdbc69f1a 0499b29c6f64c705faaf5860dc4600fca23671f4 Johannes Gäßler johannesg@5d6.de 2025-09-27T18:45:07+02:00 GitHub noreply@github.com 2025-09-27T18:45:07+02:00 CUDA: refactor and deduplicate vector FA kernels (#16208)
0499b29c6f64c705faaf5860dc4600fca23671f4 234e2ff8ed09716fb553437596779399bee31b11 Dmytro Minochkin dmytro.minochkin@gmail.com 2025-09-27T19:26:46+03:00 GitHub noreply@github.com 2025-09-27T18:26:46+02:00 vulkan: throw system error instead of SIGABRT during init on older devices (#16156)
234e2ff8ed09716fb553437596779399bee31b11 3f81b4e91c1d5f098148af117e3f13cf4b077f52 Adrien Gallouët angt@huggingface.co 2025-09-27T18:17:08+02:00 GitHub noreply@github.com 2025-09-27T19:17:08+03:00 server : remove old LLAMA_SERVER_SSL (#16290)
3f81b4e91c1d5f098148af117e3f13cf4b077f52 ace6a54565444b6377bee8e7ac693238e7766279 Jeff Bolz jbolz@nvidia.com 2025-09-27T06:36:11-04:00 GitHub noreply@github.com 2025-09-27T12:36:11+02:00 vulkan: support GET_ROWS for k-quants (#16235)
ace6a54565444b6377bee8e7ac693238e7766279 72b24d96c6888c609d562779a23787304ae4609c Adrien Gallouët angt@huggingface.co 2025-09-27T11:12:46+02:00 GitHub noreply@github.com 2025-09-27T12:12:46+03:00 build : add LLAMA_OPENSSL option (#16287)
72b24d96c6888c609d562779a23787304ae4609c 624207e676ab5eb3ce7af631902bb45fb73a8359 Vinkal vinkal-chudgar@users.noreply.github.com 2025-09-27T02:58:29+05:30 GitHub noreply@github.com 2025-09-26T23:28:29+02:00 model : make minicpm embedding_scale, residual_scale and logit_scale optional with legacy defaults (#16273)
624207e676ab5eb3ce7af631902bb45fb73a8359 807e8c6d310952f2f5656afc63dab9d7083dcb5c Aaron Teo aaron.teo1@ibm.com 2025-09-27T02:03:33+08:00 GitHub noreply@github.com 2025-09-27T02:03:33+08:00 devops: add s390x & ppc64le CI (#15925)
807e8c6d310952f2f5656afc63dab9d7083dcb5c 1a189278944d030211f336e103e96b65f976c361 Aleksander Grygier aleksander.grygier@gmail.com 2025-09-26T19:25:29+02:00 GitHub noreply@github.com 2025-09-26T19:25:29+02:00 Enhance text file detection logic for file attachments (#16199)
1a189278944d030211f336e103e96b65f976c361 e0539eb6aed346d4b25a6ea019044e88771e7690 Aleksander Grygier aleksander.grygier@gmail.com 2025-09-26T18:35:42+02:00 GitHub noreply@github.com 2025-09-26T18:35:42+02:00 Allow viewing conversations even when llama server is down (#16255)
e0539eb6aed346d4b25a6ea019044e88771e7690 5d0a40f390732cbf85d8a3b7b0fc3cbebffe780a Isaac McFadyen isaac@imcf.me 2025-09-26T11:36:48-04:00 GitHub noreply@github.com 2025-09-26T18:36:48+03:00 webui: switch to hash-based routing (alternative of #16079) (#16157)
5d0a40f390732cbf85d8a3b7b0fc3cbebffe780a d12a9836597b1ae4440d64d5a6ed727d27ac0702 Aleksander Grygier aleksander.grygier@gmail.com 2025-09-26T15:59:07+02:00 GitHub noreply@github.com 2025-09-26T15:59:07+02:00 Always show message actions for mobile UI + improvements for user message sizing (#16076)
d12a9836597b1ae4440d64d5a6ed727d27ac0702 cc1cfa277b3aae1f6cc9180472072336597a78d4 Radoslav Gerganov rgerganov@gmail.com 2025-09-26T16:09:34+03:00 GitHub noreply@github.com 2025-09-26T16:09:34+03:00 codeowners : add rgerganov as owner of RPC [no ci] (#16279)
cc1cfa277b3aae1f6cc9180472072336597a78d4 54dbc37053f7d75ea5b0631d5ee88f19391e4314 Aleksei Nikiforov 103434461+AlekseiNikiforovIBM@users.noreply.github.com 2025-09-26T15:00:44+02:00 GitHub noreply@github.com 2025-09-26T15:00:44+02:00 mtmd : fix uninitialized variable in bicubic_resize (#16275)
54dbc37053f7d75ea5b0631d5ee88f19391e4314 b995a10760cb93d23d617d76ecb82a5f95b5e0d3 Georgi Gerganov ggerganov@gmail.com 2025-09-26T14:14:28+03:00 GitHub noreply@github.com 2025-09-26T14:14:28+03:00 metal : report OOM errors (#16274)
b995a10760cb93d23d617d76ecb82a5f95b5e0d3 4710dd31bbcef79d04f85a3a6a8c7d9439c5c79a Adrien Gallouët angt@huggingface.co 2025-09-26T13:12:19+02:00 GitHub noreply@github.com 2025-09-26T14:12:19+03:00 common : use cpp-httplib as a cURL alternative for downloads (#16185)
4710dd31bbcef79d04f85a3a6a8c7d9439c5c79a 9b26511857ac09ae69ab485168fe2d3ee5fb1d6e Adrien Gallouët angt@huggingface.co 2025-09-26T12:39:35+02:00 GitHub noreply@github.com 2025-09-26T13:39:35+03:00 build : fix build-ios-device (#16257)
9b26511857ac09ae69ab485168fe2d3ee5fb1d6e 00217cd41388328c93e9e9644921c4319bb03bcc Aaron Teo aaron.teo1@ibm.com 2025-09-26T18:27:25+08:00 GitHub noreply@github.com 2025-09-26T13:27:25+03:00 ggml-cpu: implement MXFP4 SIMD for s390x (#16193)
00217cd41388328c93e9e9644921c4319bb03bcc 3b337b01a1a85c2f5b49376e8c0bbdb3f521528e Radoslav Gerganov rgerganov@gmail.com 2025-09-26T13:19:23+03:00 GitHub noreply@github.com 2025-09-26T10:19:23Z ci : create git tags for released docker images (#16008)
3b337b01a1a85c2f5b49376e8c0bbdb3f521528e a86a580a6692edd1da076d5422f944c344b4fe5e Daniel Bevenius daniel.bevenius@gmail.com 2025-09-26T07:53:36+02:00 GitHub noreply@github.com 2025-09-26T08:53:36+03:00 codeowners : add danbev as owner of build-xcframework.sh [no ci] (#16268)
a86a580a6692edd1da076d5422f944c344b4fe5e 0f7c69689f4e681948a6005adea9bee9c08ff903 R0CKSTAR yeahdongcn@gmail.com 2025-09-26T08:56:38+08:00 GitHub noreply@github.com 2025-09-26T02:56:38+02:00 musa: upgrade musa sdk to 4.3.0 (#16240)
0f7c69689f4e681948a6005adea9bee9c08ff903 835b2b915c52bcabcd688d025eacff9a07b65f52 R0CKSTAR yeahdongcn@gmail.com 2025-09-26T08:56:10+08:00 GitHub noreply@github.com 2025-09-26T02:56:10+02:00 musa: fix build warnings (#15611)
835b2b915c52bcabcd688d025eacff9a07b65f52 b05a9d650f4da1e70e7e2cbe8fe44e61b39656db Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-25T19:50:28+02:00 GitHub noreply@github.com 2025-09-25T19:50:28+02:00 model : add GroveMoE support (#15510)
b05a9d650f4da1e70e7e2cbe8fe44e61b39656db 27052978e487957b3d39e3bd8fc8ee4aa304bff9 Aaron Teo aaron.teo1@ibm.com 2025-09-25T23:38:10+08:00 GitHub noreply@github.com 2025-09-25T23:38:10+08:00 vendors: update miniaudio version (#16212)
27052978e487957b3d39e3bd8fc8ee4aa304bff9 077c94d0caf87fbd3cf3288dbb5c0fd9670294cf rtaluyev taluyev@gmail.com 2025-09-25T18:20:34+03:00 GitHub noreply@github.com 2025-09-25T18:20:34+03:00 readme : update bindings (#16144)
077c94d0caf87fbd3cf3288dbb5c0fd9670294cf aa3ee0eb0b80efca126cedf9bcb4fb5864b46ce3 Aman Gupta amangupta052@gmail.com 2025-09-25T22:35:05+08:00 GitHub noreply@github.com 2025-09-25T16:35:05+02:00 CUDA: add a fused top-K MoE kernel (#16130)
aa3ee0eb0b80efca126cedf9bcb4fb5864b46ce3 d0991da39d3c39b3980c24cdfccb9a4c95a46870 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-25T12:02:36+02:00 GitHub noreply@github.com 2025-09-25T12:02:36+02:00 model-conversion : add embedding prompt file support (#15871)
d0991da39d3c39b3980c24cdfccb9a4c95a46870 aa719c2f886c445b78113bf1e5849f1fce4124a7 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-25T11:36:47+02:00 GitHub noreply@github.com 2025-09-25T11:36:47+02:00 server : add support for external server for tests (#16243)
aa719c2f886c445b78113bf1e5849f1fce4124a7 4cdd0bb4537f9617e9efcfef6b9454fcefe2ff08 junchao-zhao 68935141+junchao-loongson@users.noreply.github.com 2025-09-25T17:22:55+08:00 GitHub noreply@github.com 2025-09-25T12:22:55+03:00 ggml : fix loongarch lsx compilation error (#15864)
4cdd0bb4537f9617e9efcfef6b9454fcefe2ff08 b5bd037832bcb8ed3086dfe26ce9090bea989af1 Johannes Gäßler johannesg@5d6.de 2025-09-25T11:12:27+02:00 GitHub noreply@github.com 2025-09-25T12:12:27+03:00 docs: fix typo [no ci] (#16244)
b5bd037832bcb8ed3086dfe26ce9090bea989af1 dfcd53f7ecb9bb897a9d752d09c59d10be47237a Douglas Hanley thesecretaryofwar@gmail.com 2025-09-25T03:53:09-05:00 GitHub noreply@github.com 2025-09-25T11:53:09+03:00 llama : add support for qwen3 reranker (#15824)
dfcd53f7ecb9bb897a9d752d09c59d10be47237a 4ea00794b8c995b6deaf4bac159c1778dc27419a Georgi Gerganov ggerganov@gmail.com 2025-09-25T11:30:16+03:00 GitHub noreply@github.com 2025-09-25T11:30:16+03:00 metal : fuse NORM + MUL + ADD, support non-multiples of 4 (#16220)
4ea00794b8c995b6deaf4bac159c1778dc27419a 02a6a82ae7c7ddd1819ae26b0cc36675879aecee Georgi Gerganov ggerganov@gmail.com 2025-09-25T11:29:42+03:00 GitHub noreply@github.com 2025-09-25T11:29:42+03:00 metal : relax reorder conditions (#16216)
02a6a82ae7c7ddd1819ae26b0cc36675879aecee c498fc82fe5b83fc8c6e1627286bdc1f93caddbf Georgi Gerganov ggerganov@gmail.com 2025-09-25T11:29:08+03:00 GitHub noreply@github.com 2025-09-25T11:29:08+03:00 metal : restore im2col perf (#16219)
c498fc82fe5b83fc8c6e1627286bdc1f93caddbf e7a5130a20cf45a3358308356c249734ca982143 Radoslav Gerganov rgerganov@gmail.com 2025-09-25T10:20:02+03:00 GitHub noreply@github.com 2025-09-25T07:20:02Z rpc : use ggml logging facilities
e7a5130a20cf45a3358308356c249734ca982143 bee378e0988b44bbe93b0768208080951b312363 Aaron Teo aaron.teo1@ibm.com 2025-09-25T13:06:30+08:00 GitHub noreply@github.com 2025-09-25T08:06:30+03:00 codeowners: add ownership of zdnn backend [no ci] (#16232)
bee378e0988b44bbe93b0768208080951b312363 5fb557653b8756ac2b6c6a102b1e44abcadf552f Eve 139727413+netrunnereve@users.noreply.github.com 2025-09-25T05:06:06Z GitHub noreply@github.com 2025-09-25T08:06:06+03:00 ci: run the x64 and arm ci on the github machines instead (#16183)
5fb557653b8756ac2b6c6a102b1e44abcadf552f 4ae88d07d026e66b41e85afece74e88af54f4e66 Aaron Teo aaron.teo1@ibm.com 2025-09-25T11:36:30+08:00 GitHub noreply@github.com 2025-09-25T11:36:30+08:00 devops: fix s390x docker release failure (#16231)
4ae88d07d026e66b41e85afece74e88af54f4e66 e789095502b337690c7616db32d7c679a5bd2533 Aaron Teo aaron.teo1@ibm.com 2025-09-25T00:25:04+08:00 GitHub noreply@github.com 2025-09-25T00:25:04+08:00 codeowners: add ownership of zdnn backend [no ci] (#16229)
e789095502b337690c7616db32d7c679a5bd2533 f2a789e33490deb483a2694b066b37e45524bb79 Johannes Gäßler johannesg@5d6.de 2025-09-24T16:53:48+02:00 GitHub noreply@github.com 2025-09-24T16:53:48+02:00 llama: print memory breakdown on exit (#15860)
f2a789e33490deb483a2694b066b37e45524bb79 3a599719673c850647e3bb911ed6d91109bb91d2 Acly aclysia@gmail.com 2025-09-24T16:17:49+02:00 GitHub noreply@github.com 2025-09-24T16:17:49+02:00 ggml : split graph allocations according to backend max buffer size (#15815)
3a599719673c850647e3bb911ed6d91109bb91d2 63b54c81a620981be020184ab99e63a8e50e47cb Tarek Dakhran t.dakhran@gmail.com 2025-09-24T13:42:26+02:00 GitHub noreply@github.com 2025-09-24T13:42:26+02:00 model : add label for LiquidAI LFM2-2.6B model (#16204)
63b54c81a620981be020184ab99e63a8e50e47cb 152729f8848b692e1ae53c197ccca92ef10a523b Jie Fu (傅杰) jiefu@tencent.com 2025-09-24T16:25:26+08:00 GitHub noreply@github.com 2025-09-24T10:25:26+02:00 model-conversion : make causal-verify-logits fails with model names containing "." (#16215)
152729f8848b692e1ae53c197ccca92ef10a523b c0c59c1157b5aeded285a3b25d2463c866f583d3 Uilian Ries uilianries@gmail.com 2025-09-24T08:53:47+02:00 GitHub noreply@github.com 2025-09-24T09:53:47+03:00 common : add missing chrono header for common.cpp (#16211)
c0c59c1157b5aeded285a3b25d2463c866f583d3 7735706b939847df2c6c00b44c36f95a20137c61 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-24T08:53:20+02:00 GitHub noreply@github.com 2025-09-24T08:53:20+02:00 codeowners : match all requirements files (#16214)
7735706b939847df2c6c00b44c36f95a20137c61 4d9ea03d170cf504a40bf3a85788af84cccaee80 Jie Fu (傅杰) jiefu@tencent.com 2025-09-24T14:46:52+08:00 GitHub noreply@github.com 2025-09-24T08:46:52+02:00 model-conversion : run-org-model.py fails to run on mac m1 (#16213)
4d9ea03d170cf504a40bf3a85788af84cccaee80 8ba548dae251f8f2e3834ed5226b76b6e4f4a485 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-24T08:10:09+02:00 GitHub noreply@github.com 2025-09-24T08:10:09+02:00 codeowners : use slash prefix for root files [no ci] (#16210)
8ba548dae251f8f2e3834ed5226b76b6e4f4a485 f505bd83ca7a43c4585ff3d59135e77eae9c793b Jie Fu (傅杰) jiefu@tencent.com 2025-09-24T12:19:23+08:00 GitHub noreply@github.com 2025-09-24T06:19:23+02:00 model-conversion : fix the make targets in the README.md (#16209)
f505bd83ca7a43c4585ff3d59135e77eae9c793b 0889589dbe8c67dd518c58a57afbb02dde2dccbe Georgi Gerganov ggerganov@gmail.com 2025-09-23T20:41:40+03:00 GitHub noreply@github.com 2025-09-23T20:41:40+03:00 ci : disable AMD workflows + update NVIDIA workflows (#16200)
0889589dbe8c67dd518c58a57afbb02dde2dccbe 4e29084ba4104c4ea529fd3163bb6e76f64383df Georgi Gerganov ggerganov@gmail.com 2025-09-23T13:44:25+03:00 GitHub noreply@github.com 2025-09-23T13:44:25+03:00 ci : enable Vulkan workflow on Mac (#16194)
4e29084ba4104c4ea529fd3163bb6e76f64383df f6b4af3d04763b1e0130f5b5fce19c4bc6f83f1c Xiangyan Sun wishstudio@gmail.com 2025-09-23T01:58:12-07:00 GitHub noreply@github.com 2025-09-23T11:58:12+03:00 ggml-cpu: Respect cpumask settings (#16164)
f6b4af3d04763b1e0130f5b5fce19c4bc6f83f1c 264f1b51872c125e23fa0ac1da5e2a1170de9a08 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-23T10:25:20+02:00 GitHub noreply@github.com 2025-09-23T10:25:20+02:00 ggml : fix uninitialized is_on_grid in quantize_row_iq3_xxs_impl (#15928)
264f1b51872c125e23fa0ac1da5e2a1170de9a08 0bc7cc715472fe28822e57f036f0746592ed2c04 Aaron Teo aaron.teo1@ibm.com 2025-09-23T14:53:05+08:00 GitHub noreply@github.com 2025-09-23T14:53:05+08:00 zdnn: refactor codebase + add docs (#16178)
0bc7cc715472fe28822e57f036f0746592ed2c04 4b9f4cb0f89a88de4bdf97727d0457b0c648804c Daniel Bevenius daniel.bevenius@gmail.com 2025-09-23T08:13:22+02:00 GitHub noreply@github.com 2025-09-23T09:13:22+03:00 codeowners : add @danbev to model-conversion example [no ci] (#16190)
4b9f4cb0f89a88de4bdf97727d0457b0c648804c 85e72271ba1ce78adf34fd8997803c991e617ca6 Aaron Teo aaron.teo1@ibm.com 2025-09-23T13:59:34+08:00 GitHub noreply@github.com 2025-09-23T13:59:34+08:00 devops: add s390x containers (#15915)
85e72271ba1ce78adf34fd8997803c991e617ca6 1d0125bcf1cbd7195ad0faf826a20bc7cec7d3f4 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-23T05:59:03+02:00 GitHub noreply@github.com 2025-09-23T05:59:03+02:00 ggml-cpu : fix typo in gemm comments [no ci] (#16189)
1d0125bcf1cbd7195ad0faf826a20bc7cec7d3f4 351f3da39c85f59d581fc184f09283da7f099a3b Gabe Goodhart ghart@us.ibm.com 2025-09-22T12:40:10-06:00 GitHub noreply@github.com 2025-09-22T20:40:10+02:00 feat: Add conversion support in GraniteHybrid for non-hybrid (all attn) (#16177)
351f3da39c85f59d581fc184f09283da7f099a3b 3ecb2f671a2f49d56357f99d135a94e841759178 Haiyue Wang haiyuewa@163.com 2025-09-23T01:57:46+08:00 GitHub noreply@github.com 2025-09-22T19:57:46+02:00 clang-tidy : disable warning about performance enum size (#16127)
3ecb2f671a2f49d56357f99d135a94e841759178 432cf4304c5379bfbd1f3feed65ec205955b2f4b Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-22T19:13:00+02:00 GitHub noreply@github.com 2025-09-22T19:13:00+02:00 ggml : implement set_rows with i32 index (#16159)
432cf4304c5379bfbd1f3feed65ec205955b2f4b 37a23c17bdc9c99c9c6ad41168e4ced3724b72cd Georgi Gerganov ggerganov@gmail.com 2025-09-22T18:20:21+03:00 GitHub noreply@github.com 2025-09-22T18:20:21+03:00 codeowners : update + cleanup (#16174)
37a23c17bdc9c99c9c6ad41168e4ced3724b72cd 138c87ce8bd558b2cc134ada7316a3dad8eb67ac Adrien Gallouët angt@huggingface.co 2025-09-22T14:13:51+02:00 GitHub noreply@github.com 2025-09-22T15:13:51+03:00 common : enable `--offline` mode without curl support (#16137)
138c87ce8bd558b2cc134ada7316a3dad8eb67ac c6db9a10278f40b4b8d3f6931728f2cce2356daa Quentin Bramas quentin.bramas@gmail.com 2025-09-22T10:53:13+02:00 GitHub noreply@github.com 2025-09-22T11:53:13+03:00 webui : fix handling incomplete chunks (#16107)
c6db9a10278f40b4b8d3f6931728f2cce2356daa d05affbab7d1364fa7ac06c03cd33f03235ae840 GideonSerf gdserf.gs@gmail.com 2025-09-22T10:49:58+02:00 GitHub noreply@github.com 2025-09-22T11:49:58+03:00 embedding : fix typos in README (#16171)
d05affbab7d1364fa7ac06c03cd33f03235ae840 4f324a556caa5be0be2261604ef4aab0faf36eb8 Haiyue Wang haiyuewa@163.com 2025-09-22T16:48:42+08:00 GitHub noreply@github.com 2025-09-22T11:48:42+03:00 common : remove unused local variables (#16140)
4f324a556caa5be0be2261604ef4aab0faf36eb8 a71ae3ba7a857394103f12e30b387c48242c84f2 Georgi Gerganov ggerganov@gmail.com 2025-09-22T11:12:37+03:00 GitHub noreply@github.com 2025-09-22T11:12:37+03:00 ggml : extend ggml_can_fuse to work with non-sequential nodes (#16123)
a71ae3ba7a857394103f12e30b387c48242c84f2 05a2458121f9cdcc98ea6ffeeab6f23023136266 Georgi Gerganov ggerganov@gmail.com 2025-09-22T11:12:09+03:00 GitHub noreply@github.com 2025-09-22T11:12:09+03:00 ggml : add ggml_op_is_empty (#16122)
05a2458121f9cdcc98ea6ffeeab6f23023136266 96fdca043be8fa733bbd59868a75517f92633376 Xuan-Son Nguyen son@huggingface.co 2025-09-22T15:10:58+07:00 GitHub noreply@github.com 2025-09-22T11:10:58+03:00 codeowners : update ownership for @ngxson and @allozuar (#16128)
96fdca043be8fa733bbd59868a75517f92633376 b2d980fce063fa3591c59ad50b946c8da66c294f Shin-myoung-serp relent95@naver.com 2025-09-22T17:04:01+09:00 GitHub noreply@github.com 2025-09-22T10:04:01+02:00 Vulkan: add conv_transpose_2d operation (#16022)
b2d980fce063fa3591c59ad50b946c8da66c294f 5c6106a696af2b00ce01d14ba525979d5a32b199 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-22T09:59:05+02:00 GitHub noreply@github.com 2025-09-22T10:59:05+03:00 codeowners : claim responsibility for ci, models, gguf-py and convert (#16124)
5c6106a696af2b00ce01d14ba525979d5a32b199 ec65fb52f0cc890e5085a6c5995ab08a156265fb Georgi Gerganov ggerganov@gmail.com 2025-09-22T10:58:02+03:00 GitHub noreply@github.com 2025-09-22T10:58:02+03:00 contrib : update roles (#16113)
ec65fb52f0cc890e5085a6c5995ab08a156265fb 1d660d2fae42ea2e1d3569638e722bf7a37b6b19 Georgi Gerganov ggerganov@gmail.com 2025-09-22T10:16:05+03:00 GitHub noreply@github.com 2025-09-22T10:16:05+03:00 ci : remove vulkaninfo calls (#16169)
1d660d2fae42ea2e1d3569638e722bf7a37b6b19 a20d810d79adfbf82e0eb703af6f27a0e2d1a539 Georgi Gerganov ggerganov@gmail.com 2025-09-22T09:11:39+03:00 GitHub noreply@github.com 2025-09-22T09:11:39+03:00 ci : use smaller model (#16168)
a20d810d79adfbf82e0eb703af6f27a0e2d1a539 4d0a7cbc617e384fc355077a304c883b5c7d4fb6 Jeff Bolz jbolz@nvidia.com 2025-09-22T00:37:17-05:00 GitHub noreply@github.com 2025-09-22T07:37:17+02:00 vulkan: add RTE variants of exp shader (#16165)
4d0a7cbc617e384fc355077a304c883b5c7d4fb6 9073a73d82a916cea0809de225ef5175c3a86e91 Georgi Gerganov ggerganov@gmail.com 2025-09-22T08:31:40+03:00 GitHub noreply@github.com 2025-09-22T08:31:40+03:00 ci : adjust params for less runtime (#16167)
9073a73d82a916cea0809de225ef5175c3a86e91 51f5a45fbe575dcd54bdd2a339ef8e8424d1c12a Ruben Ortlam picard12@live.de 2025-09-22T07:22:43+02:00 GitHub noreply@github.com 2025-09-22T07:22:43+02:00 vulkan: vec dot matrix multiplication fix (#16151)
51f5a45fbe575dcd54bdd2a339ef8e8424d1c12a c4510dc9374e17dcb8726902ab5216067a92b3d3 lhez lih@qti.qualcomm.com 2025-09-21T16:42:10-07:00 GitHub noreply@github.com 2025-09-21T16:42:10-07:00 opencl: fix concat crash on win arm64 with Adreno (#15944)
c4510dc9374e17dcb8726902ab5216067a92b3d3 da30ab5f8696cabb2d4620cdc0aa41a298c54fd6 lhez lih@qti.qualcomm.com 2025-09-21T14:48:44-07:00 GitHub noreply@github.com 2025-09-21T14:48:44-07:00 opencl: initial `q8_0` mv support (#15732)
da30ab5f8696cabb2d4620cdc0aa41a298c54fd6 28baac9c9f491c872e2c37762d3bd90446b005e9 Georgi Gerganov ggerganov@gmail.com 2025-09-21T19:00:27+03:00 GitHub noreply@github.com 2025-09-21T19:00:27+03:00 ci : add label for the RISC-V runner (#16150)
28baac9c9f491c872e2c37762d3bd90446b005e9 1eeb523c3e0c7ffbd59469f5463dcbdecba3535e Georgi Gerganov ggerganov@gmail.com 2025-09-21T16:50:45+03:00 GitHub noreply@github.com 2025-09-21T16:50:45+03:00 ci : migrate ggml ci to self-hosted runners (#16116)
1eeb523c3e0c7ffbd59469f5463dcbdecba3535e 5bb4a3edec297e74b0f7bd4ed5d0fdd12e28d858 Giuseppe Scrivano gscrivan@redhat.com 2025-09-21T08:31:55+02:00 GitHub noreply@github.com 2025-09-21T08:31:55+02:00 vulkan: optimize UMA buffer operations and fix driver hangs (#16059)
5bb4a3edec297e74b0f7bd4ed5d0fdd12e28d858 7f766929ca8e8e01dcceb1c526ee584f7e5e1408 Jeff Bolz jbolz@nvidia.com 2025-09-21T01:23:37-05:00 GitHub noreply@github.com 2025-09-21T08:23:37+02:00 vulkan: fix validation error about VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_KHR (#16086)
7f766929ca8e8e01dcceb1c526ee584f7e5e1408 405921dcefdd4e90ed948a4bf179007c2fa92b2d Georgi Gerganov ggerganov@gmail.com 2025-09-20T12:55:47+03:00 Georgi Gerganov ggerganov@gmail.com 2025-09-20T13:02:14+03:00 sync : ggml
405921dcefdd4e90ed948a4bf179007c2fa92b2d fa6383ca7e7ccb8ca3bdfeb37e348ddc4113aa26 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-16T06:16:52+02:00 Georgi Gerganov ggerganov@gmail.com 2025-09-20T13:02:14+03:00 ggml : introduce semantic versioning (ggml/1336)
fa6383ca7e7ccb8ca3bdfeb37e348ddc4113aa26 803dac2e48ef3ba26a504eb27c4e77ec2d21f7d0 Gregor Jasny gjasny@googlemail.com 2025-09-10T17:21:11+02:00 Georgi Gerganov ggerganov@gmail.com 2025-09-20T13:02:14+03:00 CUDA : conditionally add cuda architectures (ggml/1341)
803dac2e48ef3ba26a504eb27c4e77ec2d21f7d0 459c0c2c1a400f960d7b8e8d94d31a8426f80986 Ruben Ortlam picard12@live.de 2025-09-20T10:42:56+02:00 GitHub noreply@github.com 2025-09-20T10:42:56+02:00 vulkan: use vec dot for matrix matrix multiplications (#16056)
459c0c2c1a400f960d7b8e8d94d31a8426f80986 be79d9fdd95ab8955527c4aaa67b90e8b9516718 Benni 73313922+BenjaminBruenau@users.noreply.github.com 2025-09-20T07:56:30+02:00 GitHub noreply@github.com 2025-09-20T07:56:30+02:00 server: fix SSE and OpenAI compatibility for error messages when streaming (#16109)
be79d9fdd95ab8955527c4aaa67b90e8b9516718 f432d8d83e7407073634c5e4fd81a3d23a10827f ssweens 1149151+ssweens@users.noreply.github.com 2025-09-19T15:15:21-07:00 GitHub noreply@github.com 2025-09-20T00:15:21+02:00 llama-bench: add --devices and --list-devices support (#16039)
f432d8d83e7407073634c5e4fd81a3d23a10827f 4067f07fc5aa5722b87db303b561597004696f6c shun095 8069181+shun095@users.noreply.github.com 2025-09-20T00:57:30+09:00 GitHub noreply@github.com 2025-09-19T09:57:30-06:00 chat: Fix streaming parser for granite models (#15682)
4067f07fc5aa5722b87db303b561597004696f6c 4b8560ab56fdd9819358b47c338bbc8ec357c57e Aleksander Grygier aleksander.grygier@gmail.com 2025-09-19T09:52:27+02:00 GitHub noreply@github.com 2025-09-19T09:52:27+02:00 feat: Improve mobile UI for Settings Dialog (#16084)
4b8560ab56fdd9819358b47c338bbc8ec357c57e 0dd58b6877f3dc106593d6bc68d98305f553c566 Xuan-Son Nguyen son@huggingface.co 2025-09-19T13:02:51+07:00 GitHub noreply@github.com 2025-09-19T13:02:51+07:00 chat : fix build on arm64 (#16101)
0dd58b6877f3dc106593d6bc68d98305f553c566 69ffd891631befa9e6b485fd646a16dab4f2c007 Xuan-Son Nguyen son@huggingface.co 2025-09-19T11:31:56+07:00 GitHub noreply@github.com 2025-09-19T06:31:56+02:00 ggml : refactor forward_dup for cpu backend (#16062)
69ffd891631befa9e6b485fd646a16dab4f2c007 246c0d9c795fb25da8268625d597580155a3672f Adrien Gallouët angt@huggingface.co 2025-09-18T23:07:26+02:00 GitHub noreply@github.com 2025-09-18T23:07:26+02:00 ggml-amx : fix ggml_amx_init() on generic Linux (#16049)
246c0d9c795fb25da8268625d597580155a3672f 3edd87cd055a45d885fa914d879d36d33ecfc3e1 Adrien Gallouët angt@huggingface.co 2025-09-18T23:07:18+02:00 GitHub noreply@github.com 2025-09-18T23:07:18+02:00 cmake : fix static linking for OpenMP on Unix-like systems (#16031)
3edd87cd055a45d885fa914d879d36d33ecfc3e1 c0b45097c33e2667a94444f08cc9e36bec0a5e2e Shawn Gu shawngu@qti.qualcomm.com 2025-09-18T12:03:34-07:00 GitHub noreply@github.com 2025-09-18T12:03:34-07:00 opencl: optimize mxfp4 kernels (#16037)
c0b45097c33e2667a94444f08cc9e36bec0a5e2e 38dbdf4c057515ccea9bec0ca2518f86d5e4d28e Jeff Bolz jbolz@nvidia.com 2025-09-18T13:46:17-05:00 GitHub noreply@github.com 2025-09-18T13:46:17-05:00 rename optimize_graph to graph_optimize (#16082)
38dbdf4c057515ccea9bec0ca2518f86d5e4d28e 368560a1e3b9a3bc83af741b0b2bc9e46fb420d2 Bowen Han fancycode@gmail.com 2025-09-18T11:26:03-07:00 GitHub noreply@github.com 2025-09-18T20:26:03+02:00 CUDA: Optimize PAD_REFLECT_1D (#15957)
368560a1e3b9a3bc83af741b0b2bc9e46fb420d2 4ca088b036313c2d8e682f4cfeb7c29edd85d0b9 Johannes Gäßler johannesg@5d6.de 2025-09-18T19:28:32+02:00 GitHub noreply@github.com 2025-09-18T19:28:32+02:00 CUDA: fix compilation on CC 6.0 (#16091)
4ca088b036313c2d8e682f4cfeb7c29edd85d0b9 703f9e32c4eb3166f8d63007c26e31a1466c21af Eric Curtin eric.curtin@docker.com 2025-09-18T16:22:50+01:00 GitHub noreply@github.com 2025-09-18T16:22:50+01:00 Add resumable downloads for llama-server model loading (#15963)
703f9e32c4eb3166f8d63007c26e31a1466c21af ad6bd9083bc346afd4ecd2fbb83e0306c0141c99 Georgi Gerganov ggerganov@gmail.com 2025-09-18T16:28:41+03:00 GitHub noreply@github.com 2025-09-18T16:28:41+03:00 metal : use function constants for mul_mv_ext kernels (#16074)
ad6bd9083bc346afd4ecd2fbb83e0306c0141c99 2b6b55a59f590a1e1eb2dcd09d5b8b6feb9cc748 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-18T13:28:22+02:00 GitHub noreply@github.com 2025-09-18T13:28:22+02:00 cuda : add missing F32<->I32 entries in ggml_cuda_cpy_fn (#16060)
2b6b55a59f590a1e1eb2dcd09d5b8b6feb9cc748 e58174cecbc45bf79bf653cd2c984395940c6ef4 Radoslav Gerganov rgerganov@gmail.com 2025-09-18T13:36:57+03:00 GitHub noreply@github.com 2025-09-18T10:36:57Z server : include usage statistics only when user request them (#16052)
e58174cecbc45bf79bf653cd2c984395940c6ef4 b213fce89bee8cb56b587b91e15a4278f8ed0180 Georgi Gerganov ggerganov@gmail.com 2025-09-18T12:47:56+03:00 GitHub noreply@github.com 2025-09-18T12:47:56+03:00 llama : bump max seq limit from 64 to 256 (#15916)
b213fce89bee8cb56b587b91e15a4278f8ed0180 e00f3fd8fff2cf5a8c8c9f475034bd089c8bcce4 Georgi Gerganov ggerganov@gmail.com 2025-09-18T12:33:45+03:00 GitHub noreply@github.com 2025-09-18T12:33:45+03:00 metal : improve F32, F16 and BF16 mat-vec multiplication (#16057)
e00f3fd8fff2cf5a8c8c9f475034bd089c8bcce4 f2f28380ea68939c0481df3e4216b698824a59df Jhen-Jie Hong iainst0409@gmail.com 2025-09-18T15:06:48+08:00 GitHub noreply@github.com 2025-09-18T10:06:48+03:00 metal : avoid call free for non-owned buffer (#16067)
f2f28380ea68939c0481df3e4216b698824a59df 62c3b645c56ad5288aa401901f043b050edac5a2 Georgi Gerganov ggerganov@gmail.com 2025-09-18T10:03:24+03:00 GitHub noreply@github.com 2025-09-18T10:03:24+03:00 metal : handle nil cv during pipeline creation (#16065)
62c3b645c56ad5288aa401901f043b050edac5a2 d304f459d8619399eb232b1e3689379306f439d3 Chenguang Li 757486878@qq.com 2025-09-18T09:26:33+08:00 GitHub noreply@github.com 2025-09-18T09:26:33+08:00 CANN: Remove print (#16044)
d304f459d8619399eb232b1e3689379306f439d3 0320ac5264279d74f8ee91bafa6c90e9ab9bbb91 Reese Levine reeselevine1@gmail.com 2025-09-17T13:09:40-07:00 GitHub noreply@github.com 2025-09-17T13:09:40-07:00 GGML WebGPU: Support for ADD, MUL, RMS_NORM, GET_ROWS operators (#16018)
0320ac5264279d74f8ee91bafa6c90e9ab9bbb91 a7a98e0fffed794396b3fbad4dcdbbc184963645 Georgi Gerganov ggerganov@gmail.com 2025-09-17T20:38:12+03:00 GitHub noreply@github.com 2025-09-17T20:38:12+03:00 metal : refactor + optimize v2 (#15995)
a7a98e0fffed794396b3fbad4dcdbbc184963645 8f8f2274ee3601fecf6e2d57b52f701c81bede21 Aleksander Grygier aleksander.grygier@gmail.com 2025-09-17T19:29:13+02:00 GitHub noreply@github.com 2025-09-17T19:29:13+02:00 SvelteKit-based WebUI (#14839)
8f8f2274ee3601fecf6e2d57b52f701c81bede21 c959b676be29e93f8dbc3bd6056ceba812a9eb72 Xuan-Son Nguyen son@huggingface.co 2025-09-18T00:18:21+07:00 GitHub noreply@github.com 2025-09-17T19:18:21+02:00 convert : add Llama4ForCausalLM (#16042)
c959b676be29e93f8dbc3bd6056ceba812a9eb72 cd08fc3ecc0264b4414b68af3874a6c689ed60c1 Johannes Gäßler johannesg@5d6.de 2025-09-17T15:32:42+02:00 GitHub noreply@github.com 2025-09-17T15:32:42+02:00 CUDA: fix FA occupancy, optimize tile kernel (#15982)
cd08fc3ecc0264b4414b68af3874a6c689ed60c1 cb5bb6cc05119c24e7711ca2956cd0e6d409d396 David Ribeiro Alves davidralves@gmail.com 2025-09-17T01:08:02-07:00 GitHub noreply@github.com 2025-09-17T11:08:02+03:00 common : Fix corrupted memory error on json grammar initialization (#16038)
cb5bb6cc05119c24e7711ca2956cd0e6d409d396 a91d035b901e8a9edf810f63d130ee49adf27be2 Eve 139727413+netrunnereve@users.noreply.github.com 2025-09-17T07:35:37Z GitHub noreply@github.com 2025-09-17T09:35:37+02:00 vulkan: automatically remove unsupported devices (#15976)
a91d035b901e8a9edf810f63d130ee49adf27be2 745cbcf2fe1eb88f8db615ac622f0b944d924ad6 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-17T09:34:09+02:00 GitHub noreply@github.com 2025-09-17T09:34:09+02:00 ci : revert back to macos-13 for macOS-latest-cmake-x64 (#16040)
745cbcf2fe1eb88f8db615ac622f0b944d924ad6 1cbd80f8cf80a817715b1ccc5680fe2a3c5172c8 Jie Fu (傅杰) jiefu@tencent.com 2025-09-17T15:30:55+08:00 GitHub noreply@github.com 2025-09-17T09:30:55+02:00 llama-quant : fix the verification of attention layers for encoder-decoder models (#16023)
1cbd80f8cf80a817715b1ccc5680fe2a3c5172c8 85286f354813056f6c835046c0acfa3bf6ba9432 Jie Fu (傅杰) jiefu@tencent.com 2025-09-17T15:29:00+08:00 GitHub noreply@github.com 2025-09-17T10:29:00+03:00 examples : support encoder-decoder models in the simple example (#16002)
85286f354813056f6c835046c0acfa3bf6ba9432 d5fabe3682de515fd09d6c981f7a0d1b75614455 Shane A shanea@allenai.org 2025-09-17T00:01:58-07:00 GitHub noreply@github.com 2025-09-17T09:01:58+02:00 model : add OLMo3 support (#16015)
d5fabe3682de515fd09d6c981f7a0d1b75614455 8ff206097c2bf3ca1c7aa95f9d6db779fc7bdd68 Chenguang Li 757486878@qq.com 2025-09-17T14:33:08+08:00 GitHub noreply@github.com 2025-09-17T14:33:08+08:00 CANN: Optimize ggml_cann_set_device (#15935)
8ff206097c2bf3ca1c7aa95f9d6db779fc7bdd68 77475530b8bbea3bf578632507e1284cdfe2c8c0 jacekpoplawski 67507230+jacekpoplawski@users.noreply.github.com 2025-09-16T16:17:08+02:00 GitHub noreply@github.com 2025-09-16T16:17:08+02:00 llama-bench: add --n-cpu-moe support (#15952)
77475530b8bbea3bf578632507e1284cdfe2c8c0 3913f8730ec6d6245480affc30ae3049107956f4 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-16T15:27:52+02:00 GitHub noreply@github.com 2025-09-16T15:27:52+02:00 ci : use macos-latest for arm64 webgpu build (#16029)
3913f8730ec6d6245480affc30ae3049107956f4 76888d202ed2b835ae19ea9f9db6baf39e419297 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-16T15:25:57+02:00 GitHub noreply@github.com 2025-09-16T15:25:57+02:00 ggml : fix padding in timestep embedding kernels (#15932)
76888d202ed2b835ae19ea9f9db6baf39e419297 f1fbffb5c0b34b2a68febb7da3fd0f8333f1ed4c Daniel Bevenius daniel.bevenius@gmail.com 2025-09-16T13:41:38+02:00 GitHub noreply@github.com 2025-09-16T13:41:38+02:00 ci : upload xcframework artifact from ios-xcode-build job (#16010)
f1fbffb5c0b34b2a68febb7da3fd0f8333f1ed4c 51abc96bdc52ba8cd6ad78dcf12ed9a041d7b442 Bowen Han fancycode@gmail.com 2025-09-15T23:59:19-07:00 GitHub noreply@github.com 2025-09-16T08:59:19+02:00 fix: apply clang-format to CUDA macros (#16017)
51abc96bdc52ba8cd6ad78dcf12ed9a041d7b442 07808ebb07e2b1aa19032705e332679ddf967614 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-16T05:57:16+02:00 GitHub noreply@github.com 2025-09-16T05:57:16+02:00 ci : update macos-latest* jobs to use macos-latest (#15938)
07808ebb07e2b1aa19032705e332679ddf967614 6d758839ff741d4966ca92b7f801b7a8b5b96364 Yuri Khrustalev ykhrustalev@users.noreply.github.com 2025-09-15T22:54:44-04:00 GitHub noreply@github.com 2025-09-16T09:54:44+07:00 cmake : Do not install tools on iOS targets (#15903)
6d758839ff741d4966ca92b7f801b7a8b5b96364 3d4053f77f0f78ee2b791088c02af653ebee42dd Aman Gupta amangupta052@gmail.com 2025-09-16T10:38:28+08:00 GitHub noreply@github.com 2025-09-16T10:38:28+08:00 Add LLaDA-7b-MoE diffusion model (#16003)
3d4053f77f0f78ee2b791088c02af653ebee42dd dc381aa9a6dc45f00673471d34b8bddd30e77570 Jake Karnes jake.karnes@gmail.com 2025-09-15T16:28:31-06:00 GitHub noreply@github.com 2025-09-16T00:28:31+02:00 CUDA: fix im2col_3d to respect non-contiguous inputs (views) (#15956)
dc381aa9a6dc45f00673471d34b8bddd30e77570 10d197409bd9537ff302ad09966fe406882fef9d Diego Devesa slarengh@gmail.com 2025-09-15T14:38:52-07:00 GitHub noreply@github.com 2025-09-15T23:38:52+02:00 docker : enable rocWMMA in ROCm images, add gfx1151 (#15997)
10d197409bd9537ff302ad09966fe406882fef9d b907255f4bd169b0dc7dca9553b4c54af5170865 Diego Devesa slarengh@gmail.com 2025-09-15T14:38:42-07:00 GitHub noreply@github.com 2025-09-15T23:38:42+02:00 releases : switch to rocWMMA develop branch, add gfx1151 (#15992)
b907255f4bd169b0dc7dca9553b4c54af5170865 28c39da7c645185ade5436767929d7ec33006033 yael-works 106673277+yael-works@users.noreply.github.com 2025-09-15T19:51:35+03:00 GitHub noreply@github.com 2025-09-15T18:51:35+02:00 SYCL: Add COUNT_EQUAL operator support (#15991)
28c39da7c645185ade5436767929d7ec33006033 106220562aca42b6738b8f51acfce0db1b8a2fb6 Nikolay Popov 131475237+npopov-vst@users.noreply.github.com 2025-09-15T13:08:30+03:00 GitHub noreply@github.com 2025-09-15T11:08:30+01:00 llama-run: Fix model download on Windows (#15988)
106220562aca42b6738b8f51acfce0db1b8a2fb6 a68f31edd71cc39141113f05f7133a3e9ece8c61 Aman Gupta amangupta052@gmail.com 2025-09-15T17:35:11+08:00 GitHub noreply@github.com 2025-09-15T17:35:11+08:00 CUDA: some micro-optimizations in mmf.cuh for mul_mat_id (#15926)
a68f31edd71cc39141113f05f7133a3e9ece8c61 b8e09f08b9a91c0401bc67d17a17c90756420346 ddh0 chemist-mulches-39@icloud.com 2025-09-15T02:54:57-05:00 GitHub noreply@github.com 2025-09-15T09:54:57+02:00 fix KLD percentile output (#15999)
b8e09f08b9a91c0401bc67d17a17c90756420346 6c019cb04e86e2dacfe62ce7666c64e9717dde1f Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-14T23:00:59+02:00 GitHub noreply@github.com 2025-09-14T23:00:59+02:00 model : add grok-2 support (#15539)
6c019cb04e86e2dacfe62ce7666c64e9717dde1f 9dcd200d57bc6f05a59a6a8df361d5d183af4124 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-14T21:17:04+02:00 GitHub noreply@github.com 2025-09-14T21:17:04+02:00 server : only attempt to enable thinking if using jinja (#15967)
9dcd200d57bc6f05a59a6a8df361d5d183af4124 0fa154e3502e940df914f03b41475a2b80b985b0 Georgi Gerganov ggerganov@gmail.com 2025-09-14T22:02:32+03:00 GitHub noreply@github.com 2025-09-14T22:02:32+03:00 metal : remove memory pools (#15966)
0fa154e3502e940df914f03b41475a2b80b985b0 261e6a20ffdb79c4875e674b4f6b514bc73cff8f Adam channeladam@users.noreply.github.com 2025-09-15T04:43:54+10:00 GitHub noreply@github.com 2025-09-14T20:43:54+02:00 rocm.Dockerfile: added gfx1200,gfx1201 architectures to support AMD Radeon RX 9000 series (#15994)
261e6a20ffdb79c4875e674b4f6b514bc73cff8f a0e13dcbe5bae7025660349ef3e4ead060e507f2 Ruben Ortlam picard12@live.de 2025-09-14T16:56:28+02:00 GitHub noreply@github.com 2025-09-14T16:56:28+02:00 Vulkan: Clean up mul_mm shader (#15987)
a0e13dcbe5bae7025660349ef3e4ead060e507f2 a14bd350141fb42b8bf2dd2342cebc27bfdce399 lcy lcy0321@users.noreply.github.com 2025-09-14T22:20:35+08:00 GitHub noreply@github.com 2025-09-14T07:20:35-07:00 build: fix the build failures of Windows HIP release job (#15984)
a14bd350141fb42b8bf2dd2342cebc27bfdce399 918b26f197f55d5d562446dfc876d0e637929d07 Georgi Gerganov ggerganov@gmail.com 2025-09-14T15:33:22+03:00 GitHub noreply@github.com 2025-09-14T15:33:22+03:00 metal : fix kernel requirements (#15983)
918b26f197f55d5d562446dfc876d0e637929d07 9ecb88434644c865232bb665d2f6f05049fc6456 Radoslav Gerganov rgerganov@gmail.com 2025-09-14T12:28:18+03:00 GitHub noreply@github.com 2025-09-14T12:28:18+03:00 rpc : fix regression when --device is used (#15981)
9ecb88434644c865232bb665d2f6f05049fc6456 d1c6f11f47bae063a68a6be8e4830c060a11b7bd Diego Devesa slarengh@gmail.com 2025-09-14T02:21:59-07:00 GitHub noreply@github.com 2025-09-14T02:21:59-07:00 releases : update ROCM, add gfx1200, gfx1201, gfx1151 (#15972)
d1c6f11f47bae063a68a6be8e4830c060a11b7bd 6380d6a3e709cb02a8695afdd96b40e674477332 Radoslav Gerganov rgerganov@gmail.com 2025-09-14T12:10:07+03:00 GitHub noreply@github.com 2025-09-14T12:10:07+03:00 doc : update documentation for --tensor-split (#15980)
6380d6a3e709cb02a8695afdd96b40e674477332 aa0c461efe3603639af1a1defed2438d9c16ca0f Aaron Teo aaron.teo1@ibm.com 2025-09-14T13:37:03+08:00 GitHub noreply@github.com 2025-09-14T13:37:03+08:00 ggml-zdnn: rm user mapped buffers (#15965)
aa0c461efe3603639af1a1defed2438d9c16ca0f b9c9c9f789cd57fb0b28b25223305613cd90fa10 Jeff Bolz jbolz@nvidia.com 2025-09-13T16:29:43+01:00 GitHub noreply@github.com 2025-09-13T17:29:43+02:00 vulkan: fix failing dequant shaders (#15862)
b9c9c9f789cd57fb0b28b25223305613cd90fa10 50f4281a6f5c3a5d68bdeb12f904fa01e0e2ba91 Jeff Bolz jbolz@nvidia.com 2025-09-13T16:23:30+01:00 GitHub noreply@github.com 2025-09-13T17:23:30+02:00 vulkan: initialize vulkan-hpp to allow using extension function pointers (#15705)
50f4281a6f5c3a5d68bdeb12f904fa01e0e2ba91 55758b00cae1451a0d789c50a5eb8c9bee42b9a0 Diego Devesa slarengh@gmail.com 2025-09-13T07:49:49-07:00 GitHub noreply@github.com 2025-09-13T16:49:49+02:00 llama : allow using iGPUs with --device (#15951)
55758b00cae1451a0d789c50a5eb8c9bee42b9a0 f161463a54d9f93d41246286aa4a9569a91d804d Georgi Gerganov ggerganov@gmail.com 2025-09-13T16:24:22+03:00 GitHub noreply@github.com 2025-09-13T16:24:22+03:00 metal : refactor kernel loading (#15964)
f161463a54d9f93d41246286aa4a9569a91d804d 84d7b2fca11d1be118ce776f6d72a486c4883b74 Georgi Gerganov ggerganov@gmail.com 2025-09-13T13:54:28+03:00 GitHub noreply@github.com 2025-09-13T13:54:28+03:00 metal : allow ops to run concurrently (#15929)
84d7b2fca11d1be118ce776f6d72a486c4883b74 40be51152d4dc2d47444a4ed378285139859895b Georgi Gerganov ggerganov@gmail.com 2025-09-13T12:45:04+03:00 GitHub noreply@github.com 2025-09-13T12:45:04+03:00 metal : fix memory leaks (#15962)
40be51152d4dc2d47444a4ed378285139859895b 4bf5549269d99bac936a65892da2312cc71b2421 Aaron Teo aaron.teo1@ibm.com 2025-09-13T02:39:52+08:00 GitHub noreply@github.com 2025-09-13T02:39:52+08:00 ggml-zdnn: fix #15414, activate FP16 and BF16 acceleration and incorrect zTensor free (#15839)
4bf5549269d99bac936a65892da2312cc71b2421 f4e664f838cc65d89fa845c48c372e43852112e4 Eric Curtin eric.curtin@docker.com 2025-09-12T16:31:50+01:00 GitHub noreply@github.com 2025-09-12T16:31:50+01:00 Add docker protocol support for llama-server model loading (#15790)
f4e664f838cc65d89fa845c48c372e43852112e4 f088b6a84f6aed0abb619dbb9d375e726fc888a6 Haiyue Wang haiyuewa@163.com 2025-09-12T23:16:32+08:00 GitHub noreply@github.com 2025-09-12T18:16:32+03:00 context : remove redundant explicit casting to the same type (#15948)
f088b6a84f6aed0abb619dbb9d375e726fc888a6 304ac5693d1e2124f83e0584bc5eea6311d3d3b4 Georgi Gerganov ggerganov@gmail.com 2025-09-12T17:02:55+03:00 GitHub noreply@github.com 2025-09-12T17:02:55+03:00 server : adjust prompt similarity thold + add logs (#15913)
304ac5693d1e2124f83e0584bc5eea6311d3d3b4 6c88ad8fa741a2182a2dcf8c5353ae4b5c66e656 Ruben Ortlam picard12@live.de 2025-09-12T13:24:21+02:00 GitHub noreply@github.com 2025-09-12T13:24:21+02:00 Vulkan iGPU device selection overhaul and PCI ID API support (#15947)
6c88ad8fa741a2182a2dcf8c5353ae4b5c66e656 704d90c987cdf00751567b2088c4e54742aa2d3f Mathieu Baudier mbaudier@argeo.org 2025-09-12T09:06:20+02:00 GitHub noreply@github.com 2025-09-12T09:06:20+02:00 vulkan: Make device memory check more portable (#15939)
704d90c987cdf00751567b2088c4e54742aa2d3f 360d6533db39e11577afe9b0aece20c6b5ddaf1f Neo Zhang Jianyu jianyu.zhang@intel.com 2025-09-12T09:15:12+08:00 GitHub noreply@github.com 2025-09-12T09:15:12+08:00 Revert "sycl: add usage of enqueue_functions extension (#14244)" (#15910)
360d6533db39e11577afe9b0aece20c6b5ddaf1f 0e6ff0046f4a2983b2c77950aa75960fe4b4f0e2 Diego Devesa slarengh@gmail.com 2025-09-11T13:47:38-07:00 GitHub noreply@github.com 2025-09-11T22:47:38+02:00 ggml-backend : add GGML_BACKEND_DEVICE_TYPE_IGPU device type (#15797)
0e6ff0046f4a2983b2c77950aa75960fe4b4f0e2 df082f56309073ecf885eceaa21b86e8a487e61b Johannes Gäßler johannesg@5d6.de 2025-09-11T21:19:58+02:00 GitHub noreply@github.com 2025-09-11T21:19:58+02:00 CUDA: larger SRAM reads for tile FA, AMD FP16 dot (#15927)
df082f56309073ecf885eceaa21b86e8a487e61b 24a6734daf6932ff29ba8c1ff0245c51d76f783e ddh0 chemist-mulches-39@icloud.com 2025-09-11T12:12:34-05:00 GitHub noreply@github.com 2025-09-11T19:12:34+02:00 nitpick : correct MB to MiB (#15934)
24a6734daf6932ff29ba8c1ff0245c51d76f783e 2b3efea9a4d91216850856fbb77075db26f6a6eb Daniel Bevenius daniel.bevenius@gmail.com 2025-09-11T15:39:12+02:00 GitHub noreply@github.com 2025-09-11T14:39:12+01:00 ggml-cpu : add check for ARM MATMUL_INT8/i8mm support (#15922)
2b3efea9a4d91216850856fbb77075db26f6a6eb c0389dba43d50695f9d3f57dd1f1a14cbefc100c Charles Xu charles.xu@arm.com 2025-09-11T12:45:40+02:00 GitHub noreply@github.com 2025-09-11T12:45:40+02:00 kleidiai: fix GGML_ASSERT(*cur_backend_id != -1) failed (#15614)
c0389dba43d50695f9d3f57dd1f1a14cbefc100c 00681dfc16ba4cebb9c7fbd2cf2656e06a0692a4 hipudding huafengchun@gmail.com 2025-09-11T15:59:37+08:00 GitHub noreply@github.com 2025-09-11T15:59:37+08:00 CANN: Disable acl_graph for prefill stage (#15933)
00681dfc16ba4cebb9c7fbd2cf2656e06a0692a4 4f658855fa8f2e42b7ed9a5b298fa39a2e39b096 Oliver Simons osimons@nvidia.com 2025-09-10T22:04:03+02:00 GitHub noreply@github.com 2025-09-10T22:04:03+02:00 CUDA: Add `fastdiv` to `k_bin_bcast*`, giving 1-3% E2E performance (#15872)
4f658855fa8f2e42b7ed9a5b298fa39a2e39b096 6ab397e12ba8e9f776341cdae68f7ffb2f8d2cde Jie Fu (傅杰) jiefu@tencent.com 2025-09-11T02:51:51+08:00 GitHub noreply@github.com 2025-09-10T20:51:51+02:00 llama : support T5 models with unequal number of encoder-decoder layers (#15909)
6ab397e12ba8e9f776341cdae68f7ffb2f8d2cde 9de447d94e1ae9d1a36e5a2e5bf47483352c0d9c Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-10T19:08:59+02:00 GitHub noreply@github.com 2025-09-10T19:08:59+02:00 graph : support non-contiguous Q in build_attn_mha (#15908)
9de447d94e1ae9d1a36e5a2e5bf47483352c0d9c 0f0a3c2851134d49955f3c85afbb0b1bb47c3e07 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-10T17:31:40+02:00 GitHub noreply@github.com 2025-09-10T17:31:40+02:00 ggml-cpu : fix padding in ggml_timestep_embedding (#15917)
0f0a3c2851134d49955f3c85afbb0b1bb47c3e07 33daece86b65607451d0d4378d2d04ba6a20ad55 Georgi Gerganov ggerganov@gmail.com 2025-09-10T17:52:35+03:00 GitHub noreply@github.com 2025-09-10T17:52:35+03:00 metal : make the backend async (#15906)
33daece86b65607451d0d4378d2d04ba6a20ad55 e7b6d83b524bbc24a1343d862de6dba8e8eddbd6 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-10T15:39:57+02:00 GitHub noreply@github.com 2025-09-10T15:39:57+02:00 ci : add caching for ROCm installation in release workflow (#15924)
e7b6d83b524bbc24a1343d862de6dba8e8eddbd6 2cfef4d117d67ab1dec002915b48a15d11ee1973 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-10T14:17:09+02:00 GitHub noreply@github.com 2025-09-10T14:17:09+02:00 tests : filter out no-ops from coverage report (#15900)
2cfef4d117d67ab1dec002915b48a15d11ee1973 09e72a037c77b6823fde9e1e4cf28e815b9c41ab j-k dev@j-k.io 2025-09-10T12:51:28+01:00 GitHub noreply@github.com 2025-09-10T14:51:28+03:00 media : add transparent icon svg and png [no ci] (#15891)
09e72a037c77b6823fde9e1e4cf28e815b9c41ab 10d8b2b6b0ac2cae252a80b4daea5da55ab63c2f Jesse jesse@createthis.com 2025-09-10T07:28:47-04:00 GitHub noreply@github.com 2025-09-10T14:28:47+03:00 gitignore : Ignore vim swap files in tests (#15901)
10d8b2b6b0ac2cae252a80b4daea5da55ab63c2f 28b5f190ef1dbea5edf82dbc8b4407b721fadd13 Chenguang Li 757486878@qq.com 2025-09-10T18:42:00+08:00 GitHub noreply@github.com 2025-09-10T18:42:00+08:00 CANN: Add ROPE sin/cos cache for reuse (#15912)
28b5f190ef1dbea5edf82dbc8b4407b721fadd13 86587da03bd78df8f4e7d8b111a0c1d2494d6ed0 Chenguang Li 757486878@qq.com 2025-09-10T15:29:12+08:00 GitHub noreply@github.com 2025-09-10T15:29:12+08:00 CANN: implement LRU cache for ACL graphs (#15814)
86587da03bd78df8f4e7d8b111a0c1d2494d6ed0 ff02caf9eed261423289d1531a56536fbf57bfc2 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-10T05:33:58+02:00 GitHub noreply@github.com 2025-09-10T05:33:58+02:00 llama : check returned fn ptrs from ggml_backend_reg_get_proc_address (#15893)
ff02caf9eed261423289d1531a56536fbf57bfc2 ae355f6f7108540297d8b7f7ae71d20fe610a0b7 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-10T05:23:19+02:00 GitHub noreply@github.com 2025-09-10T05:23:19+02:00 ci : cache ROCm installation in windows-latest-cmake-hip (#15887)
ae355f6f7108540297d8b7f7ae71d20fe610a0b7 4f63cd705c7b6f457f36c63fdc053e07f6f3cc6b Ruben Ortlam picard12@live.de 2025-09-09T22:26:03+02:00 GitHub noreply@github.com 2025-09-09T22:26:03+02:00 vulkan: throw the oom error instead of no memory type found (#15905)
4f63cd705c7b6f457f36c63fdc053e07f6f3cc6b 17bc5a815f0bebc1844c99796760cd7df5e9b8d9 Jeff Bolz jbolz@nvidia.com 2025-09-09T07:41:15-05:00 GitHub noreply@github.com 2025-09-09T14:41:15+02:00 vulkan: Fix OOB accesses in soft_max_back (#15861)
17bc5a815f0bebc1844c99796760cd7df5e9b8d9 ed54e32558ffe0f2c3b1d31f3227211fae05bd49 Johannes Gäßler johannesg@5d6.de 2025-09-09T14:04:43+02:00 GitHub noreply@github.com 2025-09-09T14:04:43+02:00 HIP: use v_dot2_f32_f16 instruction for FA (#15884)
ed54e32558ffe0f2c3b1d31f3227211fae05bd49 a972faebed5fdc4a3d2a844d92d476058c02e02d lksj92hs 134250687+lksj92hs@users.noreply.github.com 2025-09-09T15:01:15+03:00 GitHub noreply@github.com 2025-09-09T14:01:15+02:00 Workaround for subgroup arithmetic failing on MoltenVK with AMD GPUs (issue 15846) (#15886)
a972faebed5fdc4a3d2a844d92d476058c02e02d 550cf726e133fd0a069d991287fd3a2a3e3e1cbd Aman Gupta amangupta052@gmail.com 2025-09-09T14:38:02+08:00 GitHub noreply@github.com 2025-09-09T14:38:02+08:00 CUDA: Add mul_mat_id support for the mmf kernel (#15767)
550cf726e133fd0a069d991287fd3a2a3e3e1cbd c252ce67c4b99f056eeb18d42a289e28fee03475 Johannes Gäßler johannesg@5d6.de 2025-09-09T08:11:01+02:00 GitHub noreply@github.com 2025-09-09T08:11:01+02:00 CUDA: fix GET_ROWS for large tensors (#15882)
c252ce67c4b99f056eeb18d42a289e28fee03475 70cd37dbbebdb7a2f84f08207f18eabb0b291a55 Georgi Gerganov ggerganov@gmail.com 2025-09-09T08:42:10+03:00 GitHub noreply@github.com 2025-09-09T08:42:10+03:00 contrib : add notes about merging PRs (#15881)
70cd37dbbebdb7a2f84f08207f18eabb0b291a55 acc1b008cfd95e63d5f99a370dbffb98e5a99d2c Daniel Bevenius daniel.bevenius@gmail.com 2025-09-09T06:06:52+02:00 GitHub noreply@github.com 2025-09-09T06:06:52+02:00 requirements : update transformers/torch for Embedding Gemma (#15828)
acc1b008cfd95e63d5f99a370dbffb98e5a99d2c 7057faf64b514e991e2f70147f82bb13d544b1c0 Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-09-09T06:05:55+02:00 GitHub noreply@github.com 2025-09-09T06:05:55+02:00 model-conversion : add extra debugging support for model conversion (#15877)
7057faf64b514e991e2f70147f82bb13d544b1c0 fe1c92cd7bb491e9c5767c9de413f2b7dc1e832b Aldehir Rojas hello@alde.dev 2025-09-08T16:14:32-05:00 GitHub noreply@github.com 2025-09-08T16:14:32-05:00 json : support `enum` values within `allOf` (#15830)
fe1c92cd7bb491e9c5767c9de413f2b7dc1e832b e68aa10d8f3d26fdad5b912540362d79de5460e3 j-k dev@j-k.io 2025-09-08T19:57:01+01:00 GitHub noreply@github.com 2025-09-08T21:57:01+03:00 media : add llama1 icon (#15878)
e68aa10d8f3d26fdad5b912540362d79de5460e3 0a16bf52e6874369cb4fd2bdc4863ef984b688e7 Jeff Bolz jbolz@nvidia.com 2025-09-08T13:10:07-05:00 GitHub noreply@github.com 2025-09-09T02:10:07+08:00 vulkan: sort graph to allow more parallel execution (#15850)
0a16bf52e6874369cb4fd2bdc4863ef984b688e7 88021565f08e0b7c4e07ac089a15ec16fae9166c Aman Gupta amangupta052@gmail.com 2025-09-09T01:23:46+08:00 GitHub noreply@github.com 2025-09-09T01:23:46+08:00 CUDA: generate_cu_files.py - add missing mxfp4 (#15880)
88021565f08e0b7c4e07ac089a15ec16fae9166c 56920f56651908d5cce7a310dabf54ac4f6fbb7f Jesse jesse@createthis.com 2025-09-08T10:59:48-04:00 GitHub noreply@github.com 2025-09-08T16:59:48+02:00 chat : Deepseek V3.1 reasoning and tool calling support (OpenAI Style) (#15533)
56920f56651908d5cce7a310dabf54ac4f6fbb7f b0d52998b962bd2681c34bf52af993af79f178b8 Xuan-Son Nguyen son@huggingface.co 2025-09-08T21:50:05+07:00 GitHub noreply@github.com 2025-09-08T16:50:05+02:00 server : bring back timings_per_token (#15879)
b0d52998b962bd2681c34bf52af993af79f178b8 f28d4f4ac963f182ea9d0fe9e269f3f5f3782aaf Georgi Gerganov ggerganov@gmail.com 2025-09-08T13:56:51+03:00 GitHub noreply@github.com 2025-09-08T13:56:51+03:00 cuda : fix supports_op condition for get_rows when number of blocks is too large (#15868)
f28d4f4ac963f182ea9d0fe9e269f3f5f3782aaf 9fcb29f22f5c33c04c7f0daebb24057899d67a1a Georgi Gerganov ggerganov@gmail.com 2025-09-08T13:34:56+03:00 GitHub noreply@github.com 2025-09-08T13:34:56+03:00 metal : refactor + optimize (#15857)
9fcb29f22f5c33c04c7f0daebb24057899d67a1a 5ef22d281de9c5eaaf616874bc490b89241128cb Xuan-Son Nguyen son@huggingface.co 2025-09-08T17:33:01+07:00 GitHub noreply@github.com 2025-09-08T12:33:01+02:00 ggml: allow casting between f32 and i32 (#15783)
5ef22d281de9c5eaaf616874bc490b89241128cb 233d773d02c37982badddf3994e9953a175f34da Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-08T11:55:44+02:00 GitHub noreply@github.com 2025-09-08T12:55:44+03:00 CUDA: non-contiguous src0 not supported for PAD (#15869)
233d773d02c37982badddf3994e9953a175f34da a885dcff11a7b73f9377812d6151f6b15d307de0 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-08T09:44:34+02:00 GitHub noreply@github.com 2025-09-08T09:44:34+02:00 convert : force setting sliding_window from original config (#15867)
a885dcff11a7b73f9377812d6151f6b15d307de0 663027fd5490438ce9c27ea866a560e1e268d11f Georgi Gerganov ggerganov@gmail.com 2025-09-08T10:27:07+03:00 GitHub noreply@github.com 2025-09-08T10:27:07+03:00 batched-bench : fix llama_synchronize usage during prompt processing (#15835)
663027fd5490438ce9c27ea866a560e1e268d11f cf0e3ba1500bd23635b444f64ba23cbdb56c92ef Georgi Gerganov ggerganov@gmail.com 2025-09-08T10:26:36+03:00 GitHub noreply@github.com 2025-09-08T10:26:36+03:00 context : fix n_outputs during reserve (#15858)
cf0e3ba1500bd23635b444f64ba23cbdb56c92ef d413dca00360a7e4cb71441dacecfa32556fcc31 Georgi Gerganov ggerganov@gmail.com 2025-09-08T10:25:33+03:00 GitHub noreply@github.com 2025-09-08T10:25:33+03:00 model : avoid ggml_cont_3d for fused QKV weights (#15662)
d413dca00360a7e4cb71441dacecfa32556fcc31 85ca66a74676e6d5df4433016488e039a4b464ae Jeff Bolz jbolz@nvidia.com 2025-09-07T23:23:41-05:00 GitHub noreply@github.com 2025-09-07T23:23:41-05:00 tests: large sizes for get_rows (#15687)
85ca66a74676e6d5df4433016488e039a4b464ae 3976dfbe00f02a62c0deca32c46138e4f0ca81d8 Chenguang Li 757486878@qq.com 2025-09-08T10:03:29+08:00 GitHub noreply@github.com 2025-09-08T10:03:29+08:00 CANN: Stream sync between devices for acl_graph (#15809)
3976dfbe00f02a62c0deca32c46138e4f0ca81d8 d36e61c580bf7fc7879c443c542312a42b718e11 Jeff Bolz jbolz@nvidia.com 2025-09-07T13:50:26-05:00 GitHub noreply@github.com 2025-09-07T13:50:26-05:00 vulkan: support im2col_3d (#15795)
d36e61c580bf7fc7879c443c542312a42b718e11 c97b5e5854b47b18a248d77edb693c63018a0865 Aaron Teo aaron.teo1@ibm.com 2025-09-08T02:18:28+08:00 GitHub noreply@github.com 2025-09-08T02:18:28+08:00 ggml-cpu: clean up s390x SIMD (#15855)
c97b5e5854b47b18a248d77edb693c63018a0865 267e99867f09bec8bcc2e424ad9bcddd6cccf9d0 Jeff Bolz jbolz@nvidia.com 2025-09-07T12:00:49-05:00 GitHub noreply@github.com 2025-09-07T19:00:49+02:00 vulkan: Support pad_ext (#15794)
267e99867f09bec8bcc2e424ad9bcddd6cccf9d0 3b15924d71237a43bb5ad71f5b885ee66a821342 Jeff Bolz jbolz@nvidia.com 2025-09-07T11:53:07-05:00 GitHub noreply@github.com 2025-09-07T18:53:07+02:00 vulkan: Use larger loads in scalar/coopmat1 matmul (#15729)
3b15924d71237a43bb5ad71f5b885ee66a821342 79bc429262268ad2ac8a364cfe6c2d6b9c5f008a Daniel Bevenius daniel.bevenius@gmail.com 2025-09-07T10:19:45+02:00 GitHub noreply@github.com 2025-09-07T11:19:45+03:00 ggml WebGPU: remove userdata from request adapter callback (#15527)
79bc429262268ad2ac8a364cfe6c2d6b9c5f008a c4df49a42d396bdf7344501813e7de53bc9e7bb3 Johannes Gäßler johannesg@5d6.de 2025-09-07T00:26:28+02:00 GitHub noreply@github.com 2025-09-07T00:26:28+02:00 CUDA: faster tile FA (Pascal/AMD), headsize 256 (#15769)
c4df49a42d396bdf7344501813e7de53bc9e7bb3 3c3635d2f20424d557b5b0605a2a356214ffe048 Charles Xu charles.xu@arm.com 2025-09-06T16:08:43+02:00 GitHub noreply@github.com 2025-09-06T22:08:43+08:00 kleidiai: generalize compute_forward_kv_cache to compute_forward_fp16 (#15817)
3c3635d2f20424d557b5b0605a2a356214ffe048 61bdfd5298a78593be649a1035ee2a120b13c4f0 Xuan-Son Nguyen son@huggingface.co 2025-09-06T19:45:24+07:00 GitHub noreply@github.com 2025-09-06T14:45:24+02:00 server : speed up tests (#15836)
61bdfd5298a78593be649a1035ee2a120b13c4f0 01806e77714ae8a78130d432945b959a0956c56f Xuan-Son Nguyen son@huggingface.co 2025-09-06T18:35:04+07:00 GitHub noreply@github.com 2025-09-06T13:35:04+02:00 server : implement prompt processing progress report in stream mode (#15827)
01806e77714ae8a78130d432945b959a0956c56f 186415d59552bd5c90549cd8e9be3cc287621fd9 Johannes Gäßler johannesg@5d6.de 2025-09-06T13:28:44+02:00 GitHub noreply@github.com 2025-09-06T13:28:44+02:00 ggml-cpu: document use of "free" memory [no ci] (#15834)
186415d59552bd5c90549cd8e9be3cc287621fd9 fd621880f3fd908424e675a41715a2dc760247a2 Aaron Teo aaron.teo1@ibm.com 2025-09-06T11:27:28+08:00 GitHub noreply@github.com 2025-09-06T11:27:28+08:00 ggml-cpu: drop support for nnpa intrinsics (#15821)
fd621880f3fd908424e675a41715a2dc760247a2 4281c7b315f8a904549fe527039b976e08098d1a Gabe Goodhart ghart@us.ibm.com 2025-09-05T17:32:39-06:00 GitHub noreply@github.com 2025-09-05T17:32:39-06:00 aLoRA Support (#15327)
4281c7b315f8a904549fe527039b976e08098d1a 5fac79cbc77b6d12c9feb5f34fc63586b35fd561 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-09-06T01:21:15+02:00 GitHub noreply@github.com 2025-09-06T01:21:15+02:00 ci : exempt correct research label (#15825)
5fac79cbc77b6d12c9feb5f34fc63586b35fd561 408ff524b40baf4f51a81d42a9828200dd4fcb6b Gabe Goodhart ghart@us.ibm.com 2025-09-05T14:31:24-06:00 GitHub noreply@github.com 2025-09-05T14:31:24-06:00 Thinking model disabled assistant prefill (#15404)
408ff524b40baf4f51a81d42a9828200dd4fcb6b 5143fa895e7725c5bd2135daf7d8f793d98fa91c Eric Curtin ericcurtin17@gmail.com 2025-09-05T19:43:59+01:00 GitHub noreply@github.com 2025-09-05T19:43:59+01:00 Implement --log-colors with always/never/auto (#15792)
5143fa895e7725c5bd2135daf7d8f793d98fa91c 3a550b5ca4565c9e28f63880d47840feb27d0ff6 Johannes Gäßler johannesg@5d6.de 2025-09-05T16:07:02+02:00 GitHub noreply@github.com 2025-09-05T16:07:02+02:00 CUDA: fastdiv, launch bounds for mmvq + q8_1 quant (#15802)
3a550b5ca4565c9e28f63880d47840feb27d0ff6 a81283820a466f2ace06ce4d4bc9512761f9365f Daniel Bevenius daniel.bevenius@gmail.com 2025-09-05T14:49:21+02:00 GitHub noreply@github.com 2025-09-05T13:49:21+01:00 tests : add --list-ops and --show-coverage options (#15745)
a81283820a466f2ace06ce4d4bc9512761f9365f c610b6c11b1ef7d678671dcf15acd7187a7ad8f3 Erik Scholz Green-Sky@users.noreply.github.com 2025-09-05T11:34:28+02:00 GitHub noreply@github.com 2025-09-05T11:34:28+02:00 gguf: gguf_writer refactor (#15691)
c610b6c11b1ef7d678671dcf15acd7187a7ad8f3 5d6688de08e73acc2532d668380801ed79d704eb Georgi Gerganov ggerganov@gmail.com 2025-09-05T10:39:22+03:00 GitHub noreply@github.com 2025-09-05T10:39:22+03:00 kv-cache : fix SWA checks + disable cacheless iSWA (#15811)
dcbe998a6d8f13b6c9cb36bbe7b355ced2ac3842 87932ec016f0ec81e98a13ce5212851f8c12458a Yunzhe Jia yunzhe@calculet.tech 2025-09-05T14:34:07+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-09-05T14:34:07+08:00 add calrt API used in server side
5d6688de08e73acc2532d668380801ed79d704eb 4fd1242bef6cb2325b4ff1c1a80f3b54b64508a6 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-05T04:36:23+02:00 GitHub noreply@github.com 2025-09-05T04:36:23+02:00 model-conversion : add --embeddings flag to modelcard.template [no ci] (#15801)
4fd1242bef6cb2325b4ff1c1a80f3b54b64508a6 b2426e469e2fdb6c44216d56baa4cfff4f39ae00 ExtReMLapin 3909752+ExtReMLapin@users.noreply.github.com 2025-09-05T01:24:08+02:00 GitHub noreply@github.com 2025-09-05T01:24:08+02:00 chat : fixed crash when Hermes 2 <tool_call> had a newline before it (#15639)
b2426e469e2fdb6c44216d56baa4cfff4f39ae00 9e2b1e83c68a38ea0c64f726dd979439bd02189b Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-09-05T01:22:22+02:00 GitHub noreply@github.com 2025-09-05T01:22:22+02:00 chat : nemotron thinking & toolcalling support (#15676)
9e2b1e83c68a38ea0c64f726dd979439bd02189b fb15d649ed14ab447eeab911e0c9d21e35fb243e Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-09-05T01:05:12+02:00 GitHub noreply@github.com 2025-09-05T01:05:12+02:00 scripts : add Jinja tester PySide6 simple app (#15756)
fb15d649ed14ab447eeab911e0c9d21e35fb243e 856ed0947f27b4ec3ad269fceda0402fbab263d3 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-04T18:10:29+02:00 GitHub noreply@github.com 2025-09-04T18:10:29+02:00 llama : add support for EmbeddingGemma 300m (#15798)
856ed0947f27b4ec3ad269fceda0402fbab263d3 d1e2adba65208d6de3d7f6f23b00c562ff5bb777 Gabe Goodhart ghart@us.ibm.com 2025-09-04T09:53:22-06:00 GitHub noreply@github.com 2025-09-04T18:53:22+03:00 metal : Add template specialization for mul_mm_id w/ ne20 == 10 (#15799)
d1e2adba65208d6de3d7f6f23b00c562ff5bb777 c1c354e44c06d259679bb5bb7a8fa9f0b28480e4 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-04T15:40:44+02:00 GitHub noreply@github.com 2025-09-04T15:40:44+02:00 llama : set n_outputs to 1 to avoid 0 outputs mean-pooling (#15791)
c1c354e44c06d259679bb5bb7a8fa9f0b28480e4 a68d9144262f1d0ef4f6ba7ad4a7e73e977ba78c Chenguang Li 757486878@qq.com 2025-09-04T20:20:14+08:00 GitHub noreply@github.com 2025-09-04T20:20:14+08:00 CANN: Refactor ND to NZ workspace to be per-device (#15763)
a68d9144262f1d0ef4f6ba7ad4a7e73e977ba78c badb80cadbc40e047b30c43611aba575fc8d6845 Xuan-Son Nguyen son@huggingface.co 2025-09-04T11:50:23+02:00 GitHub noreply@github.com 2025-09-04T11:50:23+02:00 server: add exceed_context_size_error type (#15780)
badb80cadbc40e047b30c43611aba575fc8d6845 0a1b3982cd0bd18730d50a693053b88c13fd04a6 Eric Curtin ecurtin@redhat.com 2025-09-04T10:49:44+01:00 GitHub noreply@github.com 2025-09-04T10:49:44+01:00 Document the new max GPU layers default in help (#15771)
0a1b3982cd0bd18730d50a693053b88c13fd04a6 5421f63ab08ca1e1a093662a5ccd0117e461185f leejet leejet714@gmail.com 2025-09-04T16:38:49+08:00 GitHub noreply@github.com 2025-09-04T10:38:49+02:00 ggml: add ops for WAN video model (cuda && cpu) (#15669)
5421f63ab08ca1e1a093662a5ccd0117e461185f 820bc9853100708011036e10f40b923968dd9b66 hipudding huafengchun@gmail.com 2025-09-04T15:12:30+08:00 GitHub noreply@github.com 2025-09-04T15:12:30+08:00 CANN: Fix precision issue on 310I DUO multi-devices (#15784)
820bc9853100708011036e10f40b923968dd9b66 239b60e8986bbcb944f57311a02ac994431bf652 rmatif rmatif@proton.me 2025-09-04T08:30:28+02:00 GitHub noreply@github.com 2025-09-03T23:30:28-07:00 opencl: add hs=40 to FA (#15758)
239b60e8986bbcb944f57311a02ac994431bf652 dff7551bfdbdd6e57c13e523d7dcca317640e907 Chenguang Li 757486878@qq.com 2025-09-04T11:03:02+08:00 GitHub noreply@github.com 2025-09-04T11:03:02+08:00 CANN: fix acl_rstd allocation size in ggml_cann_rms_norm (#15760)
dff7551bfdbdd6e57c13e523d7dcca317640e907 0fce7a1248b74148c1eb0d368b7e18e8bcb96809 Ruben Ortlam picard12@live.de 2025-09-03T22:55:10+02:00 GitHub noreply@github.com 2025-09-03T21:55:10+01:00 vulkan: fix mmv subgroup16 selection (#15775)
0fce7a1248b74148c1eb0d368b7e18e8bcb96809 8227695d7a3e5b357cb37fad263f00a8ca6db710 Jeff Bolz jbolz@nvidia.com 2025-09-03T13:33:15-05:00 GitHub noreply@github.com 2025-09-03T20:33:15+02:00 vulkan: don't use std::string in load_shaders, to improve compile time (#15724)
8227695d7a3e5b357cb37fad263f00a8ca6db710 0014fb4add3a7d000c14b763538045c6170f57d9 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-03T20:24:50+02:00 GitHub noreply@github.com 2025-09-03T20:24:50+02:00 vulkan : update ggml_vk_instance_validation_ext_available (#15666)
0014fb4add3a7d000c14b763538045c6170f57d9 661ae31c9c68201577e70278285b349a5a662caf Shin-myoung-serp relent95@naver.com 2025-09-04T03:22:55+09:00 GitHub noreply@github.com 2025-09-03T20:22:55+02:00 ggml vulkan: add hardsigmoid and hardswish operations (#15762)
661ae31c9c68201577e70278285b349a5a662caf 407c23786dd0d3a503e9429eead96e611d3950c9 Oliver Simons osimons@nvidia.com 2025-09-03T19:59:16+02:00 GitHub noreply@github.com 2025-09-03T19:59:16+02:00 CUDA: Optimize `rms_norm_f32` kernel and its fused variants, giving 1-6% perf E2E (#15715)
407c23786dd0d3a503e9429eead96e611d3950c9 cdedb70a998cea7052560fe0b0615a839443564d Daniel Bevenius daniel.bevenius@gmail.com 2025-09-03T18:28:36+02:00 GitHub noreply@github.com 2025-09-03T18:28:36+02:00 model-conversion : fix pyright errors (#15770)
cdedb70a998cea7052560fe0b0615a839443564d 2c8dac72eb6acd4e20c0da251535dfc46d35178b Georgi Gerganov ggerganov@gmail.com 2025-09-03T18:16:26+03:00 GitHub noreply@github.com 2025-09-03T18:16:26+03:00 sampling : optimize dist sampler (#15704)
2c8dac72eb6acd4e20c0da251535dfc46d35178b 40a751ea9a94364da73537b86502a808ebe1fc3a Daniel Bevenius daniel.bevenius@gmail.com 2025-09-03T13:35:49+02:00 GitHub noreply@github.com 2025-09-03T13:35:49+02:00 llama : fix incorrect model type for Gemma 270M (#15764)
40a751ea9a94364da73537b86502a808ebe1fc3a 5eae9348835037046f33fafd852df7c952c95209 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-03T12:50:47+02:00 GitHub noreply@github.com 2025-09-03T12:50:47+02:00 model-conversion : remove hardcoded /bin/bash shebangs [no ci] (#15765)
5eae9348835037046f33fafd852df7c952c95209 05c0380f2ae5c7193445e03ebc7b6de9ce49f1a6 hipudding huafengchun@gmail.com 2025-09-03T16:46:01+08:00 GitHub noreply@github.com 2025-09-03T16:46:01+08:00 CANN: Add RoPE contiguous check for 310I DUP device (#15735)
05c0380f2ae5c7193445e03ebc7b6de9ce49f1a6 8c3fdf44ecf08335942f2ba558955f55e88c7991 xctan xc-tan@outlook.com 2025-09-03T16:16:21+08:00 GitHub noreply@github.com 2025-09-03T16:16:21+08:00 ggml-cpu : optimize RVV kernels (#15720)
8c3fdf44ecf08335942f2ba558955f55e88c7991 f6da8cb86a28f0319b40d9d2a957a26a7d875f8c Daniel Bevenius daniel.bevenius@gmail.com 2025-09-03T09:48:35+02:00 GitHub noreply@github.com 2025-09-03T09:48:35+02:00 model-conversion : add missing curl script [no ci] (#15761)
f6da8cb86a28f0319b40d9d2a957a26a7d875f8c 8a2234ea0c89f212190c176d741b7742f0082582 hipudding huafengchun@gmail.com 2025-09-03T14:08:22+08:00 GitHub noreply@github.com 2025-09-03T14:08:22+08:00 CANN: Mask unsupported TRANSPOSE_1D operator (#15733)
8a2234ea0c89f212190c176d741b7742f0082582 3de008208b9b8a33f49f979097a99b4d59e6e521 Chenguang Li 757486878@qq.com 2025-09-03T10:43:53+08:00 GitHub noreply@github.com 2025-09-03T10:43:53+08:00 CANN: Fix type float_t to float (#15736)
3de008208b9b8a33f49f979097a99b4d59e6e521 69db8a52e6dd0db81ec05c581150cc1e43b8ac46 SnA1lGo 44647694+skrandy@users.noreply.github.com 2025-09-03T03:27:30+08:00 GitHub noreply@github.com 2025-09-02T21:27:30+02:00 fix: resolve unsigned int initialization warning for n_dims/size in gguf.cpp (#15754)
69db8a52e6dd0db81ec05c581150cc1e43b8ac46 c466abe1587bde458c806e2134e37750f2edf8fb Oliver Simons osimons@nvidia.com 2025-09-02T19:40:37+02:00 GitHub noreply@github.com 2025-09-03T01:40:37+08:00 chore: Update `.clang-format` to use `BinPackArguments=true` (#15744)
c466abe1587bde458c806e2134e37750f2edf8fb 0a2a3841e8ebc570da343e8cf58a21c1010b41e7 Johannes Gäßler johannesg@5d6.de 2025-09-02T18:17:26+02:00 GitHub noreply@github.com 2025-09-02T18:17:26+02:00 llama: -fa 1/0/-1 aliases for -fa on/off/auto (#15746)
0a2a3841e8ebc570da343e8cf58a21c1010b41e7 9961d244f2df6baf40af2f1ddc0927f8d91578c8 Ruben Ortlam picard12@live.de 2025-09-02T16:02:26+02:00 GitHub noreply@github.com 2025-09-02T16:02:26+02:00 vulkan: fix shaders gen when no integer dot is available (#15740)
9961d244f2df6baf40af2f1ddc0927f8d91578c8 25f1045f07cf0daf667d63e35618842e3174a8c7 hipudding huafengchun@gmail.com 2025-09-02T17:12:37+08:00 GitHub noreply@github.com 2025-09-02T17:12:37+08:00 CANN: Resolve soft_max precision issue (#15730)
25f1045f07cf0daf667d63e35618842e3174a8c7 97669e40735d08db65ef094a8faae8e8411a97db Jeff Bolz jbolz@nvidia.com 2025-09-02T01:37:01-05:00 GitHub noreply@github.com 2025-09-02T14:37:01+08:00 vulkan: Fix macro parameter order for f32 matmul shaders (#15716)
97669e40735d08db65ef094a8faae8e8411a97db 2f853687b3bce15e143a22f678d1715060fd606c rmatif rmatif@proton.me 2025-09-02T08:26:53+02:00 GitHub noreply@github.com 2025-09-01T23:26:53-07:00 opencl: add attn sinks support for FA kernels (#15706)
2f853687b3bce15e143a22f678d1715060fd606c ef2af57ddf04ab367f6ba6e8ed100fc56785e4b7 Chenguang Li 757486878@qq.com 2025-09-02T14:07:48+08:00 GitHub noreply@github.com 2025-09-02T14:07:48+08:00 CANN: Support eager execution mode under ACL graph compilation (#15712)
ef2af57ddf04ab367f6ba6e8ed100fc56785e4b7 5d804a4938d896f9089687a8b99911cdb43b0b94 hipudding huafengchun@gmail.com 2025-09-02T14:05:23+08:00 GitHub noreply@github.com 2025-09-02T14:05:23+08:00 CANN: Support ext_factor in rope (#15710)
5d804a4938d896f9089687a8b99911cdb43b0b94 d4d8dbe383e8b9600cbe8b42016e3a4529b51219 Johannes Gäßler johannesg@5d6.de 2025-09-02T01:14:55+02:00 GitHub noreply@github.com 2025-09-01T16:14:55-07:00 ggml-backend: raise GGML_MAX_SPLIT_INPUTS (#15722)
d4d8dbe383e8b9600cbe8b42016e3a4529b51219 35a42edac84690990a75726898d975cd08d220c0 Gilad S. 7817232+giladgd@users.noreply.github.com 2025-09-01T22:17:42+03:00 GitHub noreply@github.com 2025-09-01T21:17:42+02:00 vulkan: use memory budget extension to read memory usage (#15545)
35a42edac84690990a75726898d975cd08d220c0 fec7911f8febb22c27ce1acb12441800da24cc06 Jeff Bolz jbolz@nvidia.com 2025-09-01T14:01:10-05:00 GitHub noreply@github.com 2025-09-01T21:01:10+02:00 vulkan: add missing clamps in new mul_mat_id paths (#15702)
fec7911f8febb22c27ce1acb12441800da24cc06 078ce23ea77988f2fe1a42afb18d58bc084d55fa Ruben Ortlam picard12@live.de 2025-09-01T20:58:35+02:00 GitHub noreply@github.com 2025-09-01T20:58:35+02:00 vulkan: disable large mmv subgroups on older Nvidia GPUs (#15717)
078ce23ea77988f2fe1a42afb18d58bc084d55fa a0c2b207c596d1092a08615de61ab56a7d63515f s-goto-11 206795233+s-goto-11@users.noreply.github.com 2025-09-02T03:13:49+09:00 GitHub noreply@github.com 2025-09-01T20:13:49+02:00 ggml: SVE support for exponential functions (#15145)
a0c2b207c596d1092a08615de61ab56a7d63515f 4b20d8b7e31718d3fe8b4f12d220a9d19e4f1997 Prashant Vithule 119530321+Vithulep@users.noreply.github.com 2025-09-01T23:43:16+05:30 GitHub noreply@github.com 2025-09-01T20:13:16+02:00 ggml: aarch64: Implement SVE F16 kernels for vector functions (#15115)
4b20d8b7e31718d3fe8b4f12d220a9d19e4f1997 02c1813517412f3e00aa6ca7c0273fea64edb492 Jie Fu (傅杰) jiefu@tencent.com 2025-09-01T23:53:31+08:00 GitHub noreply@github.com 2025-09-01T23:53:31+08:00 convert : remove redundant code (#15708)
02c1813517412f3e00aa6ca7c0273fea64edb492 77dee9de97be75b7143a213bc48893e0c0b29af7 Ruben Ortlam picard12@live.de 2025-09-01T16:19:07+02:00 GitHub noreply@github.com 2025-09-01T16:19:07+02:00 Vulkan: Add Integer Dot Product mul_mat_vec shader for legacy quants (#14903)
77dee9de97be75b7143a213bc48893e0c0b29af7 4795c91c32fec7165a1364763d4d4f0c93abf933 Daniel Bevenius daniel.bevenius@gmail.com 2025-09-01T14:28:49+02:00 GitHub noreply@github.com 2025-09-01T14:28:49+02:00 ggml : WebGPU add TRANSPOSE and RESHAPE to supported ops (#15695)
4795c91c32fec7165a1364763d4d4f0c93abf933 b66df9d9c942254d03209186ef24ed7c994a576e Jie Fu (傅杰) jiefu@tencent.com 2025-09-01T15:34:59+08:00 GitHub noreply@github.com 2025-09-01T10:34:59+03:00 docs : add Hunyuan to models section (#15707)
b66df9d9c942254d03209186ef24ed7c994a576e b9382c3877c6067feccf182efe9449a2d1cb24c7 Akarshan Biswas akarshan@menlo.ai 2025-09-01T06:55:06+05:30 GitHub noreply@github.com 2025-09-01T06:55:06+05:30 CUDA: fix build error from ambiguous __half conversions in conv2d (#15690)
b9382c3877c6067feccf182efe9449a2d1cb24c7 3dc7397a2799bdc07bccf637ab7ae5a1e786d1a4 hipudding huafengchun@gmail.com 2025-09-01T08:57:23+08:00 GitHub noreply@github.com 2025-09-01T08:57:23+08:00 CANN: Optimize MUL_MAT_ID (#15658)
3dc7397a2799bdc07bccf637ab7ae5a1e786d1a4 e92d53b29e393fc4c0f9f1f7c3fe651be8d36faa hipudding huafengchun@gmail.com 2025-09-01T08:57:00+08:00 GitHub noreply@github.com 2025-09-01T08:57:00+08:00 CANN: fix RoPE cache issue on multi-device (#15629)
e92d53b29e393fc4c0f9f1f7c3fe651be8d36faa 0d161f021aa33ec0e90cce96f5d1a88925557327 Georgi Gerganov ggerganov@gmail.com 2025-08-31T20:41:02+03:00 GitHub noreply@github.com 2025-08-31T20:41:02+03:00 sampling : optimize samplers by reusing bucket sort (#15665)
0d161f021aa33ec0e90cce96f5d1a88925557327 4efd5a83163ff383285b3a4c2106feabf5c69557 Georgi Gerganov ggerganov@gmail.com 2025-08-31T20:11:58+03:00 GitHub noreply@github.com 2025-08-31T20:11:58+03:00 server : enable /slots by default and make it secure (#15630)
4efd5a83163ff383285b3a4c2106feabf5c69557 274966226f87f301ac132da898280ca3142b60e5 Georgi Gerganov ggerganov@gmail.com 2025-08-31T19:43:30+03:00 GitHub noreply@github.com 2025-08-31T19:43:30+03:00 metal : fix checks for available FA kernels (#15700)
274966226f87f301ac132da898280ca3142b60e5 9777032dccd67bdc7785aeab7497014a8be8dacc Diego Devesa slarengh@gmail.com 2025-08-31T08:47:05-07:00 GitHub noreply@github.com 2025-08-31T18:47:05+03:00 llama : fix fattn reserve call n_seqs parameter (#15699)
9777032dccd67bdc7785aeab7497014a8be8dacc 7d3c9f2b217acf0ce5db81ae83d3f375f49ab2c7 Diego Devesa slarengh@gmail.com 2025-08-31T06:49:03-07:00 GitHub noreply@github.com 2025-08-31T15:49:03+02:00 llama : separate compute buffer reserve from fattn check (#15696)
7d3c9f2b217acf0ce5db81ae83d3f375f49ab2c7 bbbf5ecccb35286521f735239d499eec4279a840 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-31T15:30:20+02:00 GitHub noreply@github.com 2025-08-31T15:30:20+02:00 ci : explicitly set fa off or on (#15692)
bbbf5ecccb35286521f735239d499eec4279a840 c37052ab4d6d1ae73c0e90bc6e560cc6409e1311 Jeff Bolz jbolz@nvidia.com 2025-08-31T03:13:27-05:00 GitHub noreply@github.com 2025-08-31T10:13:27+02:00 vulkan: handle large sizes for get_rows (#15686)
c37052ab4d6d1ae73c0e90bc6e560cc6409e1311 5c16b9c87d840e4d5d55fa83c732c6b693346f40 Jeff Bolz jbolz@nvidia.com 2025-08-31T02:06:43-05:00 GitHub noreply@github.com 2025-08-31T09:06:43+02:00 vulkan: mul_mat_id coopmat2 optimizations (#15546)
5c16b9c87d840e4d5d55fa83c732c6b693346f40 b97c9edc59d4a1b4069991aa670411190f4f3a3e Daniel Bevenius daniel.bevenius@gmail.com 2025-08-31T08:46:42+02:00 GitHub noreply@github.com 2025-08-31T08:46:42+02:00 vulkan : remove unused portability_enumeration_ext variable (#15679)
b97c9edc59d4a1b4069991aa670411190f4f3a3e 94e82c7eadeb8fff0db4bfd1ab6d8cf65fa6f2e0 Jeff Bolz jbolz@nvidia.com 2025-08-31T01:30:54-05:00 GitHub noreply@github.com 2025-08-31T08:30:54+02:00 vulkan: Allow fallback to sysmem memory when vidmem is full (#15649)
94e82c7eadeb8fff0db4bfd1ab6d8cf65fa6f2e0 4d74393bcc956ccd7df68a6a06d1a0575cfa712c Jeff Bolz jbolz@nvidia.com 2025-08-31T01:27:57-05:00 GitHub noreply@github.com 2025-08-31T08:27:57+02:00 vulkan: clamp matmul and FA results to the max finite value (#15652)
4d74393bcc956ccd7df68a6a06d1a0575cfa712c dd892555b0681b7f56d38780f6fdfe00a195160f Charles Xu charles.xu@arm.com 2025-08-30T18:03:42+02:00 GitHub noreply@github.com 2025-08-31T00:03:42+08:00 ggml: update kleidiai to v1.13.0 (#15663)
dd892555b0681b7f56d38780f6fdfe00a195160f e81b8e4b7f5ab870836fad26d154a7507b341b36 Diego Devesa slarengh@gmail.com 2025-08-30T08:51:28-07:00 GitHub noreply@github.com 2025-08-30T23:51:28+08:00 Update build.md to remove MSVC arm64 notes (#15684)
e81b8e4b7f5ab870836fad26d154a7507b341b36 38ad381f9f5d4dd368a96d844fb19cf501ed9d22 Johannes Gäßler johannesg@5d6.de 2025-08-30T16:32:10+02:00 GitHub noreply@github.com 2025-08-30T16:32:10+02:00 llama: use FA + max. GPU layers by default (#15434)
38ad381f9f5d4dd368a96d844fb19cf501ed9d22 696fccf354e9dbdfbce135bc40b44c9dcc64dda9 Johannes Gäßler johannesg@5d6.de 2025-08-30T16:20:32+02:00 GitHub noreply@github.com 2025-08-30T16:20:32+02:00 CUDA: use FP32 arithmetic for conv2d (#15683)
696fccf354e9dbdfbce135bc40b44c9dcc64dda9 ef476916bba4b44f44be0c98babc1cb025968e75 Jeff Bolz jbolz@nvidia.com 2025-08-30T04:11:22-05:00 GitHub noreply@github.com 2025-08-30T11:11:22+02:00 vulkan: Skip syncing for prealloc_y when it is reused (#15544)
ef476916bba4b44f44be0c98babc1cb025968e75 d82f6aa34a216f5df1945cdfe121ba5e6cd80be0 Chenguang Li 757486878@qq.com 2025-08-30T10:18:35+08:00 GitHub noreply@github.com 2025-08-30T10:18:35+08:00 CANN: FIx compiler warnings (#15661)
d82f6aa34a216f5df1945cdfe121ba5e6cd80be0 3d16b29c3bb1ec816ac0e782f20d169097063919 Sergey Alirzaev l29ah@riseup.net 2025-08-30T00:12:53+02:00 GitHub noreply@github.com 2025-08-30T00:12:53+02:00 server : removed obsolete doc (#15670)
3d16b29c3bb1ec816ac0e782f20d169097063919 792b44f2ed9668cce7f267ff0ae4950ed9b4a5de Johannes Gäßler johannesg@5d6.de 2025-08-29T22:04:08+02:00 GitHub noreply@github.com 2025-08-29T22:04:08+02:00 scripts: strip "AMD Instinct" from GPU name (#15668)
792b44f2ed9668cce7f267ff0ae4950ed9b4a5de 81017865ee444cf49ce0136f2be1e41a0270ff91 ExtReMLapin 3909752+ExtReMLapin@users.noreply.github.com 2025-08-29T19:25:40+02:00 GitHub noreply@github.com 2025-08-29T20:25:40+03:00 server : add documentation for `parallel_tool_calls` param (#15647)
81017865ee444cf49ce0136f2be1e41a0270ff91 60e5eee31f1af9bb579ac45380e3857d610020b9 Aman Gupta amangupta052@gmail.com 2025-08-29T21:30:06+08:00 GitHub noreply@github.com 2025-08-29T21:30:06+08:00 CUDA: fix bug in rms_norm fusion (#15660)
60e5eee31f1af9bb579ac45380e3857d610020b9 009b709d6efd24820ac67765ed339a72dc797814 Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-08-29T14:53:41+02:00 GitHub noreply@github.com 2025-08-29T14:53:41+02:00 chat : Seed OSS thinking + tool call support (#15552)
009b709d6efd24820ac67765ed339a72dc797814 e8d99dd0b67f2ecc1e45fca8074a3a18c3e036d2 Aman Gupta amangupta052@gmail.com 2025-08-29T11:35:58+08:00 GitHub noreply@github.com 2025-08-29T11:35:58+08:00 CUDA: fuse adds, fuse add with rms norm (#15631)
e8d99dd0b67f2ecc1e45fca8074a3a18c3e036d2 a8bca68f727844e7dcf24a956003b3c2039ea563 Gabe Goodhart ghart@us.ibm.com 2025-08-28T18:39:31-06:00 GitHub noreply@github.com 2025-08-28T18:39:31-06:00 nvidia nemotron nano v2 (nemotronh) (#15507)
a8bca68f727844e7dcf24a956003b3c2039ea563 c97dc093912ad014f6d22743ede0d4d7fd82365a Gabe Goodhart ghart@us.ibm.com 2025-08-28T15:27:36-05:00 GitHub noreply@github.com 2025-08-28T15:27:36-05:00 fix: Compute the full sum in llama-eval-callback, not just the sum of printed values (#15637)
c97dc093912ad014f6d22743ede0d4d7fd82365a 6c442f42ff25564a0cd6b1435d9abc1b0178eac5 mnehete32 33429707+mnehete32@users.noreply.github.com 2025-08-29T00:03:03+05:30 GitHub noreply@github.com 2025-08-28T20:33:03+02:00 CUDA: add conv2d (#15635)
6c442f42ff25564a0cd6b1435d9abc1b0178eac5 73804145ab6c06888e314046abf08f66a0133679 Aaron Teo aaron.teo1@ibm.com 2025-08-28T22:39:27+08:00 GitHub noreply@github.com 2025-08-28T22:39:27+08:00 ggml-cpu: fix invalid hsum build in debug s390x (#15634)
73804145ab6c06888e314046abf08f66a0133679 c8d0d14e77c3c45df5cbbddde9c2b4c377ec4e7a compilade git@compilade.net 2025-08-28T10:11:36-04:00 GitHub noreply@github.com 2025-08-28T10:11:36-04:00 ggml : fix SSM_SCAN for n_groups > 1 (#15625)
c8d0d14e77c3c45df5cbbddde9c2b4c377ec4e7a 84ab83cc0b4b7e769451ee48e4c7d1acef91ef25 Georgi Gerganov ggerganov@gmail.com 2025-08-28T17:09:05+03:00 GitHub noreply@github.com 2025-08-28T17:09:05+03:00 kv-cache : fix find_slot to not search for continuous slot (#15638)
84ab83cc0b4b7e769451ee48e4c7d1acef91ef25 55042b3692cb1467c9ee15c62c4a9fbf180f89e3 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-28T15:49:50+02:00 GitHub noreply@github.com 2025-08-28T15:49:50+02:00 model : jina-embeddings-v3 support (#13693)
55042b3692cb1467c9ee15c62c4a9fbf180f89e3 8a4280ce431da6b33e5a95ae1fd61472c8c3f8cc Aman Gupta amangupta052@gmail.com 2025-08-28T19:23:22+08:00 GitHub noreply@github.com 2025-08-28T19:23:22+08:00 scripts: add sqlite3 check for compare-commits.sh (#15633)
8a4280ce431da6b33e5a95ae1fd61472c8c3f8cc 64387f6e95434b393ac3df285864692b7fd9c4d2 Georgi Gerganov ggerganov@gmail.com 2025-08-28T12:27:02+03:00 GitHub noreply@github.com 2025-08-28T12:27:02+03:00 kv-cache : remove LLAMA_SET_ROWS checks (#15505)
64387f6e95434b393ac3df285864692b7fd9c4d2 d35a1e8c41f747548775225973a99507896a8c61 Aleksei Nikiforov 103434461+AlekseiNikiforovIBM@users.noreply.github.com 2025-08-28T10:56:41+02:00 GitHub noreply@github.com 2025-08-28T16:56:41+08:00 gguf-py: byteswapping improvements (#12851)
d35a1e8c41f747548775225973a99507896a8c61 46d9caa27a0281150e8cf082308c0f9e7576ebe5 Joshua Cogliati jrincayc@users.noreply.github.com 2025-08-28T01:48:20-06:00 GitHub noreply@github.com 2025-08-28T10:48:20+03:00 cli : change log to warning to explain reason for stopping (#15604)
46d9caa27a0281150e8cf082308c0f9e7576ebe5 5a0e3ef6f00c658fbae53797f02d5a360ebf8fec Daniel Bevenius daniel.bevenius@gmail.com 2025-08-28T09:26:48+02:00 GitHub noreply@github.com 2025-08-28T09:26:48+02:00 model-conversion : add mmproj conversion target (#15628)
5a0e3ef6f00c658fbae53797f02d5a360ebf8fec fbef0fad7a7c765939f6c9e322fa05cd52cf0c15 matiaslin 45382001+matiaslin@users.noreply.github.com 2025-08-27T17:32:36-07:00 GitHub noreply@github.com 2025-08-28T02:32:36+02:00 cuda: Add cublasLt_static linking when GGML_STATIC is enabled (#15622)
fbef0fad7a7c765939f6c9e322fa05cd52cf0c15 da54f9f1a2db07aaae390024ac466e7867685d94 Johannes Gäßler johannesg@5d6.de 2025-08-27T20:58:09+02:00 GitHub noreply@github.com 2025-08-27T20:58:09+02:00 server: higher timeout for tests (#15621)
da54f9f1a2db07aaae390024ac466e7867685d94 47373271f971aa5a0a6462b286f4a7b5bd4ba644 Georgi Gerganov ggerganov@gmail.com 2025-08-27T15:48:07+03:00 GitHub noreply@github.com 2025-08-27T15:48:07+03:00 presets : add qwen3-30B-a3b FIM (#15616)
47373271f971aa5a0a6462b286f4a7b5bd4ba644 1bded5a3b3376b2aa3ba2f11ade5910c550f354b uvos carl@uvos.xyz 2025-08-27T13:58:54+02:00 GitHub noreply@github.com 2025-08-27T13:58:54+02:00 HIP: Enable support for ggml_backend_cuda_register_host_buffer (#15615)
1bded5a3b3376b2aa3ba2f11ade5910c550f354b 1e7489745a74996fc36e8fd05b73aa16bc184e0c Georgi Gerganov ggerganov@gmail.com 2025-08-27T13:55:12+03:00 GitHub noreply@github.com 2025-08-27T13:55:12+03:00 kv-cache : better estimate of n_kv for multi-sequence batches (#15610)
1e7489745a74996fc36e8fd05b73aa16bc184e0c 1cf123a343ab7ca5586aacb9e0a1d2de7fe33be4 Chenguang Li 757486878@qq.com 2025-08-27T17:21:41+08:00 GitHub noreply@github.com 2025-08-27T17:21:41+08:00 CANN: refactor mask handling and improve performance in FA (#15561)
1cf123a343ab7ca5586aacb9e0a1d2de7fe33be4 fcca2182a18c786c57d02d1d1927204b133f1fdc xctan xc-tan@outlook.com 2025-08-27T16:44:22+08:00 GitHub noreply@github.com 2025-08-27T16:44:22+08:00 ggml-cpu : add basic RVV support for vector f32 ops (#15057)
fcca2182a18c786c57d02d1d1927204b133f1fdc 86076f92de1a547ef87c304facca3b4b9fae6c21 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-27T10:28:53+02:00 GitHub noreply@github.com 2025-08-27T10:28:53+02:00 common : add -m to bash completion for --model [no ci] (#15591)
86076f92de1a547ef87c304facca3b4b9fae6c21 bcbddcd54f0d5c22eab180831fdea6484107112f rmatif rmatif@proton.me 2025-08-27T08:36:05+02:00 GitHub noreply@github.com 2025-08-26T23:36:05-07:00 OpenCL: add fused group_norm/norm, mul, add (#15314)
87932ec016f0ec81e98a13ce5212851f8c12458a 0115cfb7c1dec0753aecb34f3af4338bf9de8eeb Yunzhe Jia yunzhe@calculet.tech 2025-08-27T14:29:18+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-08-27T14:29:18+08:00 fix merge problem
bcbddcd54f0d5c22eab180831fdea6484107112f 8b696861364360770e9f61a3422d32941a477824 Diego Devesa slarengh@gmail.com 2025-08-26T13:14:38-07:00 GitHub noreply@github.com 2025-08-26T22:14:38+02:00 tests : fix test-opt with GGML_BACKEND_DL (#15599)
8b696861364360770e9f61a3422d32941a477824 8ce3ff1d91245e158d98d8062cd64b0dd98dcfe3 Akarshan Biswas akarshan@menlo.ai 2025-08-27T00:27:49+05:30 GitHub noreply@github.com 2025-08-27T00:27:49+05:30 SYCL: fix rms_norm_mul_add for tensor dim not a multiple of sg_size (#15592)
8ce3ff1d91245e158d98d8062cd64b0dd98dcfe3 44b1efa41acd4df3f56ee0e46f898135ecd1a054 fidoriel 49869342+fidoriel@users.noreply.github.com 2025-08-26T20:05:50+02:00 GitHub noreply@github.com 2025-08-26T20:05:50+02:00 mtmd : fix mtmd ios build (#15579)
44b1efa41acd4df3f56ee0e46f898135ecd1a054 a6a58d64785cb458ed9de52f391aa38142d38d64 Eve 139727413+netrunnereve@users.noreply.github.com 2025-08-26T15:42:49Z GitHub noreply@github.com 2025-08-26T15:42:49Z tests: add performance test for mul mat id (#15543)
a6a58d64785cb458ed9de52f391aa38142d38d64 0373486dbc0dccbdcb3b5fdd65759d88cec06196 shalinib-ibm Shalini.Salomi.Bodapati@ibm.com 2025-08-26T21:05:25+05:30 GitHub noreply@github.com 2025-08-26T23:35:25+08:00 llamafile: PowerPC Sgemm Optimization (#15558)
0373486dbc0dccbdcb3b5fdd65759d88cec06196 62cef26ac5b6b7acb635d3dc963813b43952dc2b Georgi Gerganov ggerganov@gmail.com 2025-08-26T17:45:17+03:00 GitHub noreply@github.com 2025-08-26T17:45:17+03:00 graph : fix assert in memory-less build_attn (#15590)
62cef26ac5b6b7acb635d3dc963813b43952dc2b 8f5afa94c4f929da71f560db7c9f38ef6a783d95 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-26T16:12:29+02:00 GitHub noreply@github.com 2025-08-26T16:12:29+02:00 model-conversion : add qat-q4 quantization targets (#15588)
8f5afa94c4f929da71f560db7c9f38ef6a783d95 b3964c1e890ef8c947afb36a5124ce6fcb2136d4 Johannes Gäßler johannesg@5d6.de 2025-08-26T16:01:20+02:00 GitHub noreply@github.com 2025-08-26T16:01:20+02:00 CUDA: return -1 for nonexistent compiled arch (#15587)
b3964c1e890ef8c947afb36a5124ce6fcb2136d4 79a546220c719e6a70627b243a478ab8d84dc9e1 Georgi Gerganov ggerganov@gmail.com 2025-08-26T14:22:14+03:00 GitHub noreply@github.com 2025-08-26T14:22:14+03:00 metal : optimize FA vec for large sequences and BS <= 8 (#15566)
79a546220c719e6a70627b243a478ab8d84dc9e1 85cc1ae998e4898d9fa992cb9b8620338cee97bf Xuan-Son Nguyen son@huggingface.co 2025-08-26T12:54:19+02:00 GitHub noreply@github.com 2025-08-26T12:54:19+02:00 mtmd : support Kimi VL model (#15458)
85cc1ae998e4898d9fa992cb9b8620338cee97bf 1d8d83deaa48d4a5491820d58ff1c0d8cf9d196c Georgi Gerganov ggerganov@gmail.com 2025-08-26T12:47:00+03:00 GitHub noreply@github.com 2025-08-26T12:47:00+03:00 context : print graph stats for memory-less contexts (#15586)
1d8d83deaa48d4a5491820d58ff1c0d8cf9d196c c4e9239064a564de7b94ee2b401ae907235a8fca Georgi Gerganov ggerganov@gmail.com 2025-08-26T12:46:15+03:00 GitHub noreply@github.com 2025-08-26T12:46:15+03:00 metal : improve `MUL_MAT_ID` (#15541)
c4e9239064a564de7b94ee2b401ae907235a8fca 39842a7f73012eb42816ca4f26411782bd3da7c5 tc-mb 157115220+tc-mb@users.noreply.github.com 2025-08-26T16:05:55+08:00 GitHub noreply@github.com 2025-08-26T10:05:55+02:00 model : support MiniCPM-V 4.5 (#15575)
0115cfb7c1dec0753aecb34f3af4338bf9de8eeb 3a617cb18121a85eed17ac5585788810c5d9c1e0 0d5a470223fc90b6b6807921d68011ff06ae7f9e Yunzhe Jia yunzhe@calculet.tech 2025-08-26T15:19:56+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-08-26T15:19:56+08:00 Merge branch 'master' into dev
39842a7f73012eb42816ca4f26411782bd3da7c5 0fd90db5858e325358a5fbdaa4e327ee2a79d0d4 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-26T09:08:08+02:00 GitHub noreply@github.com 2025-08-26T09:08:08+02:00 gguf-py : remove erroneous FFN_GATE entry (#15583)
0fd90db5858e325358a5fbdaa4e327ee2a79d0d4 4c37636b3ea96f2574eeb7668b93fcc0e64b05dd Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-26T08:51:43+02:00 GitHub noreply@github.com 2025-08-26T09:51:43+03:00 metal : remove contiguous assertion for src0 in IM2COL (#15577)
4c37636b3ea96f2574eeb7668b93fcc0e64b05dd 34bdbbd7c2b70b848718e95bc48010f6aecd2816 Yoshi_likes_e4 104140648+pt13762104@users.noreply.github.com 2025-08-26T13:15:33+07:00 GitHub noreply@github.com 2025-08-26T08:15:33+02:00 Add a warning for special devices (#15563)
34bdbbd7c2b70b848718e95bc48010f6aecd2816 74f52f77f28a5ad6d6075231afcb8d1ad763ca32 Jeff Bolz jbolz@nvidia.com 2025-08-25T23:42:44-05:00 GitHub noreply@github.com 2025-08-26T06:42:44+02:00 vulkan: Remove splitting for mul_mat_id (#15568)
74f52f77f28a5ad6d6075231afcb8d1ad763ca32 f7207b0415986dd7f48447149da7de3a82338276 Qeeweew 68716978+Qeeweew@users.noreply.github.com 2025-08-26T05:21:22+08:00 GitHub noreply@github.com 2025-08-25T23:21:22+02:00 CUDA: Accelerate MXFP4 table lookup using `__byte_perm` (#15451)
f7207b0415986dd7f48447149da7de3a82338276 4d917cd4f64cc37744e76d084659475819fb0728 lhez lih@qti.qualcomm.com 2025-08-25T14:18:09-07:00 GitHub noreply@github.com 2025-08-25T14:18:09-07:00 opencl: fix support ops condition for `rms_norm` (#15560)
4d917cd4f64cc37744e76d084659475819fb0728 886b97a5d693550c2da470c091d9d27bf38398f8 Ruben Ortlam picard12@live.de 2025-08-25T17:56:59+02:00 GitHub noreply@github.com 2025-08-25T17:56:59+02:00 vulkan: fix min subgroup 16 condition for mmid subgroup optimization (#15565)
886b97a5d693550c2da470c091d9d27bf38398f8 111f8d06f0b9169059779a35f655b343242ccbb6 Jeff Bolz jbolz@nvidia.com 2025-08-25T10:47:16-05:00 GitHub noreply@github.com 2025-08-25T10:47:16-05:00 tests: Generate unique input values for count_equal (#15487)
111f8d06f0b9169059779a35f655b343242ccbb6 5eff6ec9b1220b599a43b594b1110487ab6aca08 Ihar Hrachyshka ihar.hrachyshka@gmail.com 2025-08-25T11:27:34-04:00 GitHub noreply@github.com 2025-08-25T18:27:34+03:00 metal: fix regression when no metal devices are present (#15531)
5eff6ec9b1220b599a43b594b1110487ab6aca08 dfd9b5f6c7586c88588f06a644c131bec071a0a1 Johannes Gäßler johannesg@5d6.de 2025-08-25T17:23:40+02:00 GitHub noreply@github.com 2025-08-25T17:23:40+02:00 CUDA: MoE helper in device code, better tile sizes (#15525)
dfd9b5f6c7586c88588f06a644c131bec071a0a1 5a6bc6b1a6cb665a944426c2055794950e524bf5 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-25T15:00:43+02:00 GitHub noreply@github.com 2025-08-25T15:00:43+02:00 model-conversion : set pooling type to none in logits.cpp (#15564)
5a6bc6b1a6cb665a944426c2055794950e524bf5 6b64f74b55628e4193f4fb00313f07dbd8556528 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-25T14:25:25+02:00 GitHub noreply@github.com 2025-08-25T14:25:25+02:00 model-conversion : add model card template for embeddings [no ci] (#15557)
6b64f74b55628e4193f4fb00313f07dbd8556528 0d5a470223fc90b6b6807921d68011ff06ae7f9e Georgi Gerganov ggerganov@gmail.com 2025-08-25T13:56:43+03:00 GitHub noreply@github.com 2025-08-25T13:56:43+03:00 batched-bench : fix unified KV cache handling + pp timing (#15562)
0d5a470223fc90b6b6807921d68011ff06ae7f9e b0ba31f525a2ad7fb539660be0c8f088c9869f6a Weizhao Ouyang o451686892@gmail.com 2025-08-25T17:15:06+08:00 GitHub noreply@github.com 2025-08-25T11:15:06+02:00 origin/master, origin/HEAD, master convert : update Ernie 4.5 dense architecture name (#15555)
b0ba31f525a2ad7fb539660be0c8f088c9869f6a 7da9fed0d6f1750a8783436f6f313b87e76e6378 Georgi Gerganov ggerganov@gmail.com 2025-08-25T10:14:48+03:00 GitHub noreply@github.com 2025-08-25T10:14:48+03:00 metal : add FA kernels for HS=40 (#15559)
7da9fed0d6f1750a8783436f6f313b87e76e6378 c247d06f38fc09059c9607a28aa44f5ff6be208d RunningLeon mnsheng@yeah.net 2025-08-25T14:32:16+08:00 GitHub noreply@github.com 2025-08-25T08:32:16+02:00 convert : support interns1-mini (#15412)
c247d06f38fc09059c9607a28aa44f5ff6be208d 043fb27d3808766d8ea8195bbd12359727264402 Chenguang Li 757486878@qq.com 2025-08-25T10:32:21+08:00 GitHub noreply@github.com 2025-08-25T10:32:21+08:00 CANN: ROPE cache sin/cos repeat (#15501)
043fb27d3808766d8ea8195bbd12359727264402 b730706a49e576fb882dc34d9966345778b3ab0b Ruben Ortlam picard12@live.de 2025-08-24T19:36:36+02:00 GitHub noreply@github.com 2025-08-24T19:36:36+02:00 vulkan: apply MUL_MAT_ID subgroup optimization to non-coopmat devices (#15524)
b730706a49e576fb882dc34d9966345778b3ab0b c9a24fb93208fbbd3da6d903eb75431bfa97e59e Georgi Gerganov ggerganov@gmail.com 2025-08-24T13:07:07+03:00 GitHub noreply@github.com 2025-08-24T13:07:07+03:00 kv-cache : support layer reuse (#15504)
c9a24fb93208fbbd3da6d903eb75431bfa97e59e a9c6ffcbfacee092bfaaa400306fceda18199737 Jeff Bolz jbolz@nvidia.com 2025-08-24T04:24:25-05:00 GitHub noreply@github.com 2025-08-24T11:24:25+02:00 vulkan: Support FA with any multiple of 8 head sizes (#15537)
a9c6ffcbfacee092bfaaa400306fceda18199737 e78cf0d4b1bdbbc2479f11d58ce0c8f51f755875 Ruben Ortlam picard12@live.de 2025-08-24T10:48:53+02:00 GitHub noreply@github.com 2025-08-24T10:48:53+02:00 vulkan: enable Conv2D for Apple after MoltenVK fixed the bug (#15526)
e78cf0d4b1bdbbc2479f11d58ce0c8f51f755875 710dfc465a68f7443b87d9f792cffba00ed739fe Jeff Bolz jbolz@nvidia.com 2025-08-24T03:48:21-05:00 GitHub noreply@github.com 2025-08-24T10:48:21+02:00 vulkan: workaround MoltenVK compile failure in multi_add (#15506)
710dfc465a68f7443b87d9f792cffba00ed739fe 611f419cff11e4952228162a1c44cb35dff2274a Johannes Gäßler johannesg@5d6.de 2025-08-23T21:37:06+02:00 GitHub noreply@github.com 2025-08-23T21:37:06+02:00 CUDA: fix half2 -> half conversion for HIP (#15529)
611f419cff11e4952228162a1c44cb35dff2274a b1afcab804e3281867a5471fbd701e32eb32e512 Jeff Bolz jbolz@nvidia.com 2025-08-23T13:16:17-05:00 GitHub noreply@github.com 2025-08-23T13:16:17-05:00 vulkan: optimize rms_norm, and allow the work to spread across multiple SMs (#15281)
b1afcab804e3281867a5471fbd701e32eb32e512 9ef536907de1b50c30e0369284898d30472a755a Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-08-23T15:21:52+02:00 GitHub noreply@github.com 2025-08-23T15:21:52+02:00 model : add support for Seed-OSS (#15490)
9ef536907de1b50c30e0369284898d30472a755a 21dc4ddaf21b8ed551d717e7606abd2cffbacdbf Johannes Gäßler johannesg@5d6.de 2025-08-23T12:58:58+02:00 GitHub noreply@github.com 2025-08-23T13:58:58+03:00 scripts: fix compare-llama-bench.py (#15521)
21dc4ddaf21b8ed551d717e7606abd2cffbacdbf 289bf4113ef5c02d8f5eb0cf2d86683d8b8bc4d9 LaffeyNyaa 112215776+LaffeyNyaa@users.noreply.github.com 2025-08-23T16:38:30+08:00 GitHub noreply@github.com 2025-08-23T10:38:30+02:00 chat : fix debug build assertion in trim function (#15520)
289bf4113ef5c02d8f5eb0cf2d86683d8b8bc4d9 b55f06e1aa67fb10e89f53e31bbccf37eb2678ea Jeff Bolz jbolz@nvidia.com 2025-08-23T02:33:36-05:00 GitHub noreply@github.com 2025-08-23T09:33:36+02:00 vulkan: Rewrite synchronization to allow some overlap between nodes (#15489)
b55f06e1aa67fb10e89f53e31bbccf37eb2678ea 0a9b43e507a359ca392c037cf341f55137ad0b69 R0CKSTAR yeahdongcn@gmail.com 2025-08-23T14:58:57+08:00 GitHub noreply@github.com 2025-08-23T08:58:57+02:00 vulkan.Dockerfile: install vulkan SDK using tarball (#15282)
0a9b43e507a359ca392c037cf341f55137ad0b69 330c3d2d21b55bca5517db7d2eea2ea8f131df4a Acly aclysia@gmail.com 2025-08-23T08:35:21+02:00 GitHub noreply@github.com 2025-08-23T08:35:21+02:00 vulkan : support ggml_mean (#15393)
330c3d2d21b55bca5517db7d2eea2ea8f131df4a e92734d51bcb82cc35f0a6b5a14928f0036b2c90 Jeff Bolz jbolz@nvidia.com 2025-08-23T01:31:54-05:00 GitHub noreply@github.com 2025-08-23T08:31:54+02:00 vulkan: optimize mul_mat_id loading row ids into shared memory (#15427)
e92734d51bcb82cc35f0a6b5a14928f0036b2c90 45363632cbd593537d541e81b600242e0b3d47fc Johannes Gäßler johannesg@5d6.de 2025-08-22T23:47:01+02:00 GitHub noreply@github.com 2025-08-22T23:47:01+02:00 test-opt: allow slight inprecision (#15503)
45363632cbd593537d541e81b600242e0b3d47fc 32732f2459a598606055f0403f0e4ec148d06d68 Reese Levine reeselevine1@gmail.com 2025-08-22T11:28:03-07:00 GitHub noreply@github.com 2025-08-22T11:28:03-07:00 ggml WebGPU: add support for quantization types (#15440)
32732f2459a598606055f0403f0e4ec148d06d68 92f7f0a53cf6484e16a8084ed90807c35a164809 Aldehir Rojas hello@alde.dev 2025-08-22T11:04:08-05:00 GitHub noreply@github.com 2025-08-22T11:04:08-05:00 model : gpt-oss add response_format support (#15494)
92f7f0a53cf6484e16a8084ed90807c35a164809 b1ab91821f980f8993423c3f2a82a0a0f60c09d2 rmatif rmatif@proton.me 2025-08-22T15:33:15+02:00 GitHub noreply@github.com 2025-08-22T15:33:15+02:00 ggml: add `conv3d` op (#15182)
b1ab91821f980f8993423c3f2a82a0a0f60c09d2 9ebebef62fd0adf8685874f154e227ea87b7c6f4 Yavor Ivanov yavorgenadiev@gmail.com 2025-08-22T14:06:29+03:00 GitHub noreply@github.com 2025-08-22T13:06:29+02:00 cuda : add Pad Reflect 1D support (#14659)
9ebebef62fd0adf8685874f154e227ea87b7c6f4 ad5c975c2d0297124fad210776ef8eed6b90d578 Georgi Gerganov ggerganov@gmail.com 2025-08-22T12:22:13+03:00 GitHub noreply@github.com 2025-08-22T12:22:13+03:00 llama : remove KV cache defragmentation logic (#15473)
ad5c975c2d0297124fad210776ef8eed6b90d578 4afb0a746f22abaa545b3ebdb76a400d7da3a713 Aaron Teo aaron.teo1@ibm.com 2025-08-22T16:11:04+08:00 GitHub noreply@github.com 2025-08-22T16:11:04+08:00 ggml-cpu: Support Q5_0 and Q5_1 on s390x (#15486)
4afb0a746f22abaa545b3ebdb76a400d7da3a713 e288693669cf9d0a71e2f2b8bd57305f06340257 65a 10104049+65a@users.noreply.github.com 2025-08-22T08:10:14Z GitHub noreply@github.com 2025-08-22T10:10:14+02:00 server : Support multimodal completion and embeddings prompts in JSON format (#15108)
e288693669cf9d0a71e2f2b8bd57305f06340257 a0f98dd604d34826eb5ea2560d1e23fe726921df Tarek Dakhran tarek@liquid.ai 2025-08-22T09:29:08+02:00 GitHub noreply@github.com 2025-08-22T09:29:08+02:00 readme : model : mtdm : lfm2 improvements (#15476)
a0f98dd604d34826eb5ea2560d1e23fe726921df 54a241f505d515d625767b993bfd573ecee306b9 Chenguang Li 757486878@qq.com 2025-08-22T14:12:07+08:00 GitHub noreply@github.com 2025-08-22T14:12:07+08:00 CANN: Optimize RMS_NORM using cache (#15419)
54a241f505d515d625767b993bfd573ecee306b9 cd36b5e5c7fed2a3ac671dd542d579ca40b48b54 Diego Devesa slarengh@gmail.com 2025-08-21T14:09:32-07:00 GitHub noreply@github.com 2025-08-21T23:09:32+02:00 sched : fix possible use of wrong ids tensor when offloading moe prompt processing (#15488)
cd36b5e5c7fed2a3ac671dd542d579ca40b48b54 3f196be84b1376945737163e35a91e29e3e24d2f Georgi Gerganov ggerganov@gmail.com 2025-08-21T19:13:45+03:00 GitHub noreply@github.com 2025-08-21T19:13:45+03:00 llama : remove deprecated llama_kv_self API (#15472)
3f196be84b1376945737163e35a91e29e3e24d2f 97ae5961a4d9ff9c60f51bac304b97de18e75eaf Georgi Gerganov ggerganov@gmail.com 2025-08-21T18:44:45+03:00 GitHub noreply@github.com 2025-08-21T18:44:45+03:00 graph : remove build_attn_with_sinks overload (#15469)
97ae5961a4d9ff9c60f51bac304b97de18e75eaf 20c2dac8c6e05f2ad5295ddef1aebaf2d266090e Acly aclysia@gmail.com 2025-08-21T17:01:51+02:00 GitHub noreply@github.com 2025-08-21T17:01:51+02:00 vulkan : support conv_2d_dw with f16 weights (#15392)
20c2dac8c6e05f2ad5295ddef1aebaf2d266090e 96452a3fa426de83494fb0268a636aa1bfe557fe Dong Won Kim 63934649+ddwkim@users.noreply.github.com 2025-08-22T00:00:16+09:00 GitHub noreply@github.com 2025-08-21T17:00:16+02:00 vulkan: add exp operation (#15456)
96452a3fa426de83494fb0268a636aa1bfe557fe 9ad5e60dba38a6718366b7ac43e7d8e8abdc36c9 Jeff Bolz jbolz@nvidia.com 2025-08-21T09:55:00-05:00 GitHub noreply@github.com 2025-08-21T16:55:00+02:00 vulkan: Reuse conversion results in prealloc_y (#15410)
9ad5e60dba38a6718366b7ac43e7d8e8abdc36c9 715a6db02ccb16284837885f2c6fab05d8f7a6ee Jie Fu (傅杰) jiefu@tencent.com 2025-08-21T22:53:13+08:00 GitHub noreply@github.com 2025-08-21T16:53:13+02:00 examples : fix some typos in examples/model-conversion/README.md (#15477)
715a6db02ccb16284837885f2c6fab05d8f7a6ee ad294df03ff2dccd227c3fee653166f3d78b23a4 Georgi Gerganov ggerganov@gmail.com 2025-08-21T17:00:33+03:00 GitHub noreply@github.com 2025-08-21T17:00:33+03:00 kv-cache : drop the "unified" prefix (#15467)
ad294df03ff2dccd227c3fee653166f3d78b23a4 029bb39eb1e7ae0e5df817ce97931abcd5fa52a9 Jie Fu (傅杰) jiefu@tencent.com 2025-08-21T21:42:34+08:00 GitHub noreply@github.com 2025-08-21T15:42:34+02:00 examples : install torch-cpu for model conversion tool/example (#15475)
029bb39eb1e7ae0e5df817ce97931abcd5fa52a9 30649cab657d87ac46692332a76e1b75d5d22e00 Ali Tariq alitariq4589@gmail.com 2025-08-21T17:52:16+05:00 GitHub noreply@github.com 2025-08-21T14:52:16+02:00 ci : enable RVV1.0 native build (#15386)
30649cab657d87ac46692332a76e1b75d5d22e00 2758fa10dab0556e6c3f130e664750fd6773dc7c Georgi Gerganov ggerganov@gmail.com 2025-08-21T13:42:55+03:00 GitHub noreply@github.com 2025-08-21T13:42:55+03:00 ci : continue file download with wget (#15471)
2758fa10dab0556e6c3f130e664750fd6773dc7c b108e429043ee5c9fc8fa4957a0a52c3e490d5c9 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-21T12:16:54+02:00 GitHub noreply@github.com 2025-08-21T12:16:54+02:00 examples : add model conversion tool/example (#15455)
b108e429043ee5c9fc8fa4957a0a52c3e490d5c9 245be739df942861ddc52331b095b40f18e2a3f1 Michael Giba michaelgiba@gmail.com 2025-08-21T05:06:46-05:00 GitHub noreply@github.com 2025-08-21T12:06:46+02:00 ci : fix -Werror=return-type in clip.cpp so ci/run.sh can run without issue (#15221)
245be739df942861ddc52331b095b40f18e2a3f1 b2caf67db1208fd38a0570785c39f7370d906d8a Copilot 198982749+Copilot@users.noreply.github.com 2025-08-21T11:47:52+02:00 GitHub noreply@github.com 2025-08-21T11:47:52+02:00 ci : add copilot-instructions.md (#15286)
b2caf67db1208fd38a0570785c39f7370d906d8a 2f3dbffb17ef782edfd50e5a130cec6e8a7e47f8 Julien Denize 40604584+juliendenize@users.noreply.github.com 2025-08-21T11:19:50+02:00 GitHub noreply@github.com 2025-08-21T11:19:50+02:00 convert : make Mistral community chat templates optional via parameter (#15420)
2f3dbffb17ef782edfd50e5a130cec6e8a7e47f8 945e1f12a6b586ebf82fa4fd7f347225e58174c5 Jie Fu (傅杰) jiefu@tencent.com 2025-08-21T16:54:34+08:00 GitHub noreply@github.com 2025-08-21T11:54:34+03:00 common : fix incorrect print of non-ascii characters in the logging (#15466)
945e1f12a6b586ebf82fa4fd7f347225e58174c5 1b0db8f6e08951969e2447c2d18bf638effb8f75 Xuan-Son Nguyen son@huggingface.co 2025-08-21T07:32:26+02:00 GitHub noreply@github.com 2025-08-21T08:32:26+03:00 ggml : fix condition of im2col on Metal backend (#15460)
1b0db8f6e08951969e2447c2d18bf638effb8f75 29f538ac630d6544406a0702476e36808a6bd1b3 stduhpf stephduh@live.fr 2025-08-21T07:19:22+02:00 GitHub noreply@github.com 2025-08-21T08:19:22+03:00 server : fix webui (#15462)
29f538ac630d6544406a0702476e36808a6bd1b3 8ad038c0fdc719ced9fbf921a02cbae9ad79287f Daniel Bevenius daniel.bevenius@gmail.com 2025-08-21T06:12:28+02:00 GitHub noreply@github.com 2025-08-21T06:12:28+02:00 examples : remove references to `make` in examples [no ci] (#15457)
8ad038c0fdc719ced9fbf921a02cbae9ad79287f 5682a3745f2b653dcb855d5766d8edc318fb3336 R0CKSTAR yeahdongcn@gmail.com 2025-08-21T11:06:05+08:00 GitHub noreply@github.com 2025-08-21T11:06:05+08:00 musa: add GGML_UNUSED_VARS (#15446)
5682a3745f2b653dcb855d5766d8edc318fb3336 1bc664a26a1d93a48baf2483a7ff95291ca8bcb8 Diego Devesa slarengh@gmail.com 2025-08-20T16:35:28-07:00 GitHub noreply@github.com 2025-08-21T01:35:28+02:00 sched : copy only the used experts when offloading prompt processing (#15346)
1bc664a26a1d93a48baf2483a7ff95291ca8bcb8 13aeb7aef284daed4ad07dc14b4b3e2a42b5ea97 teo TeoZosa@users.noreply.github.com 2025-08-21T07:10:08+09:00 GitHub noreply@github.com 2025-08-21T00:10:08+02:00 server: fix OpenAI API compatibility for usage statistics in chat streams (#15444)
13aeb7aef284daed4ad07dc14b4b3e2a42b5ea97 7a6e91ad26160dd6dfb33d29ac441617422f28e7 Johannes Gäßler johannesg@5d6.de 2025-08-20T23:14:14+02:00 GitHub noreply@github.com 2025-08-20T23:14:14+02:00 CUDA: refactor FA support/selection code (#15454)
7a6e91ad26160dd6dfb33d29ac441617422f28e7 fec9519802ae1567048abb126cdd5ea160a22d0f Johannes Gäßler johannesg@5d6.de 2025-08-20T16:58:49+02:00 GitHub noreply@github.com 2025-08-20T16:58:49+02:00 CUDA: replace GGML_CUDA_F16 with CUDA arch checks (#15433)
fec9519802ae1567048abb126cdd5ea160a22d0f 657b8a77bd01854f99d37a47318fa24f2e7e298f Jeff Bolz jbolz@nvidia.com 2025-08-20T09:33:14-05:00 GitHub noreply@github.com 2025-08-20T16:33:14+02:00 vulkan: shorten pipeline name strings (#15431)
657b8a77bd01854f99d37a47318fa24f2e7e298f ec5ab1a36c11dd3efcf4ec8d1ac89a13a8117bc3 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-20T14:26:01+02:00 GitHub noreply@github.com 2025-08-20T14:26:01+02:00 chat: handle gpt-oss return/end token inconsistency (#15421)
ec5ab1a36c11dd3efcf4ec8d1ac89a13a8117bc3 1a99c2d948209d9ea5eac8b6dc0a297107244540 Jie Fu (傅杰) fujie_email@sina.com 2025-08-20T18:33:30+08:00 GitHub noreply@github.com 2025-08-20T13:33:30+03:00 common : fix context shift help message (#15448)
1a99c2d948209d9ea5eac8b6dc0a297107244540 37f10f955f70e0158d50343d0b9a3f92d194daae xiaobing318 71554036+xiaobing318@users.noreply.github.com 2025-08-20T18:32:05+08:00 GitHub noreply@github.com 2025-08-20T13:32:05+03:00 cmake : fix target include directories (#15450)
37f10f955f70e0158d50343d0b9a3f92d194daae 2f37014073f4c6ddc8f241c927db87337c71aa52 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-20T12:31:16+02:00 GitHub noreply@github.com 2025-08-20T13:31:16+03:00 make : remove make in favor of CMake (#15449)
2f37014073f4c6ddc8f241c927db87337c71aa52 a094f381432d92c4bf92d2d6167284316ba73a62 Georgi Gerganov ggerganov@gmail.com 2025-08-20T13:30:46+03:00 GitHub noreply@github.com 2025-08-20T13:30:46+03:00 lookahead : add sample command to readme (#15447)
a094f381432d92c4bf92d2d6167284316ba73a62 fb22dd07a639e81c7415e30b146f545f1a2f2caf R0CKSTAR yeahdongcn@gmail.com 2025-08-20T10:17:37+08:00 GitHub noreply@github.com 2025-08-20T10:17:37+08:00 musa: fix build warnings (#15258)
fb22dd07a639e81c7415e30b146f545f1a2f2caf 9ef6b0b835450ee10cd7be934ad8aef681dc1f43 lhez lih@qti.qualcomm.com 2025-08-20T02:25:51+08:00 GitHub noreply@github.com 2025-08-19T11:25:51-07:00 opencl: mark `argsort` unsupported if cols exceed workgroup limit (#15375)
9ef6b0b835450ee10cd7be934ad8aef681dc1f43 1e19f5d462b6df490a13103ca555d851f47e5fa5 Georgi Gerganov ggerganov@gmail.com 2025-08-19T19:58:28+03:00 GitHub noreply@github.com 2025-08-19T19:58:28+03:00 model : add gpt-oss type strings (#15424)
1e19f5d462b6df490a13103ca555d851f47e5fa5 d2fcd91cf96b46f4485ce46b4e3a32bf0df37715 Gian-Carlo Pascutto gcp@sjeng.org 2025-08-19T18:58:14+02:00 GitHub noreply@github.com 2025-08-19T19:58:14+03:00 common : Add top-nsigma sampler to help globally (#15428)
d2fcd91cf96b46f4485ce46b4e3a32bf0df37715 a6d3cfe7fa6ea1fb0e1ba8243b731db22ddc0b49 Georgi Gerganov ggerganov@gmail.com 2025-08-19T16:46:37+03:00 GitHub noreply@github.com 2025-08-19T16:46:37+03:00 server : disable context shift by default (#15416)
a6d3cfe7fa6ea1fb0e1ba8243b731db22ddc0b49 67f09a3a27db443f9870aac87e163dba0d08131e SHUAI YANG shuaiyang047@163.com 2025-08-19T21:28:22+08:00 GitHub noreply@github.com 2025-08-19T21:28:22+08:00 CANN: optimize rope operator (#15335)
67f09a3a27db443f9870aac87e163dba0d08131e 6424594c56f4dbd0573455d89a0d89a0ac093d13 R0CKSTAR yeahdongcn@gmail.com 2025-08-19T18:33:47+08:00 GitHub noreply@github.com 2025-08-19T12:33:47+02:00 musa: handle __hgt2_mask, available starting from MUSA SDK rc4.3.0 (#15413)
6424594c56f4dbd0573455d89a0d89a0ac093d13 e9288e886970884f288533cd597b3798995b4099 Marvin Gießing marvin.giessing@gmail.com 2025-08-19T10:54:31+02:00 GitHub noreply@github.com 2025-08-19T11:54:31+03:00 ggml-cpu: add mxfp4 VSX intrinsics for Power9+ (ppc64le) hardware (#15385)
e9288e886970884f288533cd597b3798995b4099 9d262f4bad0d37838100133537aaf0a83835ed12 Xuan-Son Nguyen son@huggingface.co 2025-08-19T10:29:36+02:00 GitHub noreply@github.com 2025-08-19T10:29:36+02:00 chat : clarify the meaning of reasoning_format (#15408)
9d262f4bad0d37838100133537aaf0a83835ed12 f0d3c7405c323784a60f14ddddfbac3f7404d417 Georgi Gerganov ggerganov@gmail.com 2025-08-19T08:45:26+03:00 GitHub noreply@github.com 2025-08-19T08:45:26+03:00 server : remove swa_full warning (#15399)
f0d3c7405c323784a60f14ddddfbac3f7404d417 f08c4c0d8d0cb6caaf8b7ad316039232b9fa059c Georgi Gerganov ggerganov@gmail.com 2025-08-19T08:45:12+03:00 GitHub noreply@github.com 2025-08-19T08:45:12+03:00 batched-bench : use rand tokens (#15398)
f08c4c0d8d0cb6caaf8b7ad316039232b9fa059c 6d7f1117e3e3285d0c5c11b5ebb0439e27920082 Xuan-Son Nguyen son@huggingface.co 2025-08-18T22:53:52+02:00 GitHub noreply@github.com 2025-08-18T22:53:52+02:00 mtmd : clean up clip_n_output_tokens (#15391)
6d7f1117e3e3285d0c5c11b5ebb0439e27920082 60212f1ead2dce9bf1ac69633a7069258ae604d8 Georgi Gerganov ggerganov@gmail.com 2025-08-18T22:02:50+03:00 Georgi Gerganov ggerganov@gmail.com 2025-08-18T22:06:44+03:00 codeowners : remove mmv.*
60212f1ead2dce9bf1ac69633a7069258ae604d8 f0c541d315e97b297b3421c52ebde53340ee66b3 Georgi Gerganov ggerganov@gmail.com 2025-08-18T22:02:11+03:00 Georgi Gerganov ggerganov@gmail.com 2025-08-18T22:06:44+03:00 sync : ggml
f0c541d315e97b297b3421c52ebde53340ee66b3 baa9255a45105d2d3b4ec432af13b7a6eda3ff35 Georgi Gerganov ggerganov@gmail.com 2025-08-18T20:35:47+03:00 Georgi Gerganov ggerganov@gmail.com 2025-08-18T22:06:44+03:00 scripts : update sync scripts
baa9255a45105d2d3b4ec432af13b7a6eda3ff35 3007baf201e7ffcda17dbdb0335997fa50a6595b Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-18T19:30:17+02:00 GitHub noreply@github.com 2025-08-18T19:30:17+02:00 llama : merge conts and reshapes and remove unnecessary cont (#15380)
3007baf201e7ffcda17dbdb0335997fa50a6595b d1d82416006e7ff41780cb0e9b5f28d30a267497 Georgi Gerganov ggerganov@gmail.com 2025-08-18T18:11:44+03:00 GitHub noreply@github.com 2025-08-18T18:11:44+03:00 readme : update hot topics (#15397)
d1d82416006e7ff41780cb0e9b5f28d30a267497 618575c5825d7d4f170e686e772178d2aae148ae davidef davidef1986@gmail.com 2025-08-18T16:51:42+02:00 GitHub noreply@github.com 2025-08-18T17:51:42+03:00 server : fix incoming tasks not process in order (#15395)
618575c5825d7d4f170e686e772178d2aae148ae f44f7931729022c57319a0124931120a169e0da9 Dobri Danchev 12420863+danchev@users.noreply.github.com 2025-08-18T05:50:48-05:00 GitHub noreply@github.com 2025-08-18T12:50:48+02:00 Fix broken build: require updated pip to support --break-system-packages (#15357)
f44f7931729022c57319a0124931120a169e0da9 ae532eac2c1df1d8edc3d2719145895b966de1bf compilade git@compilade.net 2025-08-18T03:23:56-04:00 GitHub noreply@github.com 2025-08-18T09:23:56+02:00 ggml-quants : fix make_qp_quants NANs and IQ1 assertion errors (#15379)
ae532eac2c1df1d8edc3d2719145895b966de1bf e5155e698645242d4f019267ecc40ea9bad81b09 Jeff Bolz jbolz@nvidia.com 2025-08-18T00:56:29-05:00 GitHub noreply@github.com 2025-08-18T07:56:29+02:00 vulkan: disable spirv-opt for bfloat16 shaders (#15352)
e5155e698645242d4f019267ecc40ea9bad81b09 21c17b5befc5f6be5992bc87fc1ba99d388561df Oleksandr Kuvshynov 661042+okuvshynov@users.noreply.github.com 2025-08-17T18:28:58-04:00 GitHub noreply@github.com 2025-08-18T00:28:58+02:00 server : export max observed n_past value (#15361)
21c17b5befc5f6be5992bc87fc1ba99d388561df 19f4decae0ead52debe56095ba8d693b4f14e4df Jeff Bolz jbolz@nvidia.com 2025-08-17T11:08:57-05:00 GitHub noreply@github.com 2025-08-17T18:08:57+02:00 vulkan: Use larger workgroups for mul_mat_vec when M is small (#15355)
19f4decae0ead52debe56095ba8d693b4f14e4df 4d196981d4db79e0105b939eaa7ecd40385b721c Dong Won Kim 63934649+ddwkim@users.noreply.github.com 2025-08-17T23:03:09+09:00 GitHub noreply@github.com 2025-08-17T16:03:09+02:00 vulkan: support sqrt (#15370)
4d196981d4db79e0105b939eaa7ecd40385b721c b143fbc87af0324aa49e16cd91faf3ba8bb22231 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-17T14:47:42+02:00 GitHub noreply@github.com 2025-08-17T14:47:42+02:00 convert : force patch_embd weights to F16 or F32 to avoid broken GGUFs (#15367)
b143fbc87af0324aa49e16cd91faf3ba8bb22231 de5627910df74298c998e6bb36ee3217375a5719 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-17T13:30:23+02:00 GitHub noreply@github.com 2025-08-17T13:30:23+02:00 ci : fix hang in windows-hip build/release (#15365)
de5627910df74298c998e6bb36ee3217375a5719 65349f26f2299e06477ec8e85e46243046801358 Jeff Bolz jbolz@nvidia.com 2025-08-17T03:41:45-05:00 GitHub noreply@github.com 2025-08-17T10:41:45+02:00 vulkan: Optimize argsort (#15354)
65349f26f2299e06477ec8e85e46243046801358 1fe00296f587dfca0957e006d146f5875b61e43d Tarek Dakhran t.dakhran@gmail.com 2025-08-16T23:33:54+02:00 GitHub noreply@github.com 2025-08-16T23:33:54+02:00 model : support vision LiquidAI LFM2-VL family (#15347)
1fe00296f587dfca0957e006d146f5875b61e43d de2192794f4e8e04f2e8167ef2424905145e88fc Jeff Bolz jbolz@nvidia.com 2025-08-16T11:48:22-05:00 GitHub noreply@github.com 2025-08-16T11:48:22-05:00 vulkan: fuse adds (#15252)
de2192794f4e8e04f2e8167ef2424905145e88fc 2e2b22ba6607414a5d619ac6d2f034b5b02214e5 Jeff Bolz jbolz@nvidia.com 2025-08-16T04:18:31-05:00 GitHub noreply@github.com 2025-08-16T11:18:31+02:00 vulkan: Support mul_mat_id with f32 accumulators (#15337)
2e2b22ba6607414a5d619ac6d2f034b5b02214e5 912ff8c119f01ae029543c7fdf7a84f91a0437a3 Jeff Bolz jbolz@nvidia.com 2025-08-16T03:58:38-05:00 GitHub noreply@github.com 2025-08-16T10:58:38+02:00 vulkan: Add missing bounds checking to scalar/coopmat1 mul_mat_id (#15334)
912ff8c119f01ae029543c7fdf7a84f91a0437a3 5e6229a8409ac786e62cb133d09f1679a9aec13e rmatif kingrealriadh@gmail.com 2025-08-16T10:05:55+02:00 GitHub noreply@github.com 2025-08-16T01:05:55-07:00 OpenCL: add initial FA support (#14987)
5e6229a8409ac786e62cb133d09f1679a9aec13e e2c1bfff5305c661ac53e9d57cb732ff626a2242 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-15T19:50:52+02:00 GitHub noreply@github.com 2025-08-15T19:50:52+02:00 common : fix double bos, use common_chat_templates for add_bos and add_eos (#15326)
e2c1bfff5305c661ac53e9d57cb732ff626a2242 5edf1592fdb9131d01321aeef4241c6a34969e27 lhez lih@qti.qualcomm.com 2025-08-16T00:52:14+08:00 GitHub noreply@github.com 2025-08-15T09:52:14-07:00 opencl: add initial mxfp4 support via mv (#15270)
5edf1592fdb9131d01321aeef4241c6a34969e27 db3010bd23980bad5f5d93ec3e6757ec531a413b Georgi Gerganov ggerganov@gmail.com 2025-08-15T17:16:36+03:00 GitHub noreply@github.com 2025-08-15T16:16:36+02:00 vulkan : fix out-of-bounds access in argmax kernel (#15342)
db3010bd23980bad5f5d93ec3e6757ec531a413b ff27f80a74bbe5303acd511a6781a1de6d619b3c Georgi Gerganov ggerganov@gmail.com 2025-08-15T16:28:28+03:00 GitHub noreply@github.com 2025-08-15T15:28:28+02:00 vulkan : fix compile warnings on macos (#15340)
ff27f80a74bbe5303acd511a6781a1de6d619b3c d3248d9b6557c75d59954c594bb53cf517591e91 Aaron Teo aaron.teo1@ibm.com 2025-08-15T21:11:22+08:00 GitHub noreply@github.com 2025-08-15T21:11:22+08:00 ggml: initial IBM zDNN backend (#14975)
d3248d9b6557c75d59954c594bb53cf517591e91 7aeee88cfe9c0434eac722b0f7c21404f48758a5 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-15T14:02:39+02:00 GitHub noreply@github.com 2025-08-15T14:02:39+02:00 ci : fix ios-xcode-build (#15324)
7aeee88cfe9c0434eac722b0f7c21404f48758a5 b07791aa1d4831a08ad54ca19aa206c0c5c0f34a Diego Devesa slarengh@gmail.com 2025-08-15T03:27:02-07:00 GitHub noreply@github.com 2025-08-15T12:27:02+02:00 ci : move ccache action to ggml-org fork (#15328)
b07791aa1d4831a08ad54ca19aa206c0c5c0f34a 4227c9be4268ac844921b90f31595f81236bd317 Johannes Gäßler johannesg@5d6.de 2025-08-15T11:23:17+02:00 GitHub noreply@github.com 2025-08-15T11:23:17+02:00 test-opt: fix backend support check (#15317)
4227c9be4268ac844921b90f31595f81236bd317 df36bce667bf14f8e538645547754386f9516326 Johannes Gäßler johannesg@5d6.de 2025-08-14T23:21:24+02:00 GitHub noreply@github.com 2025-08-14T23:21:24+02:00 CUDA: fix negative KV_max values in FA (#15321)
df36bce667bf14f8e538645547754386f9516326 f75b8306472811ecc4e4457d5573a09201f02183 Georgi Gerganov ggerganov@gmail.com 2025-08-14T22:10:51+03:00 GitHub noreply@github.com 2025-08-14T22:10:51+03:00 eval-callback : stop on first NaN (#15320)
f75b8306472811ecc4e4457d5573a09201f02183 7a0de960452f9a57de7f1d167e57b6f3ee5ac1b6 Diego Devesa slarengh@gmail.com 2025-08-14T10:28:29-07:00 GitHub noreply@github.com 2025-08-14T10:28:29-07:00 chat : include kwargs in template example (#15309)
7a0de960452f9a57de7f1d167e57b6f3ee5ac1b6 e4e915912cfd2ee15c5a4a0074813232134892f6 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-14T17:56:26+02:00 GitHub noreply@github.com 2025-08-14T17:56:26+02:00 llama : add 18-layer model type for Gemma 3-270m (#15319)
e4e915912cfd2ee15c5a4a0074813232134892f6 5ba36f61033d2819be27e2abb70e9a3bd20c0fde simevo github@simevo.com 2025-08-14T17:45:27+02:00 GitHub noreply@github.com 2025-08-14T18:45:27+03:00 devops : fix compile bug when the BASE_CUDA_DEV_CONTAINER is based on Ubuntu 24.04 (#15005)
5ba36f61033d2819be27e2abb70e9a3bd20c0fde b204a5a234f4cf0b3f449476da639a5510c6d157 uvos carl@uvos.xyz 2025-08-14T16:23:56+02:00 GitHub noreply@github.com 2025-08-14T16:23:56+02:00 HIP: Cleanup hipification header (#15285)
b204a5a234f4cf0b3f449476da639a5510c6d157 646944cfa8961afd914dd6637739b3cda9a72e11 Aldehir Rojas hello@alde.dev 2025-08-14T09:23:11-05:00 GitHub noreply@github.com 2025-08-14T17:23:11+03:00 gpt-oss: implement harmony parsing (#15181)
646944cfa8961afd914dd6637739b3cda9a72e11 1a01899b612ae8f99a174ad076207090e08d4d7b Christian Kastner ckk@kvr.at 2025-08-14T16:22:58+02:00 GitHub noreply@github.com 2025-08-14T16:22:58+02:00 docker : Enable GGML_CPU_ALL_VARIANTS for ARM (#15267)
1a01899b612ae8f99a174ad076207090e08d4d7b 863d341eeb81db104902b14b6f9413daa515e957 Georgi Gerganov ggerganov@gmail.com 2025-08-14T17:16:03+03:00 GitHub noreply@github.com 2025-08-14T17:16:03+03:00 readme : update hot topics (#15315)
863d341eeb81db104902b14b6f9413daa515e957 d32e03f4495d3efa1c5126f53b449f1d429c5664 Jeff Bolz jbolz@nvidia.com 2025-08-14T08:38:10-05:00 GitHub noreply@github.com 2025-08-14T08:38:10-05:00 vulkan: perf_logger improvements (#15246)
d32e03f4495d3efa1c5126f53b449f1d429c5664 3973163bff40f7f5161b0f08a0011729e2b0406a Georgi Gerganov ggerganov@gmail.com 2025-08-14T14:59:50+03:00 GitHub noreply@github.com 2025-08-14T14:59:50+03:00 server : add SWA checkpoints (#15293)
3973163bff40f7f5161b0f08a0011729e2b0406a 5ade3000bd6040bf6878eafd6c77168552f73c47 Georgi Gerganov ggerganov@gmail.com 2025-08-14T14:19:23+03:00 Georgi Gerganov ggerganov@gmail.com 2025-08-14T14:59:27+03:00 sync : ggml
5ade3000bd6040bf6878eafd6c77168552f73c47 8b2483730f90d6f23e2b188a54a210498bba937f Jason Ni jason.ni.py@gmail.com 2025-08-14T19:17:51+08:00 Georgi Gerganov ggerganov@gmail.com 2025-08-14T14:59:27+03:00 ggml: fix ggml_conv_1d_dw bug (ggml/1323)
8b2483730f90d6f23e2b188a54a210498bba937f 810b9fc8b99dd55517a3e94c6dee29748d482559 Georgi Gerganov ggerganov@gmail.com 2025-08-14T13:41:03+03:00 Georgi Gerganov ggerganov@gmail.com 2025-08-14T14:59:27+03:00 tests : remove unused includes (ggml/0)
810b9fc8b99dd55517a3e94c6dee29748d482559 4ebd0c125b24a0d7a78b0ffc1d9567530ed8f0c4 kallewoof karljohan-alm@garage.co.jp 2025-08-14T20:03:30+09:00 GitHub noreply@github.com 2025-08-14T14:03:30+03:00 perplexity : provide a helpful hint for has_cpl case in split_equal error. (#15304)
4ebd0c125b24a0d7a78b0ffc1d9567530ed8f0c4 5cdb27e0917479d2d742cea7beee089574bb09fa Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-14T12:22:07+02:00 GitHub noreply@github.com 2025-08-14T13:22:07+03:00 cuda : fix GGML_CUDA_GRAPHS=OFF (#15300)
5cdb27e0917479d2d742cea7beee089574bb09fa 3ea913f1ce9567289aedd866a569dbab8fb8e419 Jonathan Graehl 99024+graehl@users.noreply.github.com 2025-08-14T03:03:57-07:00 GitHub noreply@github.com 2025-08-14T12:03:57+02:00 finetune: SGD optimizer, more CLI args (#13873)
3ea913f1ce9567289aedd866a569dbab8fb8e419 29c8fbe4e05fd23c44950d0958299e25fbeabc5c kallewoof karljohan-alm@garage.co.jp 2025-08-14T15:16:32+09:00 GitHub noreply@github.com 2025-08-14T09:16:32+03:00 perplexity: give more information about constraints on failure (#15303)
29c8fbe4e05fd23c44950d0958299e25fbeabc5c 1adc9812bd33dc85489bf093528d61c22917d54f uvos carl@uvos.xyz 2025-08-13T20:44:30+02:00 GitHub noreply@github.com 2025-08-13T20:44:30+02:00 HIP: bump requirement to rocm 6.1 (#15296)
1adc9812bd33dc85489bf093528d61c22917d54f b3e16665e14032d1276f9d0263acef8321b6f518 Bas Nijholt basnijholt@gmail.com 2025-08-13T11:21:31-07:00 GitHub noreply@github.com 2025-08-13T11:21:31-07:00 fix(nix): remove non-functional llama-cpp cachix cache from flake.nix (#15295)
b3e16665e14032d1276f9d0263acef8321b6f518 c24f4e26883a91219af5acfaf1471ca7ef582683 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-13T15:43:00+02:00 GitHub noreply@github.com 2025-08-13T15:43:00+02:00 server : enable -td and -tbd parameters (#15172)
c24f4e26883a91219af5acfaf1471ca7ef582683 d8914fc47e8a69b28c670325cb1c8ce33e3a2960 Judd 4046440+foldl@users.noreply.github.com 2025-08-13T18:45:15+08:00 GitHub noreply@github.com 2025-08-13T13:45:15+03:00 ggml : update `ggml_rope_multi` (#12665)
d8914fc47e8a69b28c670325cb1c8ce33e3a2960 e885445bc1719d2e289a9b2e11714e0f994e68be Copilot 198982749+Copilot@users.noreply.github.com 2025-08-13T12:44:40+02:00 GitHub noreply@github.com 2025-08-13T12:44:40+02:00 common : add --override-tensor-draft, --cpu-moe-draft and --n-cpu-moe-draft parameters (#15191)
e885445bc1719d2e289a9b2e11714e0f994e68be 648ebcdb739abfb5a9be14f4c5c0fd19ff268ef0 Aldehir Rojas hello@alde.dev 2025-08-13T05:28:21-05:00 GitHub noreply@github.com 2025-08-13T12:28:21+02:00 server : filter out harmony thought messages (#15278)
648ebcdb739abfb5a9be14f4c5c0fd19ff268ef0 07aa869a91837d95fcb5612c65a188763ac38647 Ali Tariq alitariq4589@gmail.com 2025-08-13T15:14:44+05:00 GitHub noreply@github.com 2025-08-13T13:14:44+03:00 ci : Added CI with RISC-V RVV1.0 Hardware (#14439)
07aa869a91837d95fcb5612c65a188763ac38647 00f35d509e5367bf19ccaf4b4adddc38db4811c6 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-13T11:30:45+02:00 GitHub noreply@github.com 2025-08-13T11:30:45+02:00 ci : add more python requirements to copilot-setup-steps (#15289)
00f35d509e5367bf19ccaf4b4adddc38db4811c6 6028bf74351d35a06bd98498624f8c2f029f7d1a Georgi Gerganov ggerganov@gmail.com 2025-08-13T11:09:39+03:00 GitHub noreply@github.com 2025-08-13T11:09:39+03:00 ggml : repack block_iq4_nlx8 (#14904)
6028bf74351d35a06bd98498624f8c2f029f7d1a bc5182272c373267352bc689e5fca276934bea2d Oliver Simons osimons@nvidia.com 2025-08-13T10:04:46+02:00 GitHub noreply@github.com 2025-08-13T10:04:46+02:00 CUDA: Optimize `reduce_rows_f32` kernel, leading up to 25x perf improvement on kernel-level and 10% perf increase for Gemma3n (#15132)
bc5182272c373267352bc689e5fca276934bea2d e71d48e3265027351e44a8e198f933c98f242c2e Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-13T09:07:13+02:00 GitHub noreply@github.com 2025-08-13T09:07:13+02:00 ci : add copilot-setup-steps.yml (#15214)
e71d48e3265027351e44a8e198f933c98f242c2e b0493156fa8622694f21f42460a84da3eded0bc0 Tak-RS snosk.t@gmail.com 2025-08-13T14:54:30+09:00 GitHub noreply@github.com 2025-08-13T08:54:30+03:00 ggml-rpc: chunk send()/recv() to avoid EINVAL for very large tensors over RPC (macOS & others) (#15188)
3a617cb18121a85eed17ac5585788810c5d9c1e0 c42293848dd9f03eba9a6c1b85a600b398f06541 Yunzhe Jia yunzhe@calculet.tech 2025-08-13T10:58:59+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-08-13T10:58:59+08:00 add calrt_decode
b0493156fa8622694f21f42460a84da3eded0bc0 f4586ee5986d6f965becb37876d6f3666478a961 uvos carl@uvos.xyz 2025-08-12T22:15:12+02:00 GitHub noreply@github.com 2025-08-12T22:15:12+02:00 HIP: disable sync warp shuffel operators from clr amd_warp_sync_functions.h (#15273)
f4586ee5986d6f965becb37876d6f3666478a961 60a76588106d22ccacd6b1a0d15d8545861d0a0d Romain Biessy romain.biessy@codeplay.com 2025-08-12T13:58:22+02:00 GitHub noreply@github.com 2025-08-12T13:58:22+02:00 sycl: Fix and disable more configurations of mul_mat (#15151)
60a76588106d22ccacd6b1a0d15d8545861d0a0d efe3a90996ca6e67ca90f337fc03ad551082c408 rmatif kingrealriadh@gmail.com 2025-08-12T11:42:41+02:00 GitHub noreply@github.com 2025-08-12T02:42:41-07:00 opencl: allow mixed f16/f32 `add` (#15140)
efe3a90996ca6e67ca90f337fc03ad551082c408 bbd57b7eafb3b32e3f7f3a2175bce0b35abc7de8 Aman Gupta amangupta052@gmail.com 2025-08-12T17:21:45+08:00 GitHub noreply@github.com 2025-08-12T17:21:45+08:00 CUDA cmake: add `-lineinfo` for easier debug (#15260)
bbd57b7eafb3b32e3f7f3a2175bce0b35abc7de8 25ff6f7659f6a5c47d6a73eada5813f0495331f0 Chenguang Li 757486878@qq.com 2025-08-12T16:12:13+08:00 GitHub noreply@github.com 2025-08-12T16:12:13+08:00 CANN: GGML_OP_CPY optimization (#15070)
25ff6f7659f6a5c47d6a73eada5813f0495331f0 be48528b068111304e4a0bb82c028558b5705f05 R0CKSTAR yeahdongcn@gmail.com 2025-08-12T10:02:51+08:00 GitHub noreply@github.com 2025-08-12T10:02:51+08:00 musa: fix failures in test-backend-ops for mul_mat_id op (#15236)
be48528b068111304e4a0bb82c028558b5705f05 cf9e5648a7ae02f7728fefcbcfd2979d83c96e8f hipudding huafengchun@gmail.com 2025-08-11T22:50:31+08:00 GitHub noreply@github.com 2025-08-11T22:50:31+08:00 CANN: Add broadcast for softmax and FA (#15208)
cf9e5648a7ae02f7728fefcbcfd2979d83c96e8f fba5c0d680b555cbac563fef57b33bc5c9621e08 rainred 107027757+gryffindor-rr@users.noreply.github.com 2025-08-11T22:12:12+08:00 GitHub noreply@github.com 2025-08-11T16:12:12+02:00 mtmd : Fix MinicpmV model converter and clip to avoid using hardcode. (#14750)
fba5c0d680b555cbac563fef57b33bc5c9621e08 53d0a1265826f40c5dbc01d06aeab9e14fcbd69b Xuan-Son Nguyen son@huggingface.co 2025-08-11T15:31:35+02:00 GitHub noreply@github.com 2025-08-11T15:31:35+02:00 chat : hotfix gpt-oss jinja raising an exception (#15243)
53d0a1265826f40c5dbc01d06aeab9e14fcbd69b 27093afe78912494073eb043fec93a007e49653c Xuan-Son Nguyen son@huggingface.co 2025-08-11T14:48:41+02:00 GitHub noreply@github.com 2025-08-11T14:48:41+02:00 server : allow specifying reasoning_format in HTTP request (#15238)
27093afe78912494073eb043fec93a007e49653c 228f724d9ce6c56e8cec75bfdabab4dd013def7f Zagaj m.zagajewska@gmail.com 2025-08-11T14:27:54+02:00 GitHub noreply@github.com 2025-08-11T15:27:54+03:00 readme : update infra list (#15234)
228f724d9ce6c56e8cec75bfdabab4dd013def7f cd3069dfcbeee8e0e96cffb93b0ebb9e595e273a Georgi Gerganov ggerganov@gmail.com 2025-08-11T13:58:24+03:00 GitHub noreply@github.com 2025-08-11T13:58:24+03:00 kv-cache : fix seq_rm with seq_id == -1 (#15226)
cd3069dfcbeee8e0e96cffb93b0ebb9e595e273a 50e81bdf5db563ab57a9a722b08f96fa8a76c927 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-11T11:21:19+02:00 GitHub noreply@github.com 2025-08-11T11:21:19+02:00 kv-cache : log (debug) all streams in find_slot (#15176)
50e81bdf5db563ab57a9a722b08f96fa8a76c927 1ebbaddff2a44b0599df659175d3274bd5bbeb81 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-11T11:15:44+02:00 GitHub noreply@github.com 2025-08-11T11:15:44+02:00 convert : fix merge conflicts (#15229)
1ebbaddff2a44b0599df659175d3274bd5bbeb81 a3a7874272e5a060079658eb5cca4617b7f99062 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-11T10:21:24+02:00 GitHub noreply@github.com 2025-08-11T11:21:24+03:00 perplexity : update comments/error msg to use decode [no ci] (#15227)
a3a7874272e5a060079658eb5cca4617b7f99062 002cb1bb3345967fbe0fa7766c2d94c2da31ef45 Julien Denize 40604584+juliendenize@users.noreply.github.com 2025-08-11T10:07:49+02:00 GitHub noreply@github.com 2025-08-11T10:07:49+02:00 convert : improve Mistral models integration (#14737)
002cb1bb3345967fbe0fa7766c2d94c2da31ef45 79c1160b073b8148a404f3dd2584be1606dccc66 Charles Xu charles.xu@arm.com 2025-08-11T09:59:26+02:00 GitHub noreply@github.com 2025-08-11T09:59:26+02:00 kleidiai: fix unsigned overflow bug (#15150)
79c1160b073b8148a404f3dd2584be1606dccc66 34c9d765bf173c551398f1e7fa4595019bc53bab David Zhao 90013954+Your-Cheese@users.noreply.github.com 2025-08-09T13:29:43-05:00 GitHub noreply@github.com 2025-08-09T20:29:43+02:00 cuda: refactored ssm_scan and use CUB (#13291)
34c9d765bf173c551398f1e7fa4595019bc53bab e54d41befcc1575f4c898c5ff4ef43970cead75f Aman Gupta amangupta052@gmail.com 2025-08-09T20:00:24+08:00 GitHub noreply@github.com 2025-08-09T20:00:24+08:00 CUDA: add attention sinks for tile and wmma (#15178)
e54d41befcc1575f4c898c5ff4ef43970cead75f 4850b52aedceeb70bb4fe49f2d7cd1df6ee98682 compilade git@compilade.net 2025-08-08T17:48:26-04:00 GitHub noreply@github.com 2025-08-08T17:48:26-04:00 gguf-py : add Numpy MXFP4 de/quantization support (#15111)
4850b52aedceeb70bb4fe49f2d7cd1df6ee98682 cd6983d56d2cce94ecb86bb114ae8379a609073c Johannes Gäßler johannesg@5d6.de 2025-08-08T23:04:36+02:00 GitHub noreply@github.com 2025-08-08T23:04:36+02:00 server-bench: external OAI servers, sqlite (#15179)
cd6983d56d2cce94ecb86bb114ae8379a609073c 6c7e9a54406dbba5e53754a8f70a285414717b06 AN Long aisk@users.noreply.github.com 2025-08-08T21:37:22+09:00 GitHub noreply@github.com 2025-08-08T14:37:22+02:00 ggml : fix field name when new ggml_backend (#14944)
6c7e9a54406dbba5e53754a8f70a285414717b06 1425f587a82bc303469b5c32759a2746ba4e1e20 Olivier Chafik olivier.chafik@gmail.com 2025-08-08T10:45:18+01:00 GitHub noreply@github.com 2025-08-08T10:45:18+01:00 vendor: sync minja (#15161)
1425f587a82bc303469b5c32759a2746ba4e1e20 aaa3d07ae749b781d6135eaff23c7fa8a4ab404a Johannes Gäßler johannesg@5d6.de 2025-08-08T08:19:58+02:00 GitHub noreply@github.com 2025-08-08T08:19:58+02:00 CUDA: attention sinks for mma FlashAttention (#15157)
aaa3d07ae749b781d6135eaff23c7fa8a4ab404a 50aa9389014bba2dd12234132aa6b8ca3601a17f lhez lih@qti.qualcomm.com 2025-08-08T13:47:03+09:00 GitHub noreply@github.com 2025-08-07T21:47:03-07:00 opencl: support sink in `soft_max` (attn sinks) (#15152)
50aa9389014bba2dd12234132aa6b8ca3601a17f c4f53563df4575196ea13f5ed669ea8ea659a6be Xuan-Son Nguyen son@huggingface.co 2025-08-07T23:26:03+02:00 GitHub noreply@github.com 2025-08-07T23:26:03+02:00 convert : support non-mxfp4 HF model (#15153)
c4f53563df4575196ea13f5ed669ea8ea659a6be a0552c8beef74e843bb085c8ef0c63f9ed7a2b27 Jeff Bolz jbolz@nvidia.com 2025-08-07T15:44:20-05:00 GitHub noreply@github.com 2025-08-07T22:44:20+02:00 vulkan: support fattn sinks (#15126)
a0552c8beef74e843bb085c8ef0c63f9ed7a2b27 99acbc9921b119aa7ed929eb5780a66a8f06e6d9 Jeff Bolz jbolz@nvidia.com 2025-08-07T15:07:11-05:00 GitHub noreply@github.com 2025-08-07T22:07:11+02:00 vulkan: Add env var to disable host visible vidmem (#15109)
99acbc9921b119aa7ed929eb5780a66a8f06e6d9 7ad67ba9fe2b909e271dd31b99c5fce3aba35899 RunningLeon mnsheng@yeah.net 2025-08-08T00:20:40+08:00 GitHub noreply@github.com 2025-08-07T18:20:40+02:00 llama : Support intern-s1 (#14875)
7ad67ba9fe2b909e271dd31b99c5fce3aba35899 9a96389544a08fd829fccda28142ce2066017fde uvos carl@uvos.xyz 2025-08-07T16:44:14+02:00 GitHub noreply@github.com 2025-08-07T16:44:14+02:00 HIP: add cmake option to enable compiler output of kernel resource usage metrics (#15103)
9a96389544a08fd829fccda28142ce2066017fde 1d72c841888b9450916bdd5a9b3274da380f5b36 Christian Kastner ckk@kvr.at 2025-08-07T13:45:41+02:00 GitHub noreply@github.com 2025-08-07T13:45:41+02:00 ggml: Skip backend library linking code when GGML_BACKEND_DL=ON (#15094)
1d72c841888b9450916bdd5a9b3274da380f5b36 20638e4f16fcc21f836c7556e83bbf532bb5a0f0 Johannes Gäßler johannesg@5d6.de 2025-08-07T10:53:21+02:00 GitHub noreply@github.com 2025-08-07T10:53:21+02:00 CUDA: GEMM for FP32/FP16/BF16 and ne11 <= 16 (#15131)
20638e4f16fcc21f836c7556e83bbf532bb5a0f0 36d3f00e142696f708ab297b9f8f1c825594712d Johannes Gäßler johannesg@5d6.de 2025-08-07T08:50:30+02:00 GitHub noreply@github.com 2025-08-07T08:50:30+02:00 scripts: fix crash when --tool is not set (#15133)
36d3f00e142696f708ab297b9f8f1c825594712d 5fd160bbd9d70b94b5b11b0001fd7f477005e4a0 Daniel Bevenius daniel.bevenius@gmail.com 2025-08-07T05:31:48+02:00 GitHub noreply@github.com 2025-08-07T05:31:48+02:00 requirements : fix PyTorch uint64 compatibility (#15134)
5fd160bbd9d70b94b5b11b0001fd7f477005e4a0 756cfea82608911bbfcbf45164b8fdaddbafaa31 Reese Levine reeselevine1@gmail.com 2025-08-06T15:14:40-07:00 GitHub noreply@github.com 2025-08-06T15:14:40-07:00 ggml: Add basic SET_ROWS support in WebGPU (#15137)
756cfea82608911bbfcbf45164b8fdaddbafaa31 e725a1a982ca870404a9c4935df52466327bbd02 rmatif kingrealriadh@gmail.com 2025-08-06T23:17:51+02:00 GitHub noreply@github.com 2025-08-06T14:17:51-07:00 fix profiling crash (#15072)
e725a1a982ca870404a9c4935df52466327bbd02 3db4da56a5a00c4b47bbb3b18f85fb4473662adb lhez lih@qti.qualcomm.com 2025-08-07T04:12:17+09:00 GitHub noreply@github.com 2025-08-06T12:12:17-07:00 opencl: add `swiglu_oai` and `add_id` (#15121)
3db4da56a5a00c4b47bbb3b18f85fb4473662adb 476aa3fd5779b32d06cd84338f777da82341195c Sachin Desai smdesai@gmail.com 2025-08-06T11:27:30-07:00 GitHub noreply@github.com 2025-08-06T20:27:30+02:00 chat : support Granite model reasoning and tool call (#14864)
476aa3fd5779b32d06cd84338f777da82341195c 0d8831543cdc368fb248bae6f1b4aa5516684edc Juk Armstrong 69222624+jukofyork@users.noreply.github.com 2025-08-06T17:28:48+01:00 GitHub noreply@github.com 2025-08-06T17:28:48+01:00 Fixed name `-override-tensors` to `-override-tensor` (#15129)
0d8831543cdc368fb248bae6f1b4aa5516684edc 65c797c4fad4d9966695ac4b4a1560be44109267 Diego Devesa slarengh@gmail.com 2025-08-06T05:37:35-07:00 GitHub noreply@github.com 2025-08-06T14:37:35+02:00 ggml : fix fallback to CPU for ununsupported ops (#15118)
65c797c4fad4d9966695ac4b4a1560be44109267 25726898e855ec6dffba227f2233a63c57184036 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-06T13:26:49+02:00 GitHub noreply@github.com 2025-08-06T13:26:49+02:00 chat : fix yandex chat template (#15116)
25726898e855ec6dffba227f2233a63c57184036 2241453252147bb7362a286977ee9f9a92130062 stevenkuang stevenkuang@tencent.com 2025-08-06T17:48:30+08:00 GitHub noreply@github.com 2025-08-06T11:48:30+02:00 chat : fix hunyuan auto-detection (#15114)
c42293848dd9f03eba9a6c1b85a600b398f06541 1cbf442f50a154aac0bf35d7b8b3436a2cdfd6f5 Yunzhe Jia yunzhe@calculet.tech 2025-08-06T15:46:06+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-08-06T15:46:06+08:00 sync with calrt API
1cbf442f50a154aac0bf35d7b8b3436a2cdfd6f5 ba67d6c03fed3193987a4ef13050e6e2a4451df4 Yunzhe Jia yunzhe@calculet.tech 2025-07-24T15:50:19+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-08-06T15:45:10+08:00 1.sync with calrt prefill,decode,kv_update op 2.add hw pattern untile function
2241453252147bb7362a286977ee9f9a92130062 9515c6131aecaccc955fdedcfe16c3e030aaefcb Chenguang Li 757486878@qq.com 2025-08-06T14:12:42+08:00 GitHub noreply@github.com 2025-08-06T14:12:42+08:00 CANN: add support for ACL Graph (#15065)
9515c6131aecaccc955fdedcfe16c3e030aaefcb fd1234cb468935ea087d6929b2487926c3afff4b Reese Levine reeselevine1@gmail.com 2025-08-05T16:26:38-07:00 GitHub noreply@github.com 2025-08-05T16:26:38-07:00 ggml: WebGPU disable SET_ROWS for now (#15078)
fd1234cb468935ea087d6929b2487926c3afff4b f324a3b715d5c1081c110ce459f8a8486fb1ee89 Georgi Gerganov ggerganov@gmail.com 2025-08-05T22:10:36+03:00 GitHub noreply@github.com 2025-08-05T22:10:36+03:00 llama : add gpt-oss (#15091)
f324a3b715d5c1081c110ce459f8a8486fb1ee89 be426425817bc3e6a2d91dae476dba6fa85894be Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-05T20:43:36+02:00 GitHub noreply@github.com 2025-08-05T20:43:36+02:00 chat : only remove double bos/eos if added (#15086)
be426425817bc3e6a2d91dae476dba6fa85894be 3306ceabf02e3df66666e5851800e843c7ca207e Georgi Gerganov ggerganov@gmail.com 2025-08-05T20:19:33+03:00 GitHub noreply@github.com 2025-08-05T20:19:33+03:00 readme : update hot topics (#15097)
3306ceabf02e3df66666e5851800e843c7ca207e c81de6e107ef51ef76aadcb8a6f008711c462517 Romain Biessy romain.biessy@codeplay.com 2025-08-05T18:39:55+02:00 GitHub noreply@github.com 2025-08-05T18:39:55+02:00 sycl: fix mul_mat selection (#15092)
c81de6e107ef51ef76aadcb8a6f008711c462517 22f060c9c4b5ef49a83a20eda25fcb792419580b Juk Armstrong 69222624+jukofyork@users.noreply.github.com 2025-08-05T13:56:44+01:00 GitHub noreply@github.com 2025-08-05T13:56:44+01:00 Fix `glm4moe` bug (#15088)
22f060c9c4b5ef49a83a20eda25fcb792419580b ee3a9fcf88fe5b5e1213711e05861b83cd4fdfe6 Alex Wu dindinw@users.noreply.github.com 2025-08-05T19:56:44+08:00 GitHub noreply@github.com 2025-08-05T13:56:44+02:00 webui: fix markdown table (#15081)
ee3a9fcf88fe5b5e1213711e05861b83cd4fdfe6 ec428b02c347767f24c78111309e3f30d2ada289 compilade git@compilade.net 2025-08-05T05:27:45-04:00 GitHub noreply@github.com 2025-08-05T11:27:45+02:00 context : fix index overflow on huge outputs (#15080)
ec428b02c347767f24c78111309e3f30d2ada289 19f68fa5a4c3bf796de52a6db9008e77d29f423a Diego Devesa slarengh@gmail.com 2025-08-04T16:05:36-07:00 GitHub noreply@github.com 2025-08-05T01:05:36+02:00 llama : add --n-cpu-moe option (#15077)
19f68fa5a4c3bf796de52a6db9008e77d29f423a 41613437ffee0dbccad684fc744788bc504ec213 compilade git@compilade.net 2025-08-04T17:26:52-04:00 GitHub noreply@github.com 2025-08-04T23:26:52+02:00 imatrix : warn when GGUF imatrix is saved without .gguf suffix (#15076)
41613437ffee0dbccad684fc744788bc504ec213 e5bebe5251cee2678e8531aa1598ca21b3c6ce1d Christian Kastner ckk@kvr.at 2025-08-04T21:29:14+02:00 GitHub noreply@github.com 2025-08-04T21:29:14+02:00 cmake: Add GGML_BACKEND_DIR option (#15074)
e5bebe5251cee2678e8531aa1598ca21b3c6ce1d ef0144c087b33e5b8da42d529ac71aaf05cb49df Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-04T21:01:48+02:00 GitHub noreply@github.com 2025-08-04T21:01:48+02:00 gguf-py : add --chat-template-file to gguf_new_metadata (#15075)
ef0144c087b33e5b8da42d529ac71aaf05cb49df 2721257e3e2c4c944ac8a08221113ee7cb503f1b Sam sammcj@users.noreply.github.com 2025-08-05T04:29:25+10:00 GitHub noreply@github.com 2025-08-04T20:29:25+02:00 model: support GLM 4.5 family of models (#14939)
2721257e3e2c4c944ac8a08221113ee7cb503f1b 587d0118f50b7e8f4bafbcdd218aefd9da0272e1 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-04T18:11:02+02:00 GitHub noreply@github.com 2025-08-04T18:11:02+02:00 quantize : fix confusing error message if ftype is invalid (#15071)
587d0118f50b7e8f4bafbcdd218aefd9da0272e1 5aa1105da24a8dd1661cea3db0582c9b2c2f54d3 Reese Levine reeselevine1@gmail.com 2025-08-04T08:52:43-07:00 GitHub noreply@github.com 2025-08-04T08:52:43-07:00 ggml: WebGPU backend host improvements and style fixing (#14978)
5aa1105da24a8dd1661cea3db0582c9b2c2f54d3 d31192b4ee1441bbbecd3cbf9e02633368bdc4f5 Jeff Bolz jbolz@nvidia.com 2025-08-04T00:09:19-05:00 GitHub noreply@github.com 2025-08-04T07:09:19+02:00 vulkan: fix build when using glslang that does not support coopmat2 (#15062)
d31192b4ee1441bbbecd3cbf9e02633368bdc4f5 0a2f5496bef9e54e5f42d6c2c3ad9eb7b379aed0 compilade git@compilade.net 2025-08-03T16:00:05-04:00 GitHub noreply@github.com 2025-08-03T22:00:05+02:00 imatrix : use GGUF by default (#14842)
0a2f5496bef9e54e5f42d6c2c3ad9eb7b379aed0 11a3811164ef2d75393c6b0a632f4c608e3e3dd2 compilade git@compilade.net 2025-08-03T15:49:13-04:00 GitHub noreply@github.com 2025-08-03T21:49:13+02:00 imatrix : fix 3d activation handling for hybrid and recurrent models (#14994)
11a3811164ef2d75393c6b0a632f4c608e3e3dd2 97366dc6abdd0bdc74260bd3c42bd06f0feb7428 compilade git@compilade.net 2025-08-03T15:43:07-04:00 GitHub noreply@github.com 2025-08-03T21:43:07+02:00 memory : handle kv_unified for hybrid models (#15050)
97366dc6abdd0bdc74260bd3c42bd06f0feb7428 83bc2f288c0e08e676d9beca9c4669197e920593 Csaba Kecskemeti csaba.kecskemeti@gmail.com 2025-08-03T12:38:18-07:00 GitHub noreply@github.com 2025-08-03T21:38:18+02:00 vocab : JetBrains Mellum pre-tokenizer (#15045)
83bc2f288c0e08e676d9beca9c4669197e920593 6c7a441161080551ce8a52ba32563b6295067192 Gabriel Larson 55459720+gabriellarson@users.noreply.github.com 2025-08-03T09:56:25-05:00 GitHub noreply@github.com 2025-08-03T16:56:25+02:00 model : add text-only support for Kimi-VL (and find special tokens in text_config) (#15051)
6c7a441161080551ce8a52ba32563b6295067192 5c0eb5ef544aeefd81c303e03208f768e158d93c Jeff Bolz jbolz@nvidia.com 2025-08-03T07:23:57-05:00 GitHub noreply@github.com 2025-08-03T14:23:57+02:00 vulkan: Use coopmat2 for conv2d (#14982)
5c0eb5ef544aeefd81c303e03208f768e158d93c 03d46982180c2fb624bd2a233e46426ab22be5d1 lhez lih@qti.qualcomm.com 2025-08-02T10:51:18-07:00 GitHub noreply@github.com 2025-08-02T19:51:18+02:00 opencl: fix adreno compiler detection logic (#15029)
03d46982180c2fb624bd2a233e46426ab22be5d1 3303c19b1691088275ee864a823697177c94a15d Johannes Gäßler johannesg@5d6.de 2025-08-02T16:37:08+02:00 GitHub noreply@github.com 2025-08-02T16:37:08+02:00 CUDA: use mma FA kernel for gqa > 4 on RTX 4000 (#15035)
3303c19b1691088275ee864a823697177c94a15d 4fdea540bda4648f98b85e8ee9dc66db4bfb5945 leejet leejet714@gmail.com 2025-08-02T22:15:36+08:00 GitHub noreply@github.com 2025-08-02T17:15:36+03:00 cuda: make im2col a little faster (#15025)
4fdea540bda4648f98b85e8ee9dc66db4bfb5945 a4569c41fd2253c89ef52fc2378687bdbf42f61a Daniel Bevenius daniel.bevenius@gmail.com 2025-08-02T16:14:57+02:00 GitHub noreply@github.com 2025-08-02T17:14:57+03:00 kv-cache : skip alignment of n_stream in kv-cache log msg [no ci] (#15040)
a4569c41fd2253c89ef52fc2378687bdbf42f61a 15e92fd33791e60a4ddb5970b47242a855c27117 Georgi Gerganov ggerganov@gmail.com 2025-08-02T17:14:21+03:00 GitHub noreply@github.com 2025-08-02T17:14:21+03:00 llama : enable LLAMA_SET_ROWS=1 by default (#14959)
15e92fd33791e60a4ddb5970b47242a855c27117 2bf3fbf0b54f97aef2b388b76d222789e1c170f1 Georgi Gerganov ggerganov@gmail.com 2025-08-02T17:13:05+03:00 GitHub noreply@github.com 2025-08-02T17:13:05+03:00 cuda, sycl : fix batched gemm when ne02 == 1 && ne03 > 1 (#15038)
2bf3fbf0b54f97aef2b388b76d222789e1c170f1 711d5e6fe66eb6cd7a10d71cec4567321848be08 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-08-02T14:39:01+02:00 GitHub noreply@github.com 2025-08-02T14:39:01+02:00 ci : check that pre-tokenizer hashes are up-to-date (#15032)
711d5e6fe66eb6cd7a10d71cec4567321848be08 f738989dcb9ccbe468c945553eafbeef7b869675 Douglas Hanley thesecretaryofwar@gmail.com 2025-08-02T05:51:02-05:00 GitHub noreply@github.com 2025-08-02T12:51:02+02:00 convert : fix Qwen3-Embedding pre-tokenizer hash (#15030)
f738989dcb9ccbe468c945553eafbeef7b869675 4cb208c93c1c938591a5b40354e2a6f9b94489bc Jhen-Jie Hong iainst0409@gmail.com 2025-08-02T18:04:48+08:00 GitHub noreply@github.com 2025-08-02T18:04:48+08:00 chat : fix multiple tool_calls on hermes-2-pro (#14962)
4cb208c93c1c938591a5b40354e2a6f9b94489bc 3025b621d12a6931ff5e9775d4f644719980ad91 Jeff Bolz jbolz@nvidia.com 2025-08-02T04:21:37-05:00 GitHub noreply@github.com 2025-08-02T11:21:37+02:00 vulkan: coopmat2 mul_mat optimizations (#14934)
3025b621d12a6931ff5e9775d4f644719980ad91 ec0b18802c91badd3ff1388ffd09ee163251bd72 R0CKSTAR yeahdongcn@gmail.com 2025-08-02T17:20:40+08:00 GitHub noreply@github.com 2025-08-02T17:20:40+08:00 llama-bench: rename DB table name from test to llama_bench (#15003)
ec0b18802c91badd3ff1388ffd09ee163251bd72 339bd0268c498c89529cd0e90c44883c211e3745 Jeff Bolz jbolz@nvidia.com 2025-08-02T03:48:30-05:00 GitHub noreply@github.com 2025-08-02T10:48:30+02:00 vulkan: Support ne[3]>1 in noncontig matrix-vector multiply (#15015)
339bd0268c498c89529cd0e90c44883c211e3745 f906275537d14c8fc7c6976d944233771fd6672c Douglas Hanley thesecretaryofwar@gmail.com 2025-08-02T03:44:50-05:00 GitHub noreply@github.com 2025-08-02T10:44:50+02:00 model : support Qwen3-Embedding (#15023)
f906275537d14c8fc7c6976d944233771fd6672c a9f7541ec25c4c8547daf5ff48700ad2836e2b7d Johannes Gäßler johannesg@5d6.de 2025-08-02T10:12:41+02:00 GitHub noreply@github.com 2025-08-02T10:12:41+02:00 server: enable token array inputs for OAI API (#15001)
a9f7541ec25c4c8547daf5ff48700ad2836e2b7d 9c35706b98ea271858acef4194f526a71b24cdc9 Jeff Bolz jbolz@nvidia.com 2025-08-02T02:57:04-05:00 GitHub noreply@github.com 2025-08-02T09:57:04+02:00 vulkan: optimizations for direct convolution (#14933)
9c35706b98ea271858acef4194f526a71b24cdc9 c76b420e4ce06f7b7cdfbb0b85d02c90e5cc5a3a Johannes Gäßler johannesg@5d6.de 2025-08-01T20:47:32+02:00 GitHub noreply@github.com 2025-08-01T20:47:32+02:00 CUDA: fix MMQ nwarps for AMD with warp_size==32 (#15014)
c76b420e4ce06f7b7cdfbb0b85d02c90e5cc5a3a 0f5ccd6fd1a1f709010312933db0316867cc30b6 l-austenfeld 53152202+l-austenfeld@users.noreply.github.com 2025-08-01T16:59:06+02:00 GitHub noreply@github.com 2025-08-01T16:59:06+02:00 vendor : update vendored copy of google/minja (#15011)
0f5ccd6fd1a1f709010312933db0316867cc30b6 1c872f71fb8a25589efa3ee9b6bf8b517cb8caa4 stevenkuang stevenkuang@tencent.com 2025-08-01T21:31:12+08:00 GitHub noreply@github.com 2025-08-01T15:31:12+02:00 model : add hunyuan dense (#14878)
1c872f71fb8a25589efa3ee9b6bf8b517cb8caa4 baad94885df512bb24ab01e2b22d1998fce4d00e lhez quic_lih@quicinc.com 2025-08-01T04:15:44-07:00 GitHub noreply@github.com 2025-08-01T13:15:44+02:00 opencl: add f16 for `add`, `sub`, `mul`, `div` (#14984)
baad94885df512bb24ab01e2b22d1998fce4d00e ba42794c9ead96ad52311ba1b23eefcbf3d6f63d Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2025-08-01T11:50:33+05:30 GitHub noreply@github.com 2025-08-01T09:20:33+03:00 ggml : Q2k interleaving implementation - x86/x64 SIMD (#14373)
ba42794c9ead96ad52311ba1b23eefcbf3d6f63d 2860d479b456e1caa026b40b829d5b13c42a8ed7 Georgi Gerganov ggerganov@gmail.com 2025-08-01T06:38:12+03:00 GitHub noreply@github.com 2025-08-01T06:38:12+03:00 graph : fix equal_seq() check (#14986)
2860d479b456e1caa026b40b829d5b13c42a8ed7 484b2091ce5017901483b5204c07878f171d1441 diannao 55k@outlook.com 2025-08-01T10:02:34+08:00 GitHub noreply@github.com 2025-08-01T10:02:34+08:00 docker : add cann build pipline (#14591)
484b2091ce5017901483b5204c07878f171d1441 daf2dd788066b8b239cb7f68210e090c2124c199 R0CKSTAR yeahdongcn@gmail.com 2025-08-01T08:47:27+08:00 GitHub noreply@github.com 2025-08-01T08:47:27+08:00 compare-commits.sh: support both llama-bench and test-backend-ops (#14392)
daf2dd788066b8b239cb7f68210e090c2124c199 a06ed5feaec9f935fbf662035b2673167bc88460 Ed Addario 29247825+EAddario@users.noreply.github.com 2025-07-31T20:32:18+01:00 GitHub noreply@github.com 2025-07-31T21:32:18+02:00 quantize : skip tensor override when in fallback mode (#14995)
a06ed5feaec9f935fbf662035b2673167bc88460 784524053d986255e383dae45235e4f76c6792a7 Diego Devesa slarengh@gmail.com 2025-07-31T11:15:41-07:00 GitHub noreply@github.com 2025-07-31T20:15:41+02:00 llama : add simple option to enable CPU for MoE weights (--cpu-moe) (#14992)
784524053d986255e383dae45235e4f76c6792a7 d6818d06a6237631523bc0f45d42e79482667948 Aman Gupta amangupta052@gmail.com 2025-08-01T01:22:58+08:00 GitHub noreply@github.com 2025-08-01T01:22:58+08:00 Fix params bug in diffusion example (#14993)
d6818d06a6237631523bc0f45d42e79482667948 e08a98826bcce70a3377592c51d0efed99eafe07 Diego Devesa slarengh@gmail.com 2025-07-31T09:11:34-07:00 GitHub noreply@github.com 2025-07-31T18:11:34+02:00 llama : allow other bufts when overriding to CPU, add --no-repack option (#14990)
e08a98826bcce70a3377592c51d0efed99eafe07 952a47f455fbd92e2659b98b9b6317a2dafeb532 Ruben Ortlam picard12@live.de 2025-07-31T17:46:54+02:00 GitHub noreply@github.com 2025-07-31T17:46:54+02:00 Vulkan: Fix minor debug mode issues (#14899)
952a47f455fbd92e2659b98b9b6317a2dafeb532 36e5fe7bcd6640f4ff974d4c5cb04a13b5b29bce tc-mb 157115220+tc-mb@users.noreply.github.com 2025-07-31T23:22:17+08:00 GitHub noreply@github.com 2025-07-31T17:22:17+02:00 mtmd : support MiniCPM-V 4.0 (#14983)
36e5fe7bcd6640f4ff974d4c5cb04a13b5b29bce 94933c8c2eeaa9a7983e3f6c08af76bd86724094 Csaba Kecskemeti csaba.kecskemeti@gmail.com 2025-07-31T07:59:49-07:00 GitHub noreply@github.com 2025-07-31T10:59:49-04:00 MODEL_TENSOR.SSM_DT_NORM has defined twice (#14991)
94933c8c2eeaa9a7983e3f6c08af76bd86724094 c1dacaa99b4ead6edbac928cd2da59436573f6b0 g2mt 166577174+g2mt@users.noreply.github.com 2025-07-31T05:25:23-07:00 GitHub noreply@github.com 2025-07-31T14:25:23+02:00 server : implement universal assisted decoding (#12635)
c1dacaa99b4ead6edbac928cd2da59436573f6b0 a9f77a8be348deeb11fb3d54d412bf583003c90d Dongliang Wei 121270393+wdl339@users.noreply.github.com 2025-07-31T20:12:20+08:00 GitHub noreply@github.com 2025-07-31T14:12:20+02:00 llama : merge build_moe_ffn_from_probs function into build_moe_ffn (#14968)
a9f77a8be348deeb11fb3d54d412bf583003c90d 8a4a85627702b569d7d2810f2de06a4321656e9d Lukas Straub lukasstraub2@web.de 2025-07-31T14:08:23+02:00 GitHub noreply@github.com 2025-07-31T14:08:23+02:00 server : add openai-style logit_bias support (#14946)
8a4a85627702b569d7d2810f2de06a4321656e9d 11490b36723d511d75fb601995c79b5c363ba3a2 Aman Gupta amangupta052@gmail.com 2025-07-31T19:49:09+08:00 GitHub noreply@github.com 2025-07-31T19:49:09+08:00 Add LLaDA 8b Diffusion model (#14771)
11490b36723d511d75fb601995c79b5c363ba3a2 66625a59a54d0a7504eda4c4e83abfcd83ba1cf8 hipudding huafengchun@gmail.com 2025-07-31T19:47:20+08:00 GitHub noreply@github.com 2025-07-31T19:47:20+08:00 CANN: Improve loading efficiency after converting weights to NZ format. (#14985)
66625a59a54d0a7504eda4c4e83abfcd83ba1cf8 6e6725459a892b49602b596339de4916c7c7965a compilade git@compilade.net 2025-07-31T01:02:46-04:00 GitHub noreply@github.com 2025-07-31T08:02:46+03:00 graph : reduce splits for recurrent and hybrid models (#14825)
6e6725459a892b49602b596339de4916c7c7965a e9192bec564780bd4313ad6524d20a0ab92797db lhez lih@qti.qualcomm.com 2025-07-30T14:56:55-07:00 GitHub noreply@github.com 2025-07-30T14:56:55-07:00 opencl: add `mul_mat_f32_f32_l4_lm` and `mul_mat_f16_f32_l4_lm` (#14809)
e9192bec564780bd4313ad6524d20a0ab92797db 41e78c567e9a8c652e405f4f909deb598deecd31 Ed Addario 29247825+EAddario@users.noreply.github.com 2025-07-30T20:11:56+01:00 GitHub noreply@github.com 2025-07-30T21:11:56+02:00 quantize : fix using combined imatrix GGUFs (multiple datasets) (#14973)
41e78c567e9a8c652e405f4f909deb598deecd31 ad4a700117d1746799d2d6599e526e2c3a7938d2 Daniel Bevenius daniel.bevenius@gmail.com 2025-07-30T18:07:11+02:00 GitHub noreply@github.com 2025-07-30T18:07:11+02:00 server : add support for `embd_normalize` parameter (#14964)
ad4a700117d1746799d2d6599e526e2c3a7938d2 e32a4ec60ee472acf3fe82b5966976fd6965ac6b uvos carl@uvos.xyz 2025-07-30T17:38:06+02:00 GitHub noreply@github.com 2025-07-30T17:38:06+02:00 HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes (#14949)
e32a4ec60ee472acf3fe82b5966976fd6965ac6b e228de94496636681b36690247d96f97d5f76c0d Georgi Gerganov ggerganov@gmail.com 2025-07-30T16:03:13+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-30T17:33:11+03:00 sync : ggml
e228de94496636681b36690247d96f97d5f76c0d 73a8e5ca0372f4dcaf1aed4e42261723da0914aa Kai Pastor dg0yt@darc.de 2025-07-30T14:53:16+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-30T17:33:11+03:00 cmake : Fix BLAS link interface (ggml/1316)
73a8e5ca0372f4dcaf1aed4e42261723da0914aa 92b8810ec7aa6d778bc287cc918443cf67b962e2 Kai Pastor dg0yt@darc.de 2025-07-30T14:52:26+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-30T17:33:11+03:00 vulkan : fix 32-bit builds (ggml/1313)
92b8810ec7aa6d778bc287cc918443cf67b962e2 00131d6eaf4df029e1ec84de868c2c5957503007 Johannes Gäßler johannesg@5d6.de 2025-07-30T15:46:13+02:00 GitHub noreply@github.com 2025-07-30T15:46:13+02:00 CUDA: skip masked KV slices for all FA kernels (#14924)
00131d6eaf4df029e1ec84de868c2c5957503007 1e15bfd42c3938506e0da5939bf7f42780965f01 Georgi Gerganov ggerganov@gmail.com 2025-07-30T15:12:02+03:00 GitHub noreply@github.com 2025-07-30T15:12:02+03:00 tests : update for LLAMA_SET_ROWS=1 (#14961)
1e15bfd42c3938506e0da5939bf7f42780965f01 a118d80233d3bf92569c051346fd2638f87bf202 Georgi Gerganov ggerganov@gmail.com 2025-07-30T13:52:11+03:00 GitHub noreply@github.com 2025-07-30T13:52:11+03:00 graph : fix stack-use-after-return (#14960)
a118d80233d3bf92569c051346fd2638f87bf202 61550f8231dc0aa478e2f537c5009ede6878ce22 Douglas Hanley thesecretaryofwar@gmail.com 2025-07-30T00:25:05-05:00 GitHub noreply@github.com 2025-07-30T08:25:05+03:00 embeddings: fix extraction of CLS pooling results (#14927)
61550f8231dc0aa478e2f537c5009ede6878ce22 aa79524c51fb014f8df17069d31d7c44b9ea6cb8 Xinpeng Dou 15529241576@163.com 2025-07-30T08:39:24+08:00 GitHub noreply@github.com 2025-07-30T08:39:24+08:00 CANN: update ops docs (#14935)
aa79524c51fb014f8df17069d31d7c44b9ea6cb8 b77d11179d7efe68ce5c913006f2ef9ee17cc8f7 uvos carl@uvos.xyz 2025-07-29T20:23:04+02:00 GitHub noreply@github.com 2025-07-29T20:23:04+02:00 HIP: remove the use of __HIP_PLATFORM_AMD__, explicitly support only AMD targets (#14945)
b77d11179d7efe68ce5c913006f2ef9ee17cc8f7 c7aa1364fd59b2ac06fd9e0a719253d968472dc3 uvos carl@uvos.xyz 2025-07-29T17:44:30+02:00 GitHub noreply@github.com 2025-07-29T17:44:30+02:00 HIP: add GGML_HIP_MMQ_MFMA option to allow disableing the MFMA path. (#14930)
c7aa1364fd59b2ac06fd9e0a719253d968472dc3 1a67fcc30677e96dda76bb1b290788e7d8852b51 uvos carl@uvos.xyz 2025-07-29T17:43:43+02:00 GitHub noreply@github.com 2025-07-29T17:43:43+02:00 HIP: Ignore unsupported unroll transformation in fattn-vec (#14931)
1a67fcc30677e96dda76bb1b290788e7d8852b51 204f2cf168dc01ca7b200b1510e0ff585ca9a92e kallewoof karljohan-alm@garage.co.jp 2025-07-30T00:05:38+09:00 GitHub noreply@github.com 2025-07-29T17:05:38+02:00 common : avoid logging partial messages (which can contain broken UTF-8 sequences) (#14937)
204f2cf168dc01ca7b200b1510e0ff585ca9a92e 138b288b594eeb19d58e0ab0d74eae32009c9c20 hipudding huafengchun@gmail.com 2025-07-29T22:36:43+08:00 GitHub noreply@github.com 2025-07-29T22:36:43+08:00 CANN: Add ggml_set_rows (#14943)
138b288b594eeb19d58e0ab0d74eae32009c9c20 bbd0f917797e9d524680f1b30d34a46eb06d7651 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-29T14:22:03+02:00 GitHub noreply@github.com 2025-07-29T14:22:03+02:00 cuda : add softcap fusion (#14907)
bbd0f917797e9d524680f1b30d34a46eb06d7651 0a5036bee9cfb946870689db4400e9e0d17844c9 Johannes Gäßler johannesg@5d6.de 2025-07-29T10:40:50+02:00 GitHub noreply@github.com 2025-07-29T10:40:50+02:00 server-bench: make seed choice configurable (#14929)
0a5036bee9cfb946870689db4400e9e0d17844c9 8ad7b3e65b5834e5574c2f5640056c9047b5d93b Aman Gupta amangupta052@gmail.com 2025-07-29T14:45:18+08:00 GitHub noreply@github.com 2025-07-29T14:45:18+08:00 CUDA: add roll (#14919)
8ad7b3e65b5834e5574c2f5640056c9047b5d93b bda62193b2a6bebbf515c3c389303094a44458c1 lhez lih@qti.qualcomm.com 2025-07-28T09:50:17-07:00 GitHub noreply@github.com 2025-07-28T18:50:17+02:00 opencl : add ops docs (#14910)
bda62193b2a6bebbf515c3c389303094a44458c1 c556418b600ad5792440942079d93e393595688b Leonard Mosescu tlemo@users.noreply.github.com 2025-07-28T09:04:27-07:00 GitHub noreply@github.com 2025-07-28T18:04:27+02:00 test-backend-ops : extend test case filtering (#14865)
c556418b600ad5792440942079d93e393595688b db16e2831c0f344f041af3d067db81c42b16eb22 Radoslav Gerganov rgerganov@gmail.com 2025-07-28T18:59:04+03:00 GitHub noreply@github.com 2025-07-28T18:59:04+03:00 llama-bench : use local GPUs along with RPC servers (#14917)
db16e2831c0f344f041af3d067db81c42b16eb22 cd1fce6d4f9c191f1c7429cc96f61281c3b63ffc xctan xc-tan@outlook.com 2025-07-28T23:40:24+08:00 GitHub noreply@github.com 2025-07-28T17:40:24+02:00 ggml-cpu : deduplicate scalar implementations (#14897)
cd1fce6d4f9c191f1c7429cc96f61281c3b63ffc 00fa15fedc79263fa0285e6a3bbb0cfb3e3878a2 Akarshan Biswas akarshan@menlo.ai 2025-07-28T20:32:15+05:30 GitHub noreply@github.com 2025-07-28T20:32:15+05:30 SYCL: Add set_rows support for quantized types (#14883)
00fa15fedc79263fa0285e6a3bbb0cfb3e3878a2 946b1f685909c8c9c044f145bce819c02f327eaa Xuan-Son Nguyen son@huggingface.co 2025-07-28T15:01:48+02:00 GitHub noreply@github.com 2025-07-28T15:01:48+02:00 mtmd : add support for Voxtral (#14862)
946b1f685909c8c9c044f145bce819c02f327eaa 6c6e397affc4fac717e718364fb4b635cec6433a Johannes Gäßler johannesg@5d6.de 2025-07-28T14:30:22+02:00 GitHub noreply@github.com 2025-07-28T14:30:22+02:00 CUDA: fix pointer incrementation in FA (#14916)
6c6e397affc4fac717e718364fb4b635cec6433a afc0e8969896ada62238da07b98731e5a4b12ba4 Dongliang Wei 121270393+wdl339@users.noreply.github.com 2025-07-28T19:47:00+08:00 GitHub noreply@github.com 2025-07-28T13:47:00+02:00 model : add support for SmallThinker series (#14898)
afc0e8969896ada62238da07b98731e5a4b12ba4 a5771c9eea801f573dc7375416e4c3306543563d Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-07-28T11:05:53+01:00 GitHub noreply@github.com 2025-07-28T11:05:53+01:00 sycl: refactor quantization to q8_1 (#14815)
a5771c9eea801f573dc7375416e4c3306543563d c35f9eaf095e0db3aa9b77b56846bab5d3bf5661 Georgi Gerganov ggerganov@gmail.com 2025-07-28T11:01:03+03:00 GitHub noreply@github.com 2025-07-28T10:01:03+02:00 ops : update BLAS (#14914)
c35f9eaf095e0db3aa9b77b56846bab5d3bf5661 1f45f2890ef7f365ba0a45e08a8d1f46b8bc6b9e Georgi Gerganov ggerganov@gmail.com 2025-07-28T08:22:56+03:00 GitHub noreply@github.com 2025-07-28T08:22:56+03:00 ops : update Metal (#14912)
1f45f2890ef7f365ba0a45e08a8d1f46b8bc6b9e 613c5095c33de9b7bbf1097d7c32510f51d58b01 Georgi Gerganov ggerganov@gmail.com 2025-07-28T08:14:20+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-28T08:15:01+03:00 sync : ggml
613c5095c33de9b7bbf1097d7c32510f51d58b01 7f97599581fcf0c37432dd3b1f503b91bed97695 Kai Pastor dg0yt@darc.de 2025-07-24T19:58:02+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-28T08:15:01+03:00 cmake : Indent ggml-config.cmake (ggml/1310)
7f97599581fcf0c37432dd3b1f503b91bed97695 bf78f5439ee8e82e367674043303ebf8e92b4805 Ed Addario 29247825+EAddario@users.noreply.github.com 2025-07-27T22:31:11+01:00 GitHub noreply@github.com 2025-07-27T23:31:11+02:00 quantize : update README.md (#14905)
bf78f5439ee8e82e367674043303ebf8e92b4805 bbfc84927481d1e59d8af6939b93b76850f0ab53 Ruben Ortlam picard12@live.de 2025-07-27T15:33:08+02:00 GitHub noreply@github.com 2025-07-27T15:33:08+02:00 vulkan: add ops docs (#14900)
bbfc84927481d1e59d8af6939b93b76850f0ab53 ca0ef2dddb022cb1337d775cd05cd27d7808aff4 Akarshan Biswas akarshan@menlo.ai 2025-07-27T17:52:58+05:30 GitHub noreply@github.com 2025-07-27T17:52:58+05:30 SYCL: add ops doc (#14901)
ca0ef2dddb022cb1337d775cd05cd27d7808aff4 89d1029559bd2968f76db854f9f113d73e34527c Daniel Bevenius daniel.bevenius@gmail.com 2025-07-27T12:10:51+02:00 GitHub noreply@github.com 2025-07-27T12:10:51+02:00 llama : clarify comment about pp and tg graphs [no ci] (#14895)
89d1029559bd2968f76db854f9f113d73e34527c f1a4e72de5950ab7136aeadddd675caf30dd6b3f Erik Scholz Green-Sky@users.noreply.github.com 2025-07-27T12:04:33+02:00 GitHub noreply@github.com 2025-07-27T12:04:33+02:00 vulkan : add fp16 support for the conv_2d kernel (#14872)
f1a4e72de5950ab7136aeadddd675caf30dd6b3f 4762ad7316dcdec20016ab5985fb46a27902204d Jeff Bolz jbolz@nvidia.com 2025-07-27T04:05:34-05:00 GitHub noreply@github.com 2025-07-27T11:05:34+02:00 vulkan: skip empty set_rows to avoid invalid API usage (#14860)
4762ad7316dcdec20016ab5985fb46a27902204d 1dc9614e0673e794d2e2bf88ba04f7d57b63a57b Gabriel Larson 55459720+gabriellarson@users.noreply.github.com 2025-07-27T03:18:37-05:00 GitHub noreply@github.com 2025-07-27T11:18:37+03:00 model : make rope_yarn_log_mul optional for deepseek2 (#14896)
1dc9614e0673e794d2e2bf88ba04f7d57b63a57b 446595b9b3a113d9ba10506922c3a156cca9d477 Shunta Saito shunta.saito@gmail.com 2025-07-27T16:38:44+09:00 GitHub noreply@github.com 2025-07-27T09:38:44+02:00 llama : fix kq_scale for the attention layers of PLaMo2 (#14892)
446595b9b3a113d9ba10506922c3a156cca9d477 66906cd82a4a1fd10151707cee3f66cb61fc4055 Aman Gupta amangupta052@gmail.com 2025-07-27T09:36:43+08:00 GitHub noreply@github.com 2025-07-27T09:36:43+08:00 Docs: add instructions for adding backends (#14889)
66906cd82a4a1fd10151707cee3f66cb61fc4055 11dd5a44eb180e1d69fac24d3852b5222d66fb7f deepsek 166548550+deepsek@users.noreply.github.com 2025-07-26T18:28:14-04:00 GitHub noreply@github.com 2025-07-27T00:28:14+02:00 HIP: Enable Matrix cores for MMQ Kernels, Enable stream-K for CDNA 3 (#14624)
11dd5a44eb180e1d69fac24d3852b5222d66fb7f 9b8f3c6c776d77045ee4f7f13fdf7863dfe59e5b hipudding huafengchun@gmail.com 2025-07-26T17:56:18+08:00 GitHub noreply@github.com 2025-07-26T17:56:18+08:00 CANN: Implement GLU ops (#14884)
9b8f3c6c776d77045ee4f7f13fdf7863dfe59e5b c7f3169cd523140a288095f2d79befb20a0b73f4 R0CKSTAR yeahdongcn@gmail.com 2025-07-26T10:36:02+08:00 GitHub noreply@github.com 2025-07-26T10:36:02+08:00 musa: fix build warnings (unused variable) (#14869)
c7f3169cd523140a288095f2d79befb20a0b73f4 793c0d7f46384001738c337d7afa46b45ae32745 Aaron Teo aaron.teo1@ibm.com 2025-07-26T01:09:03+08:00 GitHub noreply@github.com 2025-07-25T19:09:03+02:00 ggml-cpu : disable GGML_NNPA by default due to instability (#14880)
793c0d7f46384001738c337d7afa46b45ae32745 ce111d39d666f4a3b6a561ad020f6feb8cc67790 Gabe Goodhart ghart@us.ibm.com 2025-07-25T10:47:39-06:00 GitHub noreply@github.com 2025-07-25T10:47:39-06:00 metal: SSM_SCAN performance (#14743)
ce111d39d666f4a3b6a561ad020f6feb8cc67790 e7fecba93416c8aaf343932b5e36dd5e208a815e lhez lih@qti.qualcomm.com 2025-07-25T08:12:13-07:00 GitHub noreply@github.com 2025-07-25T17:12:13+02:00 opencl: add fused `rms_norm_mul` (#14841)
e7fecba93416c8aaf343932b5e36dd5e208a815e e2b7621e7c265a6739225125cf9c534f471b3472 wooksong wook16.song@samsung.com 2025-07-25T23:25:05+09:00 GitHub noreply@github.com 2025-07-25T16:25:05+02:00 docs : update HOWTO‑add‑model.md for ModelBase and new model classes (#14874)
e2b7621e7c265a6739225125cf9c534f471b3472 c1dbea752a630169c104118fd0c82f3ffcb19c91 Oliver Simons osimons@nvidia.com 2025-07-25T13:29:57+02:00 GitHub noreply@github.com 2025-07-25T14:29:57+03:00 ggml : remove invalid portPos specifiers from dot files (#14838)
c1dbea752a630169c104118fd0c82f3ffcb19c91 749e0d27f0247337869f4698f59dd7fafba94326 Georgi Gerganov ggerganov@gmail.com 2025-07-25T14:28:06+03:00 GitHub noreply@github.com 2025-07-25T14:28:06+03:00 context : restore preemptive sched reset when LLAMA_SET_ROWS=0 (#14870)
749e0d27f0247337869f4698f59dd7fafba94326 64bf1c3744053cf7def10aeed21ff48883ee755b kiwi 122582483+kiwi142857@users.noreply.github.com 2025-07-25T19:08:04+08:00 GitHub noreply@github.com 2025-07-25T13:08:04+02:00 mtmd : fix 32-bit narrowing issue in export-lora and mtmd clip (#14503)
64bf1c3744053cf7def10aeed21ff48883ee755b c12bbde37258c6d053322d1cedd4a9672109ae58 Chris Rohlf chris.rohlf@gmail.com 2025-07-25T06:17:02-04:00 GitHub noreply@github.com 2025-07-25T12:17:02+02:00 rpc : check for null buffers in get/set/copy tensor endpoints (#14868)
c12bbde37258c6d053322d1cedd4a9672109ae58 3f4fc97f1d745f1d5d3c853949503136d419e6de Diego Devesa slarengh@gmail.com 2025-07-25T01:07:26-07:00 GitHub noreply@github.com 2025-07-25T11:07:26+03:00 sched : fix multiple evaluations of the same graph with pipeline parallelism (#14855)
3f4fc97f1d745f1d5d3c853949503136d419e6de 2df255da3cea108de0ae9b302ffdd31674b6d88d R0CKSTAR yeahdongcn@gmail.com 2025-07-25T03:05:37+08:00 GitHub noreply@github.com 2025-07-24T20:05:37+01:00 musa: upgrade musa sdk to rc4.2.0 (#14498)
2df255da3cea108de0ae9b302ffdd31674b6d88d 60f816a79dd74007158745530e71738aa6caa67e Georgi Gerganov ggerganov@gmail.com 2025-07-24T18:30:33+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-24T20:27:23+03:00 sync : ggml
60f816a79dd74007158745530e71738aa6caa67e 5592f278b6ec6aa4a1793e89e8c61838a12ebc9d Kai Pastor dg0yt@darc.de 2025-07-22T20:13:21+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-24T20:27:23+03:00 cmake : fix usage issues (ggml/1257)
5592f278b6ec6aa4a1793e89e8c61838a12ebc9d e4868d16d24dec55e61bcaadaca28feed8f98b13 Daniel Bevenius daniel.bevenius@gmail.com 2025-07-21T15:53:12+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-24T20:27:23+03:00 ggml-cpu : remove stdlib include from repack.cpp (ggml/1276)
e4868d16d24dec55e61bcaadaca28feed8f98b13 820de57d4faa427a3d0bfb14e48057247fae036e Georgi Gerganov ggerganov@gmail.com 2025-07-24T16:31:48+03:00 GitHub noreply@github.com 2025-07-24T16:31:48+03:00 context : perform output reorder lazily upon access after sync (#14853)
820de57d4faa427a3d0bfb14e48057247fae036e cb4a63aad6650c2b536a7578403935388cb2920e Xuan-Son Nguyen son@huggingface.co 2025-07-24T13:59:56+02:00 GitHub noreply@github.com 2025-07-24T13:59:56+02:00 chat : fix kimi-k2 chat template (#14852)
cb4a63aad6650c2b536a7578403935388cb2920e 86f5623d904cfd392fdeb14a143097b4074660f6 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-07-24T11:09:57+01:00 GitHub noreply@github.com 2025-07-24T11:09:57+01:00 sycl: fixed semantics of block offset calculation (#14814)
86f5623d904cfd392fdeb14a143097b4074660f6 39cffdf18855e0d2beba62572542251d87421e73 yummy 57988893+jk3456a@users.noreply.github.com 2025-07-24T17:50:51+08:00 GitHub noreply@github.com 2025-07-24T11:50:51+02:00 llama : fix MiniCPM inference after Granite Four changes (#14850)
39cffdf18855e0d2beba62572542251d87421e73 065908cb09adff8b2a5f1173f80050d5d9a6790b Pouya PooyaGhahramanian@Gmail.com 2025-07-24T12:26:44+03:00 GitHub noreply@github.com 2025-07-24T11:26:44+02:00 docs: add libcurl-dev install hint for Linux distros (#14801)
065908cb09adff8b2a5f1173f80050d5d9a6790b 4ec6291a2407404de52239c1f9ca66c07e7fb28b Georgi Gerganov ggerganov@gmail.com 2025-07-24T10:24:05+03:00 GitHub noreply@github.com 2025-07-24T10:24:05+03:00 metal : fix fusion across different encoders (#14849)
4ec6291a2407404de52239c1f9ca66c07e7fb28b a12363bbf0ca11f787dee9756861043136edc0df Donghyeon Jeong 54725479+djeong20@users.noreply.github.com 2025-07-24T13:50:41+09:00 GitHub noreply@github.com 2025-07-24T12:50:41+08:00 sycl: fix undefined variable in work group size check (#14843)
a12363bbf0ca11f787dee9756861043136edc0df a86f52b2859dae4db5a7a0bbc0f1ad9de6b43ec6 jacekpoplawski 67507230+jacekpoplawski@users.noreply.github.com 2025-07-23T23:23:57+02:00 GitHub noreply@github.com 2025-07-23T23:23:57+02:00 convert : text-only support for GLM-4.1V-9B-Thinking (#14823)
a86f52b2859dae4db5a7a0bbc0f1ad9de6b43ec6 b284197df426fb189cdcfe56a43c863a788ac756 Johannes Gäßler johannesg@5d6.de 2025-07-23T21:43:25+02:00 GitHub noreply@github.com 2025-07-23T21:43:25+02:00 CUDA: fix overflow in FA, tune performance (#14840)
b284197df426fb189cdcfe56a43c863a788ac756 221c0e0c5841a814e95b9bfd549de9d6ae00ac6e Johannes Gäßler johannesg@5d6.de 2025-07-23T18:22:30+02:00 GitHub noreply@github.com 2025-07-23T18:22:30+02:00 CUDA: fix compilation with GGML_CUDA_F16 (#14837)
221c0e0c5841a814e95b9bfd549de9d6ae00ac6e 07a19e27a26f76d34be62da53807f93131fb3cab Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-23T14:27:54+02:00 GitHub noreply@github.com 2025-07-23T14:27:54+02:00 ci : correct label refactor->refactoring (#14832)
07a19e27a26f76d34be62da53807f93131fb3cab 18f3b5ff9e5eda4e7d04bceff8ffdccb0a696ed8 Johannes Gäßler johannesg@5d6.de 2025-07-23T12:35:53+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-23T14:08:09+03:00 CUDA: fix quantized KV cache + multiple sequences (#14822)
18f3b5ff9e5eda4e7d04bceff8ffdccb0a696ed8 7233358d29df3dfbf80693f2de7e5865401ad724 Georgi Gerganov ggerganov@gmail.com 2025-07-18T13:36:27+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-23T14:08:09+03:00 tests : add non-cont K,V FA tests
7233358d29df3dfbf80693f2de7e5865401ad724 6c88b3bb2509d980e6a64c50fdf8dd304929f770 l3utterfly gc.pthzfoldr@gmail.com 2025-07-23T16:16:41+08:00 GitHub noreply@github.com 2025-07-23T11:16:41+03:00 memory : handle saving/loading null layers in recurrent memory (#14675)
6c88b3bb2509d980e6a64c50fdf8dd304929f770 14c28dfc50a2e3915e697122cd3c5343224009be lixing-star 104126818+lixing-star@users.noreply.github.com 2025-07-23T14:39:51+08:00 GitHub noreply@github.com 2025-07-23T09:39:51+03:00 ggml: fix loongarch quantize_row_q8_1 error (#14827)
14c28dfc50a2e3915e697122cd3c5343224009be 8c988fa41db4d9dbc05abfd666c04def1259c645 chen fan 350211548@qq.com 2025-07-23T11:58:00+08:00 GitHub noreply@github.com 2025-07-23T11:58:00+08:00 CANN: weight format to NZ for Ascend310P3 (#14407)
8c988fa41db4d9dbc05abfd666c04def1259c645 acd6cb1c41676f6bbb25c2a76fa5abeb1719301e Aman Gupta amangupta052@gmail.com 2025-07-23T09:25:42+08:00 GitHub noreply@github.com 2025-07-23T09:25:42+08:00 CUDA: add fused rms norm (#14800)
acd6cb1c41676f6bbb25c2a76fa5abeb1719301e 84712b60439453fb393c2ca753cee682a4ad41f5 Csaba Kecskemeti csaba.kecskemeti@gmail.com 2025-07-22T09:29:43-07:00 GitHub noreply@github.com 2025-07-22T19:29:43+03:00 ggml : model card yaml tab->2xspace (#14819)
84712b60439453fb393c2ca753cee682a4ad41f5 d4d1522b20809a350ffb094db20f40f17d3ab80f Jeff Bolz jbolz@nvidia.com 2025-07-22T10:35:21-05:00 GitHub noreply@github.com 2025-07-22T17:35:21+02:00 vulkan: fix rms_norm_mul to handle broadcasting dim0 (#14817)
d4d1522b20809a350ffb094db20f40f17d3ab80f d1aa0cc5d13ec3fa553de5d298f9166679e58479 Molly Sophia mollysophia379@gmail.com 2025-07-22T23:01:29+08:00 GitHub noreply@github.com 2025-07-22T23:01:29+08:00 llama : add model type detection for rwkv7 7B&14B (#14816)
d1aa0cc5d13ec3fa553de5d298f9166679e58479 c8ade30036139e32108fee53d8b7164dbfda4bee Ed Addario 29247825+EAddario@users.noreply.github.com 2025-07-22T13:33:37+01:00 GitHub noreply@github.com 2025-07-22T14:33:37+02:00 imatrix: add option to display importance score statistics for a given imatrix file (#12718)
c8ade30036139e32108fee53d8b7164dbfda4bee e28c0b80c249b5618c3a0d5e960f63929f380ac9 stduhpf stephduh@live.fr 2025-07-22T12:51:03+02:00 GitHub noreply@github.com 2025-07-22T12:51:03+02:00 Mtmd: add a way to select device for vision encoder (#14236)
e28c0b80c249b5618c3a0d5e960f63929f380ac9 8e6f8bc875358968b63e08f7bbbe0a288f29d856 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-22T12:33:10+02:00 GitHub noreply@github.com 2025-07-22T12:33:10+02:00 cuda : implement bf16 cpy ops and enable bf16 cont (#14763)
8e6f8bc875358968b63e08f7bbbe0a288f29d856 adef81781a15083f218eae6c488b95cdad781971 lhez lih@qti.qualcomm.com 2025-07-21T23:53:30-07:00 GitHub noreply@github.com 2025-07-22T08:53:30+02:00 opencl: remove unreachable `return` (#14806)
adef81781a15083f218eae6c488b95cdad781971 48b86c4fdb1b1246d9fe17d2d53e3a1c2a0c3245 Molly Sophia mollysophia379@gmail.com 2025-07-22T09:24:22+08:00 GitHub noreply@github.com 2025-07-22T09:24:22+08:00 server : allow setting `--reverse-prompt` arg (#14799)
48b86c4fdb1b1246d9fe17d2d53e3a1c2a0c3245 38d3af1b73302377111e88ddd725257638339286 R0CKSTAR yeahdongcn@gmail.com 2025-07-22T07:45:26+08:00 GitHub noreply@github.com 2025-07-22T07:45:26+08:00 cuda: remove linking to cublasLt (#14790)
38d3af1b73302377111e88ddd725257638339286 6c9ee3b17e19dcc82ab93d52ae46fdd0226d4777 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-21T22:55:10+02:00 GitHub noreply@github.com 2025-07-21T13:55:10-07:00 opencl: fix `im2col` when `KW!=KH` (#14803)
6c9ee3b17e19dcc82ab93d52ae46fdd0226d4777 cd465d823c378853f1f6570eebfb77f69c7e1d39 rmatif kingrealriadh@gmail.com 2025-07-21T19:03:19+02:00 GitHub noreply@github.com 2025-07-21T10:03:19-07:00 opencl: add conv2d kernel (#14403)
cd465d823c378853f1f6570eebfb77f69c7e1d39 922042601b8a16877ccb1c2afaa2071f76734f10 Romain Biessy romain.biessy@codeplay.com 2025-07-21T18:39:29+02:00 GitHub noreply@github.com 2025-07-21T18:39:29+02:00 sycl: Fix im2col (#14797)
922042601b8a16877ccb1c2afaa2071f76734f10 2ba1333b35e471b344974dde553db11bf1c2836f Charles Xu charles.xu@arm.com 2025-07-21T15:49:52+02:00 GitHub noreply@github.com 2025-07-21T16:49:52+03:00 kleidiai: add support for get_rows (#14676)
2ba1333b35e471b344974dde553db11bf1c2836f c2e058f1b4e799f1be085560c1bcef95b7b5ed02 Radoslav Gerganov rgerganov@gmail.com 2025-07-21T15:03:49+03:00 GitHub noreply@github.com 2025-07-21T14:03:49+02:00 docs : fix backends table in README.md (#14796)
c2e058f1b4e799f1be085560c1bcef95b7b5ed02 c82d48ec23fb8749c341d0838f6891fd5f6b6da0 Jeff Bolz jbolz@nvidia.com 2025-07-21T06:35:40-05:00 GitHub noreply@github.com 2025-07-21T13:35:40+02:00 vulkan/cuda: Fix im2col when KW!=KH (#14789)
c82d48ec23fb8749c341d0838f6891fd5f6b6da0 b4efd77f8ab407836ca73a5176f041650c5b2411 Molly Sophia mollysophia379@gmail.com 2025-07-21T17:38:36+08:00 GitHub noreply@github.com 2025-07-21T17:38:36+08:00 llama : fix `--reverse-prompt` crashing issue (#14794)
b4efd77f8ab407836ca73a5176f041650c5b2411 2be60cbc2707359241c2784f9d2e30d8fc7cdabb IsaacDynamo 61521674+IsaacDynamo@users.noreply.github.com 2025-07-21T09:24:51+02:00 GitHub noreply@github.com 2025-07-21T10:24:51+03:00 server : add parse_special option to /tokenize endpoint (#14783)
2be60cbc2707359241c2784f9d2e30d8fc7cdabb b526ad2668944a7b2b1721f60679153646313831 Aman Gupta amangupta052@gmail.com 2025-07-21T02:13:47+08:00 GitHub noreply@github.com 2025-07-20T20:13:47+02:00 docs : fix link for tools/perplexity in README.md (#14780)
b526ad2668944a7b2b1721f60679153646313831 938b785764683c298e7805e712f8728489cc2f18 rspOverflow 217881046+rspOverflow@users.noreply.github.com 2025-07-20T23:55:32+07:00 GitHub noreply@github.com 2025-07-20T18:55:32+02:00 Documentation: Further revisions to the Vulkan section in build.md (#14785)
938b785764683c298e7805e712f8728489cc2f18 36c153248faf969af1b62ab231348694b2047b8b Aman Gupta amangupta052@gmail.com 2025-07-20T19:42:34+08:00 GitHub noreply@github.com 2025-07-20T19:42:34+08:00 Clang-format: local files first + fix BinPacking (#14779)
36c153248faf969af1b62ab231348694b2047b8b a979ca22db0d737af1e548a73291193655c6be99 0cc4m picard12@live.de 2025-07-19T22:47:21+02:00 GitHub noreply@github.com 2025-07-19T23:47:21+03:00 Contrib: add 0cc4m as codeowner for Vulkan backend (#14775)
a979ca22db0d737af1e548a73291193655c6be99 90083283ec254fa8d33897746dea229aee401b37 Ervin Áron Tasnádi etasnadi@protonmail.com 2025-07-19T21:59:08+02:00 GitHub noreply@github.com 2025-07-19T21:59:08+02:00 ggml: adds CONV_2D op and direct GEMM Vulkan implementation (#14316)
90083283ec254fa8d33897746dea229aee401b37 d4b91ea7b2da253e1355b503f0fcb7b428ce005d compilade git@compilade.net 2025-07-19T12:51:22-04:00 GitHub noreply@github.com 2025-07-19T12:51:22-04:00 imatrix : use GGUF to store importance matrices (#9400)
d4b91ea7b2da253e1355b503f0fcb7b428ce005d 83f5872404baa39d826af2ef66351e63c64205a8 Peter0x44 peter0x44@disroot.org 2025-07-19T16:58:03+01:00 GitHub noreply@github.com 2025-07-19T17:58:03+02:00 vulkan: Add logging for bf16 features to ggml_vk_print_gpu_info (#13274) (#14707)
83f5872404baa39d826af2ef66351e63c64205a8 f0d4d176df72734a543c29eef9f942850c13311e 0cc4m picard12@live.de 2025-07-19T17:47:53+02:00 GitHub noreply@github.com 2025-07-19T17:47:53+02:00 Vulkan: Fix fprintf format-security warning (#14770)
f0d4d176df72734a543c29eef9f942850c13311e b17230917c18a25af9cd143a941001466af845a2 rspOverflow 217881046+rspOverflow@users.noreply.github.com 2025-07-19T17:18:36+07:00 GitHub noreply@github.com 2025-07-19T12:18:36+02:00 Documentation: Update build.md's Vulkan section (#14736)
b17230917c18a25af9cd143a941001466af845a2 bf9087f59aab940cf312b85a67067ce33d9e365a Georgi Gerganov ggerganov@gmail.com 2025-07-19T11:46:12+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-19T11:46:50+03:00 sync : ggml
bf9087f59aab940cf312b85a67067ce33d9e365a 9fb1042ce6719bc46dbe88ab013148aabe3105f1 Georgi Gerganov ggerganov@gmail.com 2025-07-18T20:37:26+03:00 GitHub noreply@github.com 2025-07-18T20:37:26+03:00 metal : fuse add, mul + add tests (#14596)
9fb1042ce6719bc46dbe88ab013148aabe3105f1 2adf8d83acdb9b1bf58db6c9729ac9dc6847a58b Georgi Gerganov ggerganov@gmail.com 2025-07-18T20:08:33+03:00 GitHub noreply@github.com 2025-07-18T20:08:33+03:00 graph : fix graph reuse reset of params (#14760)
2adf8d83acdb9b1bf58db6c9729ac9dc6847a58b 021cc28bef4dd7d0bf9c91dbbd0803caa6cb15f2 Georgi Gerganov ggerganov@gmail.com 2025-07-18T17:33:41+03:00 GitHub noreply@github.com 2025-07-18T17:33:41+03:00 parallel : add option for different RNG seeds (#14757)
021cc28bef4dd7d0bf9c91dbbd0803caa6cb15f2 d498af3d5a00f96bdd37b534860f03a6d9e98d39 Oliver Simons oliver.simons@posteo.de 2025-07-18T13:35:32+02:00 GitHub noreply@github.com 2025-07-18T04:35:32-07:00 cuda : Fix Gemma3n not executed as CUDA_GRAPH on NVGPUs (#14741)
d498af3d5a00f96bdd37b534860f03a6d9e98d39 eacdeb5bfcb6c6cd54461fd0e9f04cab78bf975b Georgi Gerganov ggerganov@gmail.com 2025-07-18T14:31:15+03:00 GitHub noreply@github.com 2025-07-18T14:31:15+03:00 graph : avoid huge warm-up graphs for MoE models (#14753)
eacdeb5bfcb6c6cd54461fd0e9f04cab78bf975b e0cb5c5cb8a61ac232130cf6bf878035f93824d9 Georgi Gerganov ggerganov@gmail.com 2025-07-18T11:53:55+03:00 GitHub noreply@github.com 2025-07-18T11:53:55+03:00 model : fix build after merge conflict (#14754)
e0cb5c5cb8a61ac232130cf6bf878035f93824d9 f9a31eea06a859e34cecb88b4d020c7f03d86cc4 lgai-exaone exaonemodels@lgresearch.ai 2025-07-18T17:45:49+09:00 GitHub noreply@github.com 2025-07-18T10:45:49+02:00 model : add EXAONE 4.0 support (#14630)
f9a31eea06a859e34cecb88b4d020c7f03d86cc4 8f974bc1e980c06833504276021072e7a4088c81 Aman Gupta amangupta052@gmail.com 2025-07-18T14:54:18+08:00 GitHub noreply@github.com 2025-07-18T14:54:18+08:00 CUDA: set_rows + cpy.cu refactor (#14712)
8f974bc1e980c06833504276021072e7a4088c81 09651d09ffc1e941bd1be23163abf5495c416547 Georgi Gerganov ggerganov@gmail.com 2025-07-18T08:29:28+03:00 GitHub noreply@github.com 2025-07-18T08:29:28+03:00 graph : refactor context to not pass gf explicitly (#14629)
09651d09ffc1e941bd1be23163abf5495c416547 349ea79fcebc75b0c55bf61594a47736966d4f95 Nexes the Elder 124105151+Nexesenex@users.noreply.github.com 2025-07-18T06:25:54+02:00 GitHub noreply@github.com 2025-07-18T07:25:54+03:00 graph : Pass the graph placeholder message in debug mode (#14748)
349ea79fcebc75b0c55bf61594a47736966d4f95 670e1360cd40f242ae76ba0966542fae6cb59392 Neo Zhang Jianyu jianyu.zhang@intel.com 2025-07-18T10:23:14+08:00 GitHub noreply@github.com 2025-07-18T10:23:14+08:00 use max work group size for device to replace the magic number (#14732)
670e1360cd40f242ae76ba0966542fae6cb59392 760b4484e3c192a2649a6ffd0d90086c5558f849 Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-07-18T01:17:16+02:00 GitHub noreply@github.com 2025-07-18T01:17:16+02:00 convert : fix Ernie4.5 MoE without shared experts (#14746)
760b4484e3c192a2649a6ffd0d90086c5558f849 cb887f1bc1001c92f7b4a595b9014f3a454a07ab Wroclaw wroclaw223@outlook.com 2025-07-18T00:18:16+02:00 GitHub noreply@github.com 2025-07-17T15:18:16-07:00 nix : use optionalAttrs for env mkDerivation attrset argument (#14726)
cb887f1bc1001c92f7b4a595b9014f3a454a07ab d6fb3f6b49b27ef1c0f4cf5128e041f7e7dc03af Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-07-17T23:15:32+02:00 GitHub noreply@github.com 2025-07-17T23:15:32+02:00 model: add Ernie 4.5 MoE support (#14658)
d6fb3f6b49b27ef1c0f4cf5128e041f7e7dc03af 01612b74090df592663cfa01f661c9628f403b59 Georgi Gerganov ggerganov@gmail.com 2025-07-17T20:52:33+03:00 GitHub noreply@github.com 2025-07-17T20:52:33+03:00 kv-cache : fix k-shift for multiple streams (#14742)
01612b74090df592663cfa01f661c9628f403b59 086cf81e88fb75287b71ff19c08a206b7bc2e02f Georgi Gerganov ggerganov@gmail.com 2025-07-17T19:08:33+03:00 GitHub noreply@github.com 2025-07-17T19:08:33+03:00 llama : reuse compute graphs (#14482)
086cf81e88fb75287b71ff19c08a206b7bc2e02f d9b691081c04ec5fb0daa9d2b979f915c142963d Tarek Dakhran tarek@liquid.ai 2025-07-17T09:22:11+02:00 GitHub noreply@github.com 2025-07-17T09:22:11+02:00 llama : fix parallel processing for lfm2 (#14705)
d9b691081c04ec5fb0daa9d2b979f915c142963d ad57d3edd2f48cf6dc41a98fd9b303435ecb4fb0 Georgi Gerganov ggerganov@gmail.com 2025-07-17T09:49:15+03:00 GitHub noreply@github.com 2025-07-17T09:49:15+03:00 kv-cache : opt mask set input (#14600)
ad57d3edd2f48cf6dc41a98fd9b303435ecb4fb0 1ba45d49822c39ca9a552c7b75efe0495ff400c3 Georgi Gerganov ggerganov@gmail.com 2025-07-17T09:45:54+03:00 GitHub noreply@github.com 2025-07-17T09:45:54+03:00 batch : fix uninitialized has_cpl flag (#14733)
1ba45d49822c39ca9a552c7b75efe0495ff400c3 19e5943d9e976b59ccd5a0a87ae3d2ff3da44390 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-17T01:52:08+02:00 GitHub noreply@github.com 2025-07-16T20:52:08-03:00 ci : disable failing vulkan crossbuilds (#14723)
19e5943d9e976b59ccd5a0a87ae3d2ff3da44390 496957e1cbcb522abc63aa18521036e40efce985 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-16T23:17:43+02:00 GitHub noreply@github.com 2025-07-16T23:17:43+02:00 convert : make hf token optional (#14717)
496957e1cbcb522abc63aa18521036e40efce985 21c021745d781edf9c44b4972ef6cbbf53b0ecff Diner Burger burger@diner.name 2025-07-16T15:17:25-04:00 GitHub noreply@github.com 2025-07-16T21:17:25+02:00 llama : fix parameter order for hybrid memory initialization (#14725)
21c021745d781edf9c44b4972ef6cbbf53b0ecff b0f0ecc3dce806c68609d375a2b3edc430d8db18 Reese Levine reeselevine1@gmail.com 2025-07-16T08:18:51-07:00 GitHub noreply@github.com 2025-07-16T18:18:51+03:00 ggml: Add initial WebGPU backend (#14521)
b0f0ecc3dce806c68609d375a2b3edc430d8db18 225e7a1438f4ea85eaa7b5ef3ab3b266ee4d9c06 tempstudio 49735574+tempstudio@users.noreply.github.com 2025-07-16T10:02:06-05:00 GitHub noreply@github.com 2025-07-16T18:02:06+03:00 model : support output bias for qwen2 (#14711)
225e7a1438f4ea85eaa7b5ef3ab3b266ee4d9c06 ab140198211385b85eeeb0abd549a4bbe259e10d Georgi Gerganov ggerganov@gmail.com 2025-07-16T16:35:42+03:00 GitHub noreply@github.com 2025-07-16T16:35:42+03:00 llama : add high-throughput mode (#14363)
ab140198211385b85eeeb0abd549a4bbe259e10d 64978340b0b4a0a6e2fb74270c1509383d2eff32 Aman Gupta amangupta052@gmail.com 2025-07-16T20:03:51+08:00 GitHub noreply@github.com 2025-07-16T20:03:51+08:00 Support diffusion models: Add Dream 7B (#14644)
64978340b0b4a0a6e2fb74270c1509383d2eff32 6ffd4e9c442e99afac3d138543ebf86d5fc5ee03 Georgi Gerganov ggerganov@gmail.com 2025-07-16T14:43:32+03:00 GitHub noreply@github.com 2025-07-16T14:43:32+03:00 ggml : add asserts (#14720)
6ffd4e9c442e99afac3d138543ebf86d5fc5ee03 e4841d24d3485ea4af54f8b65a19ec3123f0ff3c Georgi Gerganov ggerganov@gmail.com 2025-07-16T14:04:12+03:00 GitHub noreply@github.com 2025-07-16T14:04:12+03:00 server : pre-calculate EOG logit biases (#14721)
e4841d24d3485ea4af54f8b65a19ec3123f0ff3c 538cc77f7f44dfa047dba6a06d90c86dda69cf1d Shunta Saito shunta.saito@gmail.com 2025-07-16T19:12:22+09:00 GitHub noreply@github.com 2025-07-16T12:12:22+02:00 llama : fix parallel processing for plamo2 (#14716)
538cc77f7f44dfa047dba6a06d90c86dda69cf1d 5cae76654113160f691f581930b69fc5535e8159 Georgi Gerganov ggerganov@gmail.com 2025-07-16T12:13:57+03:00 GitHub noreply@github.com 2025-07-16T12:13:57+03:00 server : fix handling of the ignore_eos flag (#14710)
5cae76654113160f691f581930b69fc5535e8159 4b91d6f71f14040979bcdb7b6729b3bca93ec1c1 Johannes Gäßler johannesg@5d6.de 2025-07-16T09:33:28+02:00 GitHub noreply@github.com 2025-07-16T09:33:28+02:00 scripts: synthetic prompt mode for server-bench.py (#14695)
4b91d6f71f14040979bcdb7b6729b3bca93ec1c1 cf91f217f1c8b98b6db8e4ba6b480f017f81d206 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-16T08:52:04+02:00 GitHub noreply@github.com 2025-07-16T08:52:04+02:00 convert : only check for tokenizer folder if we need it (#14704)
cf91f217f1c8b98b6db8e4ba6b480f017f81d206 79e0b68c178656bb0632cb8602d2940b755077f8 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-16T08:51:12+02:00 GitHub noreply@github.com 2025-07-16T08:51:12+02:00 convert : add pre-computed hashes first to prevent order mishaps (#14701)
79e0b68c178656bb0632cb8602d2940b755077f8 c81f4192f91a1e209c1eec7a84fe5371ef9175da Min-Hua 136287195+Min-Hua@users.noreply.github.com 2025-07-16T12:00:42+08:00 GitHub noreply@github.com 2025-07-16T07:00:42+03:00 llama: add LLAMA_API to deprecated llama_kv_self_seq_div (#14708)
c81f4192f91a1e209c1eec7a84fe5371ef9175da 4a4f426944e79b79e389f9ed7b34831cb9b637ad Ed Addario 29247825+EAddario@users.noreply.github.com 2025-07-15T23:04:42+01:00 GitHub noreply@github.com 2025-07-16T00:04:42+02:00 gguf-py : dump bpw per layer and model in markdown mode (#14703)
4a4f426944e79b79e389f9ed7b34831cb9b637ad ba1ceb34566c889a1fc500efa79799ffed25d9b0 Gabriel Larson 55459720+gabriellarson@users.noreply.github.com 2025-07-15T14:54:22-05:00 GitHub noreply@github.com 2025-07-15T21:54:22+02:00 model : add Kimi-K2 support (#14654)
ba1ceb34566c889a1fc500efa79799ffed25d9b0 10a0351a97c25471aea0bbde9cca54d32d163eec Jeff Bolz jbolz@nvidia.com 2025-07-15T14:51:09-05:00 GitHub noreply@github.com 2025-07-15T21:51:09+02:00 vulkan: fix noncontig check for mat_mul_id splitting (#14683)
10a0351a97c25471aea0bbde9cca54d32d163eec 68e37a61a7b24863f67541db65b4f6195962a268 Jeff Bolz jbolz@nvidia.com 2025-07-15T14:32:11-05:00 GitHub noreply@github.com 2025-07-15T21:32:11+02:00 vulkan: add RTE variants for glu/add/sub/mul/div (#14653)
68e37a61a7b24863f67541db65b4f6195962a268 cbc68be51d88b1d5531643b926a4b359c3cff131 Shunta Saito shunta.saito@gmail.com 2025-07-16T01:11:42+09:00 GitHub noreply@github.com 2025-07-15T18:11:42+02:00 model : add PLaMo-2 support (#14560)
cbc68be51d88b1d5531643b926a4b359c3cff131 bdca38376f7e8dd928defe01ce6a16218a64b040 R0CKSTAR yeahdongcn@gmail.com 2025-07-15T15:28:53+08:00 GitHub noreply@github.com 2025-07-15T15:28:53+08:00 cuda: fix build warnings in set-rows.cu (unused variable) (#14687)
bdca38376f7e8dd928defe01ce6a16218a64b040 55c509daf51d25bfaee9c8b8ce6abff103d4473b Anton Mitkov anton_b_mitkov@abv.bg 2025-07-14T18:12:42+01:00 GitHub noreply@github.com 2025-07-14T18:12:42+01:00 sycl: Hotfix for non dnnl codepath (#14677)
55c509daf51d25bfaee9c8b8ce6abff103d4473b 9c9e4fc6354fc811efa06a8eb7a86d3315cec9c8 shalinib-ibm Shalini.Salomi.Bodapati@ibm.com 2025-07-14T18:46:42+05:30 GitHub noreply@github.com 2025-07-14T16:16:42+03:00 ggml : refactor llamafile_sgemm PPC code (#14673)
9c9e4fc6354fc811efa06a8eb7a86d3315cec9c8 494c5899cb76859f32ddd913534f2685fd684a3d Aman Gupta amangupta052@gmail.com 2025-07-14T21:01:41+08:00 GitHub noreply@github.com 2025-07-14T21:01:41+08:00 llama-context: add ability to get logits (#14672)
494c5899cb76859f32ddd913534f2685fd684a3d 0f4c6ec0f1a9607ba67071f8a02c69b0afc2f91e Johannes Gäßler johannesg@5d6.de 2025-07-14T13:14:30+02:00 GitHub noreply@github.com 2025-07-14T13:14:30+02:00 scripts: benchmark for HTTP server throughput (#14668)
0f4c6ec0f1a9607ba67071f8a02c69b0afc2f91e 65a3ebb0aa56d6c501466e0f950ad15105fd32d8 Akarshan Biswas akarshan@menlo.ai 2025-07-14T15:07:55+05:30 GitHub noreply@github.com 2025-07-14T10:37:55+01:00 SYCL: use 1D kernel for set_rows (#14618)
65a3ebb0aa56d6c501466e0f950ad15105fd32d8 0d9226763c82562186122f3b827fa3862864a19c Anton Mitkov anton_b_mitkov@abv.bg 2025-07-14T10:37:35+01:00 GitHub noreply@github.com 2025-07-14T10:37:35+01:00 sycl: Batched mulmat rework for oneDNN dispatch (#14617)
0d9226763c82562186122f3b827fa3862864a19c 982e347255723fe6d02e60ee30cfdd0559c884c5 Molly Sophia mollysophia379@gmail.com 2025-07-14T07:43:43+08:00 GitHub noreply@github.com 2025-07-14T07:43:43+08:00 llama : add jinja template for rwkv-world (#14665)
982e347255723fe6d02e60ee30cfdd0559c884c5 923e3ea2e3c96a0b4c208f53bc3bc90cdcdf13c0 Ed Addario 29247825+EAddario@users.noreply.github.com 2025-07-13T17:02:17+01:00 GitHub noreply@github.com 2025-07-13T18:02:17+02:00 quantize : fix minor logic flaw in --tensor-type (#14572)
923e3ea2e3c96a0b4c208f53bc3bc90cdcdf13c0 e743cddb60dc3a8815b9de7dd7d5c491e61b2259 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-13T15:01:24+02:00 GitHub noreply@github.com 2025-07-13T15:01:24+02:00 cuda : add set rows for bf16 (#14664)
e743cddb60dc3a8815b9de7dd7d5c491e61b2259 05fec5bd298d3c0243cbb9336e59b8b6aff75a81 Yavor Ivanov yavorgenadiev@gmail.com 2025-07-13T02:33:16-07:00 GitHub noreply@github.com 2025-07-13T11:33:16+02:00 cuda : add ELU support (#14657)
05fec5bd298d3c0243cbb9336e59b8b6aff75a81 dcf7f2ea3c4e3cf36dd9ab5a36785c00e6033267 Georgi Gerganov ggerganov@gmail.com 2025-07-13T10:36:33+03:00 GitHub noreply@github.com 2025-07-13T10:36:33+03:00 ggml : add build-time message to remind about ggml_set_rows (#14661)
dcf7f2ea3c4e3cf36dd9ab5a36785c00e6033267 84b396e0510855a95d591afdf1f21c562cb3712a Yavor Ivanov yavorgenadiev@gmail.com 2025-07-12T22:38:13-07:00 GitHub noreply@github.com 2025-07-13T08:38:13+03:00 metal : Add missing unary ops Metal support (#14660)
84b396e0510855a95d591afdf1f21c562cb3712a c31e60647def83d671bac5ab5b35579bf25d9aa1 Yavor Ivanov yavorgenadiev@gmail.com 2025-07-12T22:12:36-07:00 GitHub noreply@github.com 2025-07-13T08:12:36+03:00 cmake : Add CMake presets for Linux and GCC (#14656)
c31e60647def83d671bac5ab5b35579bf25d9aa1 67eade1bf93f40e7c4975e5bd0782f426c89cff4 Tarek Dakhran tarek@liquid.ai 2025-07-12T19:10:14+02:00 GitHub noreply@github.com 2025-07-12T19:10:14+02:00 tests : cover lfm2 cases in test_ssm_conv (#14651)
67eade1bf93f40e7c4975e5bd0782f426c89cff4 7de5c7cab61d4da4387ed9b216f88b96297bcc2d Tarek Dakhran tarek@liquid.ai 2025-07-12T19:07:08+02:00 GitHub noreply@github.com 2025-07-12T19:07:08+02:00 docs : add LFM2 to models section (#14650)
7de5c7cab61d4da4387ed9b216f88b96297bcc2d 8eff95544e817704d44bec20f9fc956ce76a33be Aman Gupta amangupta052@gmail.com 2025-07-12T21:31:38+08:00 GitHub noreply@github.com 2025-07-12T16:31:38+03:00 CUDA: add set rows for f32 and f16 (#14551)
8eff95544e817704d44bec20f9fc956ce76a33be 3120413ccd7797d4cbdee32ef89a641765d1f6c4 Georgi Gerganov ggerganov@gmail.com 2025-07-12T16:06:12+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-12T16:13:27+03:00 sync : ggml
3120413ccd7797d4cbdee32ef89a641765d1f6c4 215535701d659e873c52d2a9163d4616a42da4f7 Georgi Gerganov ggerganov@gmail.com 2025-07-12T12:39:32+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-12T14:25:44+03:00 vulkan : remove unused vars (#0)
215535701d659e873c52d2a9163d4616a42da4f7 74bb294591054a6bf0cc3f6e4637d87414cf2c46 Georgi Gerganov ggerganov@gmail.com 2025-07-12T12:39:27+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-12T14:25:44+03:00 sync : ggml
74bb294591054a6bf0cc3f6e4637d87414cf2c46 3e303b1107e4748cd50a675544d8c7fabb40ec54 Acly aclysia@gmail.com 2025-07-12T12:37:37+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-12T14:25:44+03:00 vulkan : implement bilinear interpolation (ggml/1291)
3e303b1107e4748cd50a675544d8c7fabb40ec54 0c1df14b5f8d992805cb22d0b77b44092a18aeab Acly aclysia@gmail.com 2025-07-12T12:32:32+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-12T14:25:44+03:00 vulkan : implement ggml_roll (ggml/1290)
0c1df14b5f8d992805cb22d0b77b44092a18aeab b3ad3a0191994d6c47b2bd389d5c7431526ecd2c Douglas Hanley thesecretaryofwar@gmail.com 2025-07-12T06:21:02-04:00 GitHub noreply@github.com 2025-07-12T13:21:02+03:00 server : fix pooled embedding output (#14645)
b3ad3a0191994d6c47b2bd389d5c7431526ecd2c 98197e5c98388470030d908f355ec5937dcccaaa Jeff Bolz jbolz@nvidia.com 2025-07-12T05:12:26-05:00 GitHub noreply@github.com 2025-07-12T12:12:26+02:00 vulkan: support SET_ROWS (#14587)
98197e5c98388470030d908f355ec5937dcccaaa f5e96b368f1acc7f53c390001b936517c4d18999 Jeff Bolz jbolz@nvidia.com 2025-07-12T04:51:58-05:00 GitHub noreply@github.com 2025-07-12T11:51:58+02:00 vulkan: optimizations for deepseek prompt processing (#14555)
f5e96b368f1acc7f53c390001b936517c4d18999 756aa1020aabc70b71b9a8efcd35fbf42b3bb9ab Tarek Dakhran t.dakhran@gmail.com 2025-07-11T20:27:01+02:00 GitHub noreply@github.com 2025-07-11T20:27:01+02:00 model : support LiquidAI LFM2 hybrid family (#14620)
756aa1020aabc70b71b9a8efcd35fbf42b3bb9ab aaa088d87f9006d56866085fe46e4b2755ef723f Slobodan Josic 127323561+slojosic-amd@users.noreply.github.com 2025-07-11T18:55:00+02:00 GitHub noreply@github.com 2025-07-11T18:55:00+02:00 HIP : Add HIP 7.0+ compatibility for hipBLAS compute types (#14634)
aaa088d87f9006d56866085fe46e4b2755ef723f 0d5375d54b258ec63edd1fb5d58c37d58ce8be8b Georgi Gerganov ggerganov@gmail.com 2025-07-11T16:07:55+03:00 GitHub noreply@github.com 2025-07-11T16:07:55+03:00 readme : add hot PRs (#14636)
0d5375d54b258ec63edd1fb5d58c37d58ce8be8b 576c82eda210ca0111c04f5256bf77897a4d4cc4 Georgi Gerganov ggerganov@gmail.com 2025-07-11T13:46:07+03:00 GitHub noreply@github.com 2025-07-11T13:46:07+03:00 llama : move enum llama_vocab_pre_type to implementation (#14631)
576c82eda210ca0111c04f5256bf77897a4d4cc4 0aedae00e6fb48680324a5ac5da9cba0e35de6b5 Dowon ks2515@naver.com 2025-07-11T16:36:04+09:00 GitHub noreply@github.com 2025-07-11T09:36:04+02:00 vocab : add midm-2.0 model pre-tokenizer (#14626)
0aedae00e6fb48680324a5ac5da9cba0e35de6b5 6bdda13981d6c8189b7dc4f9fb8ecb91c21529f8 Gabe Goodhart ghart@us.ibm.com 2025-07-10T18:20:13-06:00 GitHub noreply@github.com 2025-07-11T02:20:13+02:00 model : Granite Four (#13550)
6bdda13981d6c8189b7dc4f9fb8ecb91c21529f8 0b8855775c6b873931d40b77a5e42558aacbde52 rmatif kingrealriadh@gmail.com 2025-07-10T23:58:12+02:00 GitHub noreply@github.com 2025-07-10T14:58:12-07:00 opencl: add tiled mul_mat_f16_f32 (#14535)
0b8855775c6b873931d40b77a5e42558aacbde52 4bb625b713fd9b294b4f7af87eaa752b710c7cc1 lhez quic_lih@quicinc.com 2025-07-10T11:48:52-07:00 GitHub noreply@github.com 2025-07-10T11:48:52-07:00 opencl: add `set_rows` for `f16` and `f32` (#14547)
4bb625b713fd9b294b4f7af87eaa752b710c7cc1 11ee0fea2a24da1d3206eeba8fc52b759d9dfb24 Ryan Mangeno 160974989+ryan-mangeno@users.noreply.github.com 2025-07-10T13:41:00-04:00 GitHub noreply@github.com 2025-07-10T19:41:00+02:00 Smoldocling support (#14597)
11ee0fea2a24da1d3206eeba8fc52b759d9dfb24 a457551332853ef19d0796fec12b62c538126ea5 Aman Gupta amangupta052@gmail.com 2025-07-10T23:29:01+08:00 GitHub noreply@github.com 2025-07-10T23:29:01+08:00 Docs: script to auto-generate ggml operations docs (#14598)
a457551332853ef19d0796fec12b62c538126ea5 704bb7a71c01dc07c1478b85f6322bf5dfde1eaf Eric Zhang 34133756+EZForever@users.noreply.github.com 2025-07-10T20:29:05+08:00 GitHub noreply@github.com 2025-07-10T15:29:05+03:00 cmake : do not search for curl libraries by ourselves (#14613)
704bb7a71c01dc07c1478b85f6322bf5dfde1eaf 435a6d10d618a015060e45a38c7e9f27f4243316 Akarshan Biswas akarshan@menlo.ai 2025-07-10T13:59:38+05:30 GitHub noreply@github.com 2025-07-10T09:29:38+01:00 SYCL: Initial set_rows kernel implementation (#14562)
435a6d10d618a015060e45a38c7e9f27f4243316 f9a867f5921a85f3fa64d7b067f4c8ffc5f62eb4 Xuan-Son Nguyen son@huggingface.co 2025-07-10T09:00:20+02:00 GitHub noreply@github.com 2025-07-10T10:00:20+03:00 llama : minor coding style fix for smollm3 (#14605)
f9a867f5921a85f3fa64d7b067f4c8ffc5f62eb4 ac44eb6c808bd5d677261ce86edd8c43ec54cf2c Eric Zhang 34133756+EZForever@users.noreply.github.com 2025-07-10T13:19:37+08:00 GitHub noreply@github.com 2025-07-10T08:19:37+03:00 cmake : bump llguidance version to v1.0.1 (#14609)
ac44eb6c808bd5d677261ce86edd8c43ec54cf2c a57d1bcb3c0165ac87b1f0dbb429839b0da69689 Eric Zhang 34133756+EZForever@users.noreply.github.com 2025-07-10T13:19:13+08:00 GitHub noreply@github.com 2025-07-10T08:19:13+03:00 cmake : llguidance build parser library only (#14608)
a57d1bcb3c0165ac87b1f0dbb429839b0da69689 cb9178f885d1986cc0b12feb26ff426bc8a3556c compilade git@compilade.net 2025-07-09T23:54:38-04:00 GitHub noreply@github.com 2025-07-09T23:54:38-04:00 cuda : support Falcon-H1 state size for SSM_SCAN (#14602)
cb9178f885d1986cc0b12feb26ff426bc8a3556c 4a5686da22057867c23bd4a6be941ddc8c51e585 Xuan-Son Nguyen son@huggingface.co 2025-07-09T23:09:28+02:00 GitHub noreply@github.com 2025-07-09T23:09:28+02:00 llama : remove llm_graph_input_one (#14603)
4a5686da22057867c23bd4a6be941ddc8c51e585 98bab638fb28cf95a5a66dd2d51b40d6c8f6d69a compilade git@compilade.net 2025-07-09T14:59:57-04:00 GitHub noreply@github.com 2025-07-09T14:59:57-04:00 llama : support Jamba hybrid Transformer-Mamba models (#7531)
98bab638fb28cf95a5a66dd2d51b40d6c8f6d69a 26a48ad699d50b6268900062661bd22f3e792579 Xuan-Son Nguyen son@huggingface.co 2025-07-09T18:16:12+02:00 GitHub noreply@github.com 2025-07-09T18:16:12+02:00 ggml : add ggml_scale_bias (#14417)
26a48ad699d50b6268900062661bd22f3e792579 ffd59e7d18a76459d5c31ba97073c7c9d73cb752 Miaoqian Lin linmq006@gmail.com 2025-07-09T20:33:53+08:00 GitHub noreply@github.com 2025-07-09T14:33:53+02:00 ggml : prevent integer overflow in gguf tensor size calculation (#14595)
ffd59e7d18a76459d5c31ba97073c7c9d73cb752 105554595f9a7bf3e02232ed7798201d47c2a4a2 Dowon ks2515@naver.com 2025-07-09T17:22:31+09:00 GitHub noreply@github.com 2025-07-09T11:22:31+03:00 model : add skt/A.X-4.0 model vocabulary (#14589)
105554595f9a7bf3e02232ed7798201d47c2a4a2 04655063c47af6cbced295c8c7ad369402b15300 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-09T10:19:50+02:00 GitHub noreply@github.com 2025-07-09T10:19:50+02:00 llama : remove unintended whitespace (#14592)
04655063c47af6cbced295c8c7ad369402b15300 20b7bf8a32259ad9189a4797fd3e3a859c537b99 ibrahim khadraoui 132432132+ibrahimkhadraoui@users.noreply.github.com 2025-07-09T12:03:49+04:00 GitHub noreply@github.com 2025-07-09T10:03:49+02:00 model : add support for Falcon-H1 family (#14534)
20b7bf8a32259ad9189a4797fd3e3a859c537b99 6efcd65945a98cf6883cdd9de4c8ccd8c79d219a Xuan-Son Nguyen son@huggingface.co 2025-07-09T08:26:13+02:00 GitHub noreply@github.com 2025-07-09T09:26:13+03:00 convert : fix smollm3 jinja template (#14586)
6efcd65945a98cf6883cdd9de4c8ccd8c79d219a 699f4392a33f57c3352cf8d60bdc53db7ca235e7 Jeff Bolz jbolz@nvidia.com 2025-07-08T13:11:42-05:00 GitHub noreply@github.com 2025-07-08T20:11:42+02:00 vulkan: optimize flash attention split_k_reduce (#14554)
699f4392a33f57c3352cf8d60bdc53db7ca235e7 08382869a2d6dca5d84710f2f82cc17a9696585a stevenkuang stevenkuang@tencent.com 2025-07-09T00:29:29+08:00 GitHub noreply@github.com 2025-07-08T18:29:29+02:00 model : fix hunyuan moe chat template (#14584)
08382869a2d6dca5d84710f2f82cc17a9696585a bb4f7a9e4eec171fecf0f640b1337a1c24485560 Xuan-Son Nguyen son@huggingface.co 2025-07-08T18:07:01+02:00 GitHub noreply@github.com 2025-07-08T18:07:01+02:00 model : add SmolLM3 (#14581)
bb4f7a9e4eec171fecf0f640b1337a1c24485560 b8eeb8741d4483daf576498cf90537b4de71633c compilade git@compilade.net 2025-07-08T11:37:47-04:00 GitHub noreply@github.com 2025-07-08T18:37:47+03:00 memory : fix broken batch splits for recurrent cache (#14575)
b8eeb8741d4483daf576498cf90537b4de71633c 17a1f0d2d407040ee242e18dd79be8bb212cfcef Jeff Bolz jbolz@nvidia.com 2025-07-08T08:21:21-05:00 GitHub noreply@github.com 2025-07-08T15:21:21+02:00 vulkan : fix rope with partial rotation and non-cont src (#14582)
17a1f0d2d407040ee242e18dd79be8bb212cfcef 8f22dc0a53338c629c1ef8fa878d8e39bfe627c9 Alawode Oluwandabira dabiraalawode@yahoo.com 2025-07-08T11:47:33+03:00 GitHub noreply@github.com 2025-07-08T11:47:33+03:00 server: Add ability to mount server at prefix (#14544)
8f22dc0a53338c629c1ef8fa878d8e39bfe627c9 53903ae6fa5f1caf187889c839cdd1ad25da4018 Xuan-Son Nguyen son@huggingface.co 2025-07-08T10:24:06+02:00 GitHub noreply@github.com 2025-07-08T11:24:06+03:00 model : add hunyuan moe (#14425)
53903ae6fa5f1caf187889c839cdd1ad25da4018 4d0dcd4a06080e796e6742a88f2ffa7fc41b28b8 Jeff Bolz jbolz@nvidia.com 2025-07-08T02:38:31-05:00 GitHub noreply@github.com 2025-07-08T09:38:31+02:00 vulkan: increase timeout for CI (#14574)
4d0dcd4a06080e796e6742a88f2ffa7fc41b28b8 75c91de6e955d5b8f3f28173f5040593e1964eb3 Georgi Gerganov ggerganov@gmail.com 2025-07-08T10:15:21+03:00 GitHub noreply@github.com 2025-07-08T10:15:21+03:00 cuda : fix rope with partial rotation and non-cont src (#14580)
75c91de6e955d5b8f3f28173f5040593e1964eb3 68155c66f0e76680f34442247e589a090add22d3 Aman Gupta amangupta052@gmail.com 2025-07-08T10:11:18+08:00 GitHub noreply@github.com 2025-07-08T10:11:18+08:00 CUDA: add bilinear interpolation for upscale (#14563)
68155c66f0e76680f34442247e589a090add22d3 e1a7059053a9b8958f2d57a21fc46dbc7fb24f8e R0CKSTAR yeahdongcn@gmail.com 2025-07-08T07:58:30+08:00 GitHub noreply@github.com 2025-07-08T07:58:30+08:00 musa: fix build warnings (unused variable) (#14561)
e1a7059053a9b8958f2d57a21fc46dbc7fb24f8e 12f55c302b35cfe900b84c5fe67c262026af9c44 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-07T23:35:35+02:00 GitHub noreply@github.com 2025-07-07T23:35:35+02:00 llama : fix incorrect minicpm3 v_states shape (#14571)
12f55c302b35cfe900b84c5fe67c262026af9c44 b9c3eefde1b67104bd993485ff38dd62abe9d70c Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-07T21:35:08+02:00 GitHub noreply@github.com 2025-07-07T21:35:08+02:00 llama : remove ggml_cont where possible (#14568)
b9c3eefde1b67104bd993485ff38dd62abe9d70c 6491d6e4f1caf0ad2221865b4249ae6938a6308c Aman Gupta amangupta052@gmail.com 2025-07-07T21:45:43+08:00 GitHub noreply@github.com 2025-07-07T21:45:43+08:00 CUDA: add bf16 and i32 to getrows (#14529)
6491d6e4f1caf0ad2221865b4249ae6938a6308c e592be15756ae546ba87bb5a5250b71248121971 Eve 139727413+netrunnereve@users.noreply.github.com 2025-07-06T10:29:36Z GitHub noreply@github.com 2025-07-06T12:29:36+02:00 vulkan: increase LOAD_VEC_A to 8 (IQ1/IQ2) or 4 (IQ3) (#14485)
e592be15756ae546ba87bb5a5250b71248121971 a0374a67e2924f2e845cdc59dd67d9a44065a89c Jeff Bolz jbolz@nvidia.com 2025-07-06T03:08:16-05:00 GitHub noreply@github.com 2025-07-06T10:08:16+02:00 vulkan: fix rms_norm+mul fusion (#14545)
a0374a67e2924f2e845cdc59dd67d9a44065a89c ddef99522d1ba74193b7394e803fab8db5c78bae Jeff Bolz jbolz@nvidia.com 2025-07-05T02:26:04-05:00 GitHub noreply@github.com 2025-07-05T09:26:04+02:00 vulkan: Handle updated FA dim2/3 definition (#14518)
ddef99522d1ba74193b7394e803fab8db5c78bae 668168814601d2aef6161ce49ab186bc74177ae7 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-05T09:17:14+02:00 GitHub noreply@github.com 2025-07-05T09:17:14+02:00 server : fix assistant prefilling when content is an array (#14360)
668168814601d2aef6161ce49ab186bc74177ae7 bac8bed248d15419137c5bc7f834582397baaebc Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-05T08:24:56+02:00 GitHub noreply@github.com 2025-07-04T23:24:56-07:00 opencl: add GELU_ERF (#14476)
bac8bed248d15419137c5bc7f834582397baaebc b81510a7b78f7c5a6f069c4f3d0e569b9d287913 Georgi Gerganov ggerganov@gmail.com 2025-07-05T07:18:09+03:00 GitHub noreply@github.com 2025-07-05T07:18:09+03:00 eval-callback : check for empty input (#14539)
b81510a7b78f7c5a6f069c4f3d0e569b9d287913 ef797db357e44ecb7437fa9d22f4e1614104b342 R0CKSTAR yeahdongcn@gmail.com 2025-07-05T12:10:53+08:00 GitHub noreply@github.com 2025-07-05T12:10:53+08:00 test-backend-ops: add support for specifying output format (#14368)
ef797db357e44ecb7437fa9d22f4e1614104b342 67d1ef23c68e1e09dc6d7423d1ef5fc0ea56ed5e Georgi Gerganov ggerganov@gmail.com 2025-07-04T19:19:09+03:00 GitHub noreply@github.com 2025-07-04T19:19:09+03:00 metal : disable fast math in all quantize kernels (#14528)
67d1ef23c68e1e09dc6d7423d1ef5fc0ea56ed5e 7b50f7c0257695d0f6918bffce4e68d5c13b0c9e Georgi Gerganov ggerganov@gmail.com 2025-07-04T09:08:59+03:00 GitHub noreply@github.com 2025-07-04T09:08:59+03:00 batch : add optional for sequential equal split (#14511)
7b50f7c0257695d0f6918bffce4e68d5c13b0c9e c79184d2d192489e3c918bab8ed717d22f8c02bd Georgi Gerganov ggerganov@gmail.com 2025-07-04T09:05:36+03:00 GitHub noreply@github.com 2025-07-04T09:05:36+03:00 graph : prepare for 4D mask (#14515)
c79184d2d192489e3c918bab8ed717d22f8c02bd 499a8f5a787f5fdc7f3fdfb9d1ff01b6cb5e994e Georgi Gerganov ggerganov@gmail.com 2025-07-04T09:04:59+03:00 GitHub noreply@github.com 2025-07-04T09:04:59+03:00 batch : add n_used count (#14512)
499a8f5a787f5fdc7f3fdfb9d1ff01b6cb5e994e 28657a8229b5adc6028cf1c4ed62191792d2fdb0 luyhcsu 110711054+luyhcsu@users.noreply.github.com 2025-07-04T11:50:07+08:00 GitHub noreply@github.com 2025-07-04T11:50:07+08:00 CANN: Replace aclrtMemsetSync with aclnnInplaceZero operator (#14002)
28657a8229b5adc6028cf1c4ed62191792d2fdb0 bee28421be25fd447f61cb6db64d556cbfce32ec Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-03T23:07:22+02:00 GitHub noreply@github.com 2025-07-03T23:07:22+02:00 ggml : implement GEGLU_ERF and GEGLU_QUICK ops (#14445)
bee28421be25fd447f61cb6db64d556cbfce32ec 2b72bedec198a90bb5b0cceaf1d0aff9e34ffbc2 lhez quic_lih@quicinc.com 2025-07-03T11:22:24-07:00 GitHub noreply@github.com 2025-07-03T20:22:24+02:00 opencl : broadcast for soft_max (#14510)
2b72bedec198a90bb5b0cceaf1d0aff9e34ffbc2 c8c4495b8d3a8799e2d46778f993965b0ac1ae43 Jeff Bolz jbolz@nvidia.com 2025-07-03T13:21:14-05:00 GitHub noreply@github.com 2025-07-03T20:21:14+02:00 vulkan: support mixed/deepseekR1 FA head sizes (#14509)
c8c4495b8d3a8799e2d46778f993965b0ac1ae43 7b63a71a6b0f54effe9b94073d4d0519dcf53676 Johannes Gäßler johannesg@5d6.de 2025-07-03T17:05:18+02:00 GitHub noreply@github.com 2025-07-03T17:05:18+02:00 ggml: backward pass for split swiglu (#14483)
7b63a71a6b0f54effe9b94073d4d0519dcf53676 0c2ee38ab729f3f6a7c28dc4c5dcd8d5f0cbf8fc Nicolò Scipione nicolo.scipione@codeplay.com 2025-07-03T11:00:03+02:00 GitHub noreply@github.com 2025-07-03T11:00:03+02:00 Fix conditional enabling following arch checks for ggml-sycl (#14504)
0c2ee38ab729f3f6a7c28dc4c5dcd8d5f0cbf8fc a70c8a0c4b4c1606cd9a0ba889ce61aa88610095 Xuan-Son Nguyen son@huggingface.co 2025-07-03T10:03:06+02:00 GitHub noreply@github.com 2025-07-03T10:03:06+02:00 convert : correct gemma 3n conversion (#14450)
a70c8a0c4b4c1606cd9a0ba889ce61aa88610095 9067487c4411efb20400103fcccfdd389c80d428 Georgi Gerganov ggerganov@gmail.com 2025-07-03T10:53:35+03:00 GitHub noreply@github.com 2025-07-03T10:53:35+03:00 kv-cache : use ggml_set_rows (#14285)
9067487c4411efb20400103fcccfdd389c80d428 d4cdd9c1c3cebdafca735958597de4ff7b7c0f54 Georgi Gerganov ggerganov@gmail.com 2025-07-03T10:46:57+03:00 GitHub noreply@github.com 2025-07-03T10:46:57+03:00 ggml : fix FA mask dim 2 and 3 (#14505)
d4cdd9c1c3cebdafca735958597de4ff7b7c0f54 55c2646b452419be7552157b9b11373ccc1bf2f2 Georgi Gerganov ggerganov@gmail.com 2025-07-03T07:48:32+03:00 GitHub noreply@github.com 2025-07-03T07:48:32+03:00 ggml : remove kompute backend (#14501)
55c2646b452419be7552157b9b11373ccc1bf2f2 e75ba4c0434eb759eb7ff74e034ebe729053e575 Aman Gupta amangupta052@gmail.com 2025-07-03T07:45:11+08:00 GitHub noreply@github.com 2025-07-03T07:45:11+08:00 CUDA: add dynamic shared mem to softmax, refactor general usage (#14497)
e75ba4c0434eb759eb7ff74e034ebe729053e575 5d46babdc2d4675d96ebcf23cac098a02f0d30cc Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-02T21:02:35+02:00 GitHub noreply@github.com 2025-07-02T21:02:35+02:00 gguf-py : add support for chat template jinja files (#14508)
5d46babdc2d4675d96ebcf23cac098a02f0d30cc e17991c466ac835b2c71bd813c1ca7ff8dd97b94 compilade git@compilade.net 2025-07-02T13:10:24-04:00 GitHub noreply@github.com 2025-07-02T13:10:24-04:00 llama : initial Mamba-2 support (#9126)
e17991c466ac835b2c71bd813c1ca7ff8dd97b94 c46944aa25b5560e6195f8cbcbc8947e9a6395c3 Georgi Gerganov ggerganov@gmail.com 2025-07-02T19:35:47+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-02T20:08:45+03:00 sync : ggml
c46944aa25b5560e6195f8cbcbc8947e9a6395c3 f3ed38d793fab9f97987dc52d7a41622f2056701 Daniel Bevenius daniel.bevenius@gmail.com 2025-07-02T13:55:32+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-02T20:08:45+03:00 ggml : add version function to get lib version (ggml/1286)
f3ed38d793fab9f97987dc52d7a41622f2056701 55a1c5a5fdefef808e95aabd3d5563af1068cc80 Rotem Dan rotemdan@gmail.com 2025-07-02T19:37:16+03:00 GitHub noreply@github.com 2025-07-02T18:37:16+02:00 Set RPATH to "@loader_path" / "$ORIGIN" to ensure executables and dynamic libraries search for dependencies in their origin directory. (#14309)
55a1c5a5fdefef808e95aabd3d5563af1068cc80 12a81af45f0dbbab24bd819a15f57c03ceb1be90 Aman Gupta amangupta052@gmail.com 2025-07-02T20:34:24+08:00 Georgi Gerganov ggerganov@gmail.com 2025-07-02T15:48:33+03:00 CUDA: add softmax broadcast (#14475)
12a81af45f0dbbab24bd819a15f57c03ceb1be90 8875523eb311cac832bfda0c581e852292185ae9 Johannes Gäßler johannesg@5d6.de 2025-07-02T13:42:12+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-02T15:48:33+03:00 CUDA: broadcasting for FlashAttention mask (#14500)
8875523eb311cac832bfda0c581e852292185ae9 ec68e84c32325a3417fbcd2e60d4bda6adb4e4bc Jeff Bolz jbolz@nvidia.com 2025-07-01T03:32:56-05:00 Georgi Gerganov ggerganov@gmail.com 2025-07-02T15:48:33+03:00 vulkan: support softmax/FA batch and broadcast (#14449)
ec68e84c32325a3417fbcd2e60d4bda6adb4e4bc 307e79d33d4cdd9f1d6c42fc861724e6ba12b98f Georgi Gerganov ggerganov@gmail.com 2025-06-27T21:50:57+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-02T15:48:33+03:00 ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (#14435)
307e79d33d4cdd9f1d6c42fc861724e6ba12b98f d7f5f4e578d1f60b0835d1734a50438c309b3e5c zhouwg zhouwg2000@gmail.com 2025-07-02T20:38:10+08:00 GitHub noreply@github.com 2025-07-02T14:38:10+02:00 opencl : fix possible buffer overflow in dump_tensor (#14490)
d7f5f4e578d1f60b0835d1734a50438c309b3e5c c8a4e470f65c3a932e18eddc5fba1844876d7463 Georgi Gerganov ggerganov@gmail.com 2025-07-02T14:12:07+03:00 GitHub noreply@github.com 2025-07-02T14:12:07+03:00 simple-chat : fix context-exceeded condition (#14494)
c8a4e470f65c3a932e18eddc5fba1844876d7463 603e43dc913f90d9c3291624728b731687eddc35 Eric Zhang 34133756+EZForever@users.noreply.github.com 2025-07-02T19:00:04+08:00 GitHub noreply@github.com 2025-07-02T13:00:04+02:00 opencl : skip empty nodes on cgraph compute (#14491)
603e43dc913f90d9c3291624728b731687eddc35 611ba4b264c8fbb9d2d978bd6a5c17aeab1700fb lhez quic_lih@quicinc.com 2025-07-02T00:07:42-07:00 GitHub noreply@github.com 2025-07-02T09:07:42+02:00 opencl : update upscale to support align corners (#14488)
611ba4b264c8fbb9d2d978bd6a5c17aeab1700fb 85841e121db5b96ae4d277a3cd3995eef66b98ec Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-02T09:02:51+02:00 GitHub noreply@github.com 2025-07-02T09:02:51+02:00 ci : add OpenCL to labeler workflow (#14496)
85841e121db5b96ae4d277a3cd3995eef66b98ec 68b3cd65141586ef13a698baa9029627f038f102 Eric Zhang 34133756+EZForever@users.noreply.github.com 2025-07-02T13:41:35+08:00 GitHub noreply@github.com 2025-07-02T08:41:35+03:00 github : add OpenCL backend to issue templates (#14492)
68b3cd65141586ef13a698baa9029627f038f102 de569441470332ff922c23fb0413cc957be75b25 Björn Ganster mail@bjoern-ganster.de 2025-07-02T07:19:31+02:00 GitHub noreply@github.com 2025-07-02T08:19:31+03:00 ggml : Callback before abort (#14481)
de569441470332ff922c23fb0413cc957be75b25 1b2aaf28acd632b79bc3b07acb5dad877dc7dbb1 Georgi Gerganov ggerganov@gmail.com 2025-07-01T18:04:08+03:00 GitHub noreply@github.com 2025-07-01T18:04:08+03:00 ci : disable fast-math for Metal GHA CI (#14478)
1b2aaf28acd632b79bc3b07acb5dad877dc7dbb1 343b6e94b61674ac94e64ede7b7ea3365793f7ed Grzegorz Grasza xek@redhat.com 2025-07-01T15:44:11+02:00 GitHub noreply@github.com 2025-07-01T15:44:11+02:00 Add Vulkan images to docker.md (#14472)
343b6e94b61674ac94e64ede7b7ea3365793f7ed 6a746cf9c40f8d93d13eb8a6c2b989a15e7fa61a Chenguang Li 757486878@qq.com 2025-07-01T16:47:30+08:00 GitHub noreply@github.com 2025-07-01T16:47:30+08:00 CANN: update aclnnGroupedMatmulV2 to aclnnGroupedMatmulV3 (#14411)
6a746cf9c40f8d93d13eb8a6c2b989a15e7fa61a eff5e45443a248f89cc61e64786f38b68b4da489 Jeff Bolz jbolz@nvidia.com 2025-07-01T03:43:08-05:00 GitHub noreply@github.com 2025-07-01T10:43:08+02:00 vulkan: Split large mul_mat_id to fit in shared memory (#14451)
eff5e45443a248f89cc61e64786f38b68b4da489 a6a47958a1d9edc628cedc2f9320c84b59fe1f4f Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-07-01T10:14:21+02:00 GitHub noreply@github.com 2025-07-01T10:14:21+02:00 add GELU_ERF (#14455)
a6a47958a1d9edc628cedc2f9320c84b59fe1f4f f61c05d4b1f8ad498a5cf7f0f57a40383aad563c Georgi Gerganov ggerganov@gmail.com 2025-07-01T11:05:48+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-01T11:06:39+03:00 ggml : remove trailing whitespace (#0)
f61c05d4b1f8ad498a5cf7f0f57a40383aad563c 431b2c24f31bf5411786c378b409d0202df8d53c Georgi Gerganov ggerganov@gmail.com 2025-07-01T10:27:52+03:00 Georgi Gerganov ggerganov@gmail.com 2025-07-01T11:06:39+03:00 sync : ggml
431b2c24f31bf5411786c378b409d0202df8d53c 497be7c01dab243100f9c42e0a763a746e42d038 Acly aclysia@gmail.com 2025-07-01T09:11:00+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-01T11:06:39+03:00 ggml-cpu : "align corners" for bilinear upscale/downscale (ggml/1285)
497be7c01dab243100f9c42e0a763a746e42d038 79b33b231774d5c39c8df018e9a276becae6d41a Daniel Bevenius daniel.bevenius@gmail.com 2025-06-24T06:10:16+02:00 Georgi Gerganov ggerganov@gmail.com 2025-07-01T11:06:39+03:00 ggml-quants : rename best_mad to best_error (ggml/1283)
79b33b231774d5c39c8df018e9a276becae6d41a 0a5a3b5cdfd887cf0f8e09d9ff89dee130cfcdde lhez quic_lih@quicinc.com 2025-07-01T00:19:16-07:00 GitHub noreply@github.com 2025-07-01T09:19:16+02:00 opencl : add GEGLU, REGLU, SWIGLU (#14456)
0a5a3b5cdfd887cf0f8e09d9ff89dee130cfcdde 745f11fed09ba2303f3f494efb12286d382608e5 Aman Gupta amangupta052@gmail.com 2025-06-30T23:57:04+08:00 GitHub noreply@github.com 2025-06-30T23:57:04+08:00 Add Conv2d for CPU (#14388)
745f11fed09ba2303f3f494efb12286d382608e5 5dd942de5922a22ec8446a4ad2203738dbcb9389 Georgi Gerganov ggerganov@gmail.com 2025-06-30T18:03:03+03:00 GitHub noreply@github.com 2025-06-30T18:03:03+03:00 memory : correctly handle failure in apply() (#14438)
5dd942de5922a22ec8446a4ad2203738dbcb9389 a7417f55945eb7c7a5ea6807e66564b8066a4e50 Georgi Gerganov ggerganov@gmail.com 2025-06-30T17:04:05+03:00 GitHub noreply@github.com 2025-06-30T17:04:05+03:00 metal : disable fast-math for some cpy kernels (#14460)
a7417f55945eb7c7a5ea6807e66564b8066a4e50 eb3fa2913e0ace38912011fcb331624c679656f5 Romain Biessy romain.biessy@codeplay.com 2025-06-30T14:52:02+02:00 GitHub noreply@github.com 2025-06-30T14:52:02+02:00 ggml-cpu: sycl: Re-enable exp f16 (#14462)
eb3fa2913e0ace38912011fcb331624c679656f5 c839a2da1a0d4275c3a98f70c3a7627487573a46 Diego Devesa slarengh@gmail.com 2025-06-30T03:43:15-07:00 GitHub noreply@github.com 2025-06-30T12:43:15+02:00 test-backend-ops : disable llama test (#14461)
c839a2da1a0d4275c3a98f70c3a7627487573a46 e9b6350e61d592634263a14b3d77ecbf6c1fb096 xiaobing318 71554036+xiaobing318@users.noreply.github.com 2025-06-30T17:48:24+08:00 GitHub noreply@github.com 2025-06-30T12:48:24+03:00 cmake : Remove redundant include path in CMakeLists.txt (#14452)
e9b6350e61d592634263a14b3d77ecbf6c1fb096 caf5681fcb47dfe9bafee94ef9aa8f669ac986c7 Vedran Miletić vedran@miletic.net 2025-06-30T10:17:18+02:00 GitHub noreply@github.com 2025-06-30T10:17:18+02:00 scripts : make the shell scripts cross-platform (#14341)
caf5681fcb47dfe9bafee94ef9aa8f669ac986c7 83790b0e7e09ab17238b16452a33053a71dbdfad matteo matteo.serva@gmail.com 2025-06-29T20:02:53+02:00 GitHub noreply@github.com 2025-06-29T20:02:53+02:00 server : support jinja extra template kwargs (Qwen3 enable_thinking feature), from command line and from client (#13196)
83790b0e7e09ab17238b16452a33053a71dbdfad f47c1d7106e49062279bcc57fc1077c0db61e278 Renat rntk@users.noreply.github.com 2025-06-29T19:29:57+02:00 GitHub noreply@github.com 2025-06-29T19:29:57+02:00 server : fix appearance of the chats list context menu for Safari (#14322)
f47c1d7106e49062279bcc57fc1077c0db61e278 a5d1fb6212298db1be1639db4c03adb2c522ee13 Akarshan Biswas akarshan@menlo.ai 2025-06-29T21:07:58+05:30 GitHub noreply@github.com 2025-06-29T21:07:58+05:30 SYCL: disable faulty fp16 exp kernel (#14395)
a5d1fb6212298db1be1639db4c03adb2c522ee13 a0535ffa0d35fccfec3e1a0a3bfc9dbb6054d7c0 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-29T14:38:10+02:00 GitHub noreply@github.com 2025-06-29T14:38:10+02:00 ggml : fix unmerged GGML_FPxx_TO_FPxx refactoring (#14443)
a0535ffa0d35fccfec3e1a0a3bfc9dbb6054d7c0 bd9c981d7226107f18deb8344c3301450311bb8b Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-29T11:04:10+02:00 GitHub noreply@github.com 2025-06-29T11:04:10+02:00 ggml : implement REGLU/GEGLU/SWIGLU ops (#14158)
bd9c981d7226107f18deb8344c3301450311bb8b 27208bf657cfe7262791df473927225e48efe482 Jeff Bolz jbolz@nvidia.com 2025-06-29T02:43:36-05:00 GitHub noreply@github.com 2025-06-29T09:43:36+02:00 vulkan: Add fusion support for RMS_NORM+MUL (#14366)
27208bf657cfe7262791df473927225e48efe482 63a7bb3c7e1c6b0a92d03b0a594d3cd501d6ed3e Aman Gupta amangupta052@gmail.com 2025-06-29T01:30:53+08:00 GitHub noreply@github.com 2025-06-29T01:30:53+08:00 CUDA: add bf16 and f32 support to cublas_mul_mat_batched (#14361)
63a7bb3c7e1c6b0a92d03b0a594d3cd501d6ed3e 00d5282c7f2a0bb05bb315fb81ca3b0f42cf9f07 Jeff Bolz jbolz@nvidia.com 2025-06-28T10:36:40-05:00 GitHub noreply@github.com 2025-06-28T17:36:40+02:00 vulkan: handle noncontig in the final case of ggml_vk_get_cpy_pipeline (#14378)
00d5282c7f2a0bb05bb315fb81ca3b0f42cf9f07 566c16fcce44876a167c37f159085afe6f84b28c Jeff Bolz jbolz@nvidia.com 2025-06-28T10:17:09-05:00 GitHub noreply@github.com 2025-06-28T17:17:09+02:00 vulkan: lock accesses of pinned_memory vector (#14333)
566c16fcce44876a167c37f159085afe6f84b28c b25e92774e2fa4ee3820e458d5cf43f40190f8d2 Weizhao Ouyang weizhao.ouyang@arm.com 2025-06-28T22:08:21+08:00 GitHub noreply@github.com 2025-06-28T16:08:21+02:00 model : add support for ERNIE 4.5 0.3B model (#14408)
b25e92774e2fa4ee3820e458d5cf43f40190f8d2 6609507a910aa7437aaa53fd999447de3947d998 Xinpeng Dou 15529241576@163.com 2025-06-28T17:35:41+08:00 GitHub noreply@github.com 2025-06-28T17:35:41+08:00 fix async_mode bug (#14432)
6609507a910aa7437aaa53fd999447de3947d998 ceb1bf5a34d5e66e28b23dcc7a3cd83fe1e27481 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-28T09:57:07+02:00 GitHub noreply@github.com 2025-06-28T09:57:07+02:00 ci : fix windows build and release (#14431)
ceb1bf5a34d5e66e28b23dcc7a3cd83fe1e27481 72babea5dea56c8a8e8420ccf731b12a5cf37854 Jeff Bolz jbolz@nvidia.com 2025-06-27T22:35:30-05:00 GitHub noreply@github.com 2025-06-27T22:35:30-05:00 vulkan: Fix GGML_VULKAN_SHADER_DEBUG_INFO (#14427)
72babea5dea56c8a8e8420ccf731b12a5cf37854 43678060c1f4cfab2f899466fd615e358677f807 Georgi Gerganov ggerganov@gmail.com 2025-06-27T21:42:02+03:00 GitHub noreply@github.com 2025-06-27T21:42:02+03:00 graph : make llm_graph_context destructor virtual (#14410)
43678060c1f4cfab2f899466fd615e358677f807 8d94219a4a7f2da72ee542019ca01f36af93d1d6 Georgi Gerganov ggerganov@gmail.com 2025-06-27T17:55:45+03:00 GitHub noreply@github.com 2025-06-27T17:55:45+03:00 recurrent : call balloc split_reset() in init_batch() (#14414)
8d94219a4a7f2da72ee542019ca01f36af93d1d6 f667f1e6244e1f420512fa66692b7096ff17f366 Radoslav Gerganov rgerganov@gmail.com 2025-06-27T16:41:40+03:00 GitHub noreply@github.com 2025-06-27T16:41:40+03:00 ggml : add ggml_set_rows (#14274)
f667f1e6244e1f420512fa66692b7096ff17f366 8846aace4934ad29651ea61b8c7e3f6b0556e3d2 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-27T10:42:19+02:00 GitHub noreply@github.com 2025-06-27T10:42:19+02:00 convert : fix broken sentencepiece vocab (#14416)
8846aace4934ad29651ea61b8c7e3f6b0556e3d2 a01047b041aa04aeea351933658433ed004516ab Xuan-Son Nguyen son@huggingface.co 2025-06-26T19:34:02+02:00 GitHub noreply@github.com 2025-06-26T20:34:02+03:00 model : gemma3n text-only (#14400)
a01047b041aa04aeea351933658433ed004516ab b25346221dadb9101aa9dda55431dde4d3596943 bandoti 141645996+bandoti@users.noreply.github.com 2025-06-26T13:46:53-03:00 GitHub noreply@github.com 2025-06-26T13:46:53-03:00 cmake: regen vulkan shaders when shaders-gen sources change (#14398)
b25346221dadb9101aa9dda55431dde4d3596943 e8215dbb96b8fb94a24c29cdd228166fb972dbfc Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-26T15:01:14+02:00 GitHub noreply@github.com 2025-06-26T15:01:14+02:00 llama : return mistral-v7-tekken as default template only (#14390)
e8215dbb96b8fb94a24c29cdd228166fb972dbfc 5783ae43599400b723b5da0569c1f848419ff3c7 Georgi Gerganov ggerganov@gmail.com 2025-06-26T15:51:19+03:00 GitHub noreply@github.com 2025-06-26T15:51:19+03:00 metal : add special-case mat-vec mul for ne00 == 4 (#14385)
5783ae43599400b723b5da0569c1f848419ff3c7 bf5bcd0b857db420235e03639f0a5f218a7f8cf8 Georgi Gerganov ggerganov@gmail.com 2025-06-26T15:50:15+03:00 GitHub noreply@github.com 2025-06-26T15:50:15+03:00 metal : batch rows copy in a single threadgroup (#14384)
bf5bcd0b857db420235e03639f0a5f218a7f8cf8 716301d1b03c31875ec3b24526c48c8b1bd0fd8c Aaron Teo aaron.teo1@ibm.com 2025-06-26T18:41:41+08:00 GitHub noreply@github.com 2025-06-26T12:41:41+02:00 docs: update s390x documentation + add faq (#14389)
ba67d6c03fed3193987a4ef13050e6e2a4451df4 8f52c6f8abb2ed9ad967cc0bfb90acb073011193 Yunzhe Jia yunzhe@calculet.tech 2025-06-26T17:42:32+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-26T17:42:32+08:00 llama-calrt: infer-op: copy data from output buffer to dst-tensor.data
8f52c6f8abb2ed9ad967cc0bfb90acb073011193 d2f325130c4ed77a3dcff2c23a926587fb08e2cc Yunzhe Jia yunzhe@calculet.tech 2025-06-26T15:10:45+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-26T15:10:45+08:00 Squashed commit of the following:
716301d1b03c31875ec3b24526c48c8b1bd0fd8c 60ef23d6c14d325d83eae5752e5de39ad268e9b0 R0CKSTAR yeahdongcn@gmail.com 2025-06-26T12:11:59+08:00 GitHub noreply@github.com 2025-06-26T12:11:59+08:00 musa: enable fp16 mma (all) and cublas on qy2 (#13842)
60ef23d6c14d325d83eae5752e5de39ad268e9b0 b193d5306912a2adae0fde7481819f6ee0941bc6 Aaron Teo aaron.teo1@ibm.com 2025-06-26T05:49:04+08:00 GitHub noreply@github.com 2025-06-25T23:49:04+02:00 ggml-cpu: enable IBM NNPA Vector Intrinsics (#14317)
b193d5306912a2adae0fde7481819f6ee0941bc6 2bf9d539dd158345e3a3b096e16474af535265b4 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-25T23:26:51+02:00 GitHub noreply@github.com 2025-06-25T23:26:51+02:00 ggml : do not output unprintable characters on GGUF load failure (#14381)
2bf9d539dd158345e3a3b096e16474af535265b4 73e53dc834c0a2336cd104473af6897197b96277 Anton Mitkov anton_b_mitkov@abv.bg 2025-06-25T17:09:55+01:00 GitHub noreply@github.com 2025-06-25T18:09:55+02:00 sycl: GGML_SYCL_DISABLE_OPT on by default for all Intel Devices (#13973)
73e53dc834c0a2336cd104473af6897197b96277 62af464227dafa1c55e0535bcb24346326748f46 lhez quic_lih@quicinc.com 2025-06-24T11:46:25-07:00 GitHub noreply@github.com 2025-06-24T11:46:25-07:00 opencl: ref count `ggml_backend_opencl_context` and refactor profiling (#14254)
62af464227dafa1c55e0535bcb24346326748f46 c148cf1946275952a79ad50b6199725f12a70411 Georgi Gerganov ggerganov@gmail.com 2025-06-24T18:26:30+03:00 GitHub noreply@github.com 2025-06-24T18:26:30+03:00 batch : fix check for empty sequences in memory (#14364)
c148cf1946275952a79ad50b6199725f12a70411 1b809cee225222094a0ff5be8467240487ce4ae4 Mathieu Baudier mbaudier@argeo.org 2025-06-24T15:05:31+02:00 GitHub noreply@github.com 2025-06-24T15:05:31+02:00 cmake : use LLAMA_BUILD_NUMBER when defining LLAMA_INSTALL_VERSION (#14362)
1b809cee225222094a0ff5be8467240487ce4ae4 abf241045d09cad70dc797b0fba393ad09ee2cbe Nigel Bosch pnigelb@gmail.com 2025-06-24T08:59:11Z GitHub noreply@github.com 2025-06-24T10:59:11+02:00 server : move no API key doc to /health (#14352)
abf241045d09cad70dc797b0fba393ad09ee2cbe 901e20bbe571fbde48d13eb188f4e7cdc7562fb6 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-24T09:31:00+02:00 GitHub noreply@github.com 2025-06-24T09:31:00+02:00 main : honor --verbose-prompt on interactive prompts (#14350)
901e20bbe571fbde48d13eb188f4e7cdc7562fb6 0142961a2e67909e33cdf410274b56c08c5dce7a Bartowski 3266127+bartowski1182@users.noreply.github.com 2025-06-24T02:17:58-04:00 GitHub noreply@github.com 2025-06-24T09:17:58+03:00 jinja : Add Mistral-Small-3.2-24B-Instruct-2506.jinja (#14349)
0142961a2e67909e33cdf410274b56c08c5dce7a ce82bd0117bd3598300b3a089d13d401b90279c7 uvos philipp@uvos.xyz 2025-06-24T01:12:56+02:00 GitHub noreply@github.com 2025-06-24T01:12:56+02:00 CUDA/HIP: optimize mmv paths taken for HIP devices (#14324)
ce82bd0117bd3598300b3a089d13d401b90279c7 bf2a99e3cb06a14dce2a586450d012dc31f922ae bandoti 141645996+bandoti@users.noreply.github.com 2025-06-23T15:30:51-03:00 GitHub noreply@github.com 2025-06-23T15:30:51-03:00 ci: add workflow for relocatable cmake package (#14346)
bf2a99e3cb06a14dce2a586450d012dc31f922ae 72c6bc3f3d0cf3bf160dddf1b803fed52bcbb0a3 Jeff Bolz jbolz@nvidia.com 2025-06-23T08:44:48-05:00 GitHub noreply@github.com 2025-06-23T15:44:48+02:00 vulkan: update windows SDK in release.yml (#14344)
72c6bc3f3d0cf3bf160dddf1b803fed52bcbb0a3 defe2158dd5e250b4ef53994057a4ec03131a263 Molly Sophia mollysophia379@gmail.com 2025-06-23T19:56:19+08:00 GitHub noreply@github.com 2025-06-23T19:56:19+08:00 llama : better rwkv chat template and add missing `inputs.use_jinja` setting (#14336)
defe2158dd5e250b4ef53994057a4ec03131a263 7b50d589a863c7631135c1226f6eab65cb406212 Johannes Gäßler johannesg@5d6.de 2025-06-23T13:11:31+02:00 GitHub noreply@github.com 2025-06-23T13:11:31+02:00 CUDA: mul_mat_v support for batch sizes > 1 (#14262)
7b50d589a863c7631135c1226f6eab65cb406212 3a9457df962b5883f2773f1c295e8c19df60d89f Georgi Gerganov ggerganov@gmail.com 2025-06-23T12:27:35+03:00 GitHub noreply@github.com 2025-06-23T12:27:35+03:00 kv-cells : fix tracking of seq_pos (#14339)
3a9457df962b5883f2773f1c295e8c19df60d89f fa4a9f2a1ccda2573189a9d4995bdf0bceb41156 Jeff Bolz jbolz@nvidia.com 2025-06-23T03:19:24-05:00 GitHub noreply@github.com 2025-06-23T10:19:24+02:00 vulkan: update windows SDK in CI (#14334)
fa4a9f2a1ccda2573189a9d4995bdf0bceb41156 238005c2dc67426cf678baa2d54c881701693288 Ed Addario 29247825+EAddario@users.noreply.github.com 2025-06-22T22:16:26+01:00 GitHub noreply@github.com 2025-06-22T23:16:26+02:00 quantize : handle user-defined pruning of whole layers (blocks) (#13037)
238005c2dc67426cf678baa2d54c881701693288 66aba7aca9a245d71f1ddf02c7a97223529752a8 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-22T19:46:17+02:00 GitHub noreply@github.com 2025-06-22T19:46:17+02:00 gguf-py : fix SpecialVocab parsing when post_processor is null (#14330)
66aba7aca9a245d71f1ddf02c7a97223529752a8 f1f5e82df6222dcbaca6396c0de44df259a1694f Ruikai Peng retr0@retr0.blog 2025-06-23T01:28:06+08:00 GitHub noreply@github.com 2025-06-23T01:28:06+08:00 run : avoid double tokenization (#14327)
f1f5e82df6222dcbaca6396c0de44df259a1694f af3373f1adfca56119f3e4de0e6a0a8df8edf3d9 Georgi Gerganov ggerganov@gmail.com 2025-06-22T20:10:07+03:00 GitHub noreply@github.com 2025-06-22T20:10:07+03:00 examples : fix is_first logic for tokenization (#14329)
af3373f1adfca56119f3e4de0e6a0a8df8edf3d9 5d5c066de8a3d2cb32f04c4d5ad1560945f30bf3 uvos philipp@uvos.xyz 2025-06-22T16:51:23+02:00 GitHub noreply@github.com 2025-06-22T16:51:23+02:00 HIP: enable vec fattn on RDNA4 (#14323)
5d5c066de8a3d2cb32f04c4d5ad1560945f30bf3 40bfa04c95c19fb42bafd4e21b5c2a7771846801 yuiseki yuiseki@gmail.com 2025-06-22T21:44:57+09:00 GitHub noreply@github.com 2025-06-22T14:44:57+02:00 mtmd : fix Pixtral OOM with large images by capping image_size to 1024 (#14326)
40bfa04c95c19fb42bafd4e21b5c2a7771846801 aa064b2eb7f7779db5e094a9d8d66d5033557de2 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-22T07:37:43+02:00 GitHub noreply@github.com 2025-06-22T08:37:43+03:00 common : use std::string_view now that we target c++17 (#14319)
aa064b2eb7f7779db5e094a9d8d66d5033557de2 aa0ef5c578eef4c2adc7be1282f21bab5f3e8d26 Aman Gupta amangupta052@gmail.com 2025-06-22T12:39:54+08:00 GitHub noreply@github.com 2025-06-22T12:39:54+08:00 CUDA: add mean operation (#14313)
aa0ef5c578eef4c2adc7be1282f21bab5f3e8d26 bb16041caef45cd4348cd6f84906b5dfec7a1f6a Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-21T18:12:05+02:00 GitHub noreply@github.com 2025-06-21T18:12:05+02:00 gguf-py : fix Qwen3-Embedding eos token (#14314)
bb16041caef45cd4348cd6f84906b5dfec7a1f6a 58cba76a9aab728717509d62b67c13afd8dc227a Markus Tavenrath mtavenrath@users.noreply.github.com 2025-06-21T08:17:12+02:00 GitHub noreply@github.com 2025-06-21T08:17:12+02:00 Add support for VK_EXT_debug_utils to add labels to Vulkan objects. (#13792)
58cba76a9aab728717509d62b67c13afd8dc227a 67ae5312e255ae97852a4a216e2245580bfafd72 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-21T07:33:21+02:00 GitHub noreply@github.com 2025-06-21T07:33:21+02:00 gguf-py : fix TemplateProcessing pair when bos/eos is missing (#14312)
67ae5312e255ae97852a4a216e2245580bfafd72 692e3cdd0a069ab56411b64506a67537d767683e Georgi Gerganov ggerganov@gmail.com 2025-06-21T08:04:18+03:00 GitHub noreply@github.com 2025-06-21T08:04:18+03:00 metal : fix thread-safety (#14300)
692e3cdd0a069ab56411b64506a67537d767683e b23fa0b3f40165ca3aae8ad4ee756e72f9a130dd Georgi Gerganov ggerganov@gmail.com 2025-06-21T08:03:46+03:00 GitHub noreply@github.com 2025-06-21T08:03:46+03:00 memory : rename interface to llama_memory_context_i (#14296)
b23fa0b3f40165ca3aae8ad4ee756e72f9a130dd 06cbedfca1587473df9b537f1dd4d6bfa2e3de13 Daniel Han danielhanchen@gmail.com 2025-06-20T21:32:01-07:00 GitHub noreply@github.com 2025-06-21T06:32:01+02:00 convert : fix Llama 4 conversion (#14311)
06cbedfca1587473df9b537f1dd4d6bfa2e3de13 b7147673f26c7ceb926a43c4734002bba291bcb8 Georgi Gerganov ggerganov@gmail.com 2025-06-20T20:50:24+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-20T21:02:47+03:00 sync : ggml
b7147673f26c7ceb926a43c4734002bba291bcb8 d860dd99a4178f58d1d1fa64eebc2aabc95392a7 Acly aclysia@gmail.com 2025-06-18T13:34:50+02:00 Georgi Gerganov ggerganov@gmail.com 2025-06-20T21:02:47+03:00 Add `ggml_roll` (ggml/1274)
d860dd99a4178f58d1d1fa64eebc2aabc95392a7 c959f462a0b4d42eaf930ffd72df0e435c97d5d5 David Chiu david20571015@gmail.com 2025-06-21T01:43:35+08:00 GitHub noreply@github.com 2025-06-20T19:43:35+02:00 docs : fix the link to llama.h (#14293)
c959f462a0b4d42eaf930ffd72df0e435c97d5d5 22015b2092e291022ea3cfedc6aeb8f2643807da Aman Gupta amangupta052@gmail.com 2025-06-20T22:48:24+08:00 GitHub noreply@github.com 2025-06-20T22:48:24+08:00 CUDA: add conv_2d_transpose (#14287)
22015b2092e291022ea3cfedc6aeb8f2643807da dd6e6d0b6a4bbe3ebfc931d1eb14db2f2b1d70af Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-20T16:37:44+02:00 GitHub noreply@github.com 2025-06-20T16:37:44+02:00 lint : remove trailing whitepace (#14304)
dd6e6d0b6a4bbe3ebfc931d1eb14db2f2b1d70af 8308f98c7fb778e54bf75538f5234d8bd20915e9 Ruikai Peng retr0@retr0.blog 2025-06-20T22:13:06+08:00 GitHub noreply@github.com 2025-06-20T07:13:06-07:00 vocab : prevent tokenizer overflow (#14301)
8308f98c7fb778e54bf75538f5234d8bd20915e9 6369be07359d03723f38c0a4a014ff0f698a0738 Nicolò Scipione nicolo.scipione@codeplay.com 2025-06-20T15:07:21+02:00 GitHub noreply@github.com 2025-06-20T15:07:21+02:00 sycl: add usage of enqueue_functions extension (#14244)
6369be07359d03723f38c0a4a014ff0f698a0738 88fc854b4bd2e3caf10e705e6afcbbca136f0a3c Christian Kastner ckk@kvr.at 2025-06-20T12:17:32Z GitHub noreply@github.com 2025-06-20T14:17:32+02:00 Implement GGML_CPU_ALL_VARIANTS for PowerPC (#14286)
88fc854b4bd2e3caf10e705e6afcbbca136f0a3c e28c1b93fd7d3f8faf9551d962e8a65fe2122e38 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-20T14:04:09+02:00 GitHub noreply@github.com 2025-06-20T14:04:09+02:00 llama : improve sep token handling (#14272)
e28c1b93fd7d3f8faf9551d962e8a65fe2122e38 d27b3ca1758dfb1718e333d497ef4b68ad109bc2 Diego Devesa slarengh@gmail.com 2025-06-20T04:57:36-07:00 GitHub noreply@github.com 2025-06-20T13:57:36+02:00 cuda : synchronize graph capture and cublas handle destruction (#14288)
d27b3ca1758dfb1718e333d497ef4b68ad109bc2 9230dbe2c757e2d5071329095727d0fa9d4b85c4 Georgi Gerganov ggerganov@gmail.com 2025-06-20T11:19:15+03:00 GitHub noreply@github.com 2025-06-20T11:19:15+03:00 ggml : fix repack work size for mul_mat_id (#14292)
9230dbe2c757e2d5071329095727d0fa9d4b85c4 812939a9e90f99d1bd5bb1bc6b99d12600671d50 Charles Xu charles.xu@arm.com 2025-06-20T09:51:01+02:00 GitHub noreply@github.com 2025-06-20T10:51:01+03:00 ggml: Update KleidiAI to v1.9.0 (#14277)
812939a9e90f99d1bd5bb1bc6b99d12600671d50 4c9fdfbe1580a66fd7d77c77418ce2c606a29fdd Georgi Gerganov ggerganov@gmail.com 2025-06-20T10:50:27+03:00 GitHub noreply@github.com 2025-06-20T10:50:27+03:00 model : more uniform output id handling (#14275)
4c9fdfbe1580a66fd7d77c77418ce2c606a29fdd 9eaa51e7f08593f123f00136591179a8f5956ecd Georgi Gerganov ggerganov@gmail.com 2025-06-20T10:14:14+03:00 GitHub noreply@github.com 2025-06-20T10:14:14+03:00 ubatch : new splitting logic (#14217)
d2f325130c4ed77a3dcff2c23a926587fb08e2cc 6391ee7ffd97e723dfbdcc32f001e1b95a63cba6 Yunzhe Jia yunzhe@calculet.tech 2025-06-17T11:53:33+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-20T09:57:18+08:00 calrt: add input_token in graph; input_token will be copy to calrt_in_buffer
9eaa51e7f08593f123f00136591179a8f5956ecd 8f71d0f3e86ccbba059350058af8758cafed73e6 Aman Gupta amangupta052@gmail.com 2025-06-20T09:50:24+08:00 GitHub noreply@github.com 2025-06-20T09:50:24+08:00 CUDA: add conv_2d_dw (#14265)
8f71d0f3e86ccbba059350058af8758cafed73e6 381174bbdaf10d6a80dc2099f284b20544d86962 Diego Devesa slarengh@gmail.com 2025-06-19T12:24:14-07:00 GitHub noreply@github.com 2025-06-19T21:24:14+02:00 ggml-cpu : remove unnecesary arm feature detection (#14281)
381174bbdaf10d6a80dc2099f284b20544d86962 d67341dc18fc5cc63362880ab2f8f9ecfc7932e7 Alex Trotta 44127594+Ahajha@users.noreply.github.com 2025-06-19T09:56:12-04:00 GitHub noreply@github.com 2025-06-19T15:56:12+02:00 gguf-py : make sentencepiece optional (#14200)
d67341dc18fc5cc63362880ab2f8f9ecfc7932e7 456af35eb70177b8dd5779b6d4c21bb020f9cebd aa956 aa956@users.noreply.github.com 2025-06-19T16:01:03+03:00 GitHub noreply@github.com 2025-06-19T16:01:03+03:00 server : add server parameters for draft model cache type (#13782)
456af35eb70177b8dd5779b6d4c21bb020f9cebd 600e3e9b50c1f0c9fc4a70356241fd87f00e8e14 fanyang fanyang89@outlook.com 2025-06-19T20:49:48+08:00 GitHub noreply@github.com 2025-06-19T14:49:48+02:00 build : suppress gcc15 compile warnings (#14261)
600e3e9b50c1f0c9fc4a70356241fd87f00e8e14 fffcce535ebbdc25f81966a15b758658788e7466 Anton Mitkov anton_b_mitkov@abv.bg 2025-06-19T11:40:21+01:00 GitHub noreply@github.com 2025-06-19T11:40:21+01:00 sycl: Cleanup codepaths in Get Rows in sycl backend (#14215)
fffcce535ebbdc25f81966a15b758658788e7466 5fc7856815920f828f9e90cb759ac82c7f0c1ea5 bashayer hijji bashayer.hijji@gmail.com 2025-06-19T13:24:12+03:00 GitHub noreply@github.com 2025-06-19T12:24:12+02:00 llama-bench : add --no-warmup flag (#14224) (#14270)
5fc7856815920f828f9e90cb759ac82c7f0c1ea5 faed5a5f5dde4d816adecdccb4f4682a04126f92 pqnet 119850+pqnet@users.noreply.github.com 2025-06-19T12:21:40+02:00 GitHub noreply@github.com 2025-06-19T12:21:40+02:00 convert : fix remote option in Windows (#14100)
faed5a5f5dde4d816adecdccb4f4682a04126f92 10bb545c5b54175ed9874ad8d187effa2bcb4b5f Aaron Teo aaron.teo1@ibm.com 2025-06-19T17:48:54+08:00 GitHub noreply@github.com 2025-06-19T11:48:54+02:00 llamafile : support s390x SIMD instruction set (#14273)
10bb545c5b54175ed9874ad8d187effa2bcb4b5f edc4a29effe716956fdfd2bc5b9cba2a6d8492f8 0cc4m picard12@live.de 2025-06-19T09:15:42+02:00 GitHub noreply@github.com 2025-06-19T09:15:42+02:00 Vulkan: Set device max size for host memory to avoid OOM warning and fallback to CPU buffer (#14249)
edc4a29effe716956fdfd2bc5b9cba2a6d8492f8 ed3290ab34493fd3d2e0f925f816a401da4f2dfd Gabe Goodhart gabe.l.hart@gmail.com 2025-06-19T00:08:14-05:00 GitHub noreply@github.com 2025-06-19T08:08:14+03:00 memory : Hybrid recurrent cache (#13979)
ed3290ab34493fd3d2e0f925f816a401da4f2dfd 8d947136546773f6410756f37fcc5d3e65b8135d Georgi Gerganov ggerganov@gmail.com 2025-06-19T08:05:21+03:00 GitHub noreply@github.com 2025-06-19T08:05:21+03:00 metal : add mean kernel (#14267)
8d947136546773f6410756f37fcc5d3e65b8135d 50d2227953ca9024f04255b4f116d06fcc0db74c Aaron Teo aaron.teo1@ibm.com 2025-06-19T01:10:26+08:00 GitHub noreply@github.com 2025-06-18T18:10:26+01:00 docs: add s390x build documentation (#14264)
50d2227953ca9024f04255b4f116d06fcc0db74c 6231c5cd6d49d61511f328b5f43322407df90a91 Aaron Teo aaron.teo1@ibm.com 2025-06-19T01:10:08+08:00 GitHub noreply@github.com 2025-06-18T18:10:08+01:00 ggml-cpu: reduce asm calls for hsum (#14037)
6231c5cd6d49d61511f328b5f43322407df90a91 ef035803eb9dbc306ea9a8ff82e30af12b567cf7 Aaron Teo aaron.teo1@ibm.com 2025-06-19T01:06:49+08:00 GitHub noreply@github.com 2025-06-18T18:06:49+01:00 ggml-cpu: fix uncaught underscore terminators (#14023)
ef035803eb9dbc306ea9a8ff82e30af12b567cf7 413977de32e90712ecec84d0b9c738847da8dc02 Charles Xu charles.xu@arm.com 2025-06-18T13:40:07+02:00 GitHub noreply@github.com 2025-06-18T12:40:07+01:00 ggml: Add Apple support for GGML_CPU_ALL_VARIANTS (#14258)
413977de32e90712ecec84d0b9c738847da8dc02 95402553a5effc61ddc9e29c7bcb56f71311dd4a Xuan-Son Nguyen son@huggingface.co 2025-06-18T10:43:57+02:00 GitHub noreply@github.com 2025-06-18T10:43:57+02:00 mtmd : refactor llava-uhd preprocessing logic (#14247)
95402553a5effc61ddc9e29c7bcb56f71311dd4a 3865cff4f5b84c16119590efd6ce537789a27715 Xuan-Son Nguyen son@huggingface.co 2025-06-18T09:58:43+02:00 GitHub noreply@github.com 2025-06-18T09:58:43+02:00 llama-chat : fix multiple system message for gemma, orion (#14246)
3865cff4f5b84c16119590efd6ce537789a27715 d03172cc797b6adbbc00bfc09caf614fb0f895a0 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-18T09:52:07+02:00 GitHub noreply@github.com 2025-06-18T09:52:07+02:00 convert : fix null head_dim AutoConfig regression (#14248)
d03172cc797b6adbbc00bfc09caf614fb0f895a0 dd8e59f4435342eda93c5b0cf4109e21c9c7d0eb Georgi Gerganov ggerganov@gmail.com 2025-06-18T09:58:23+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-18T09:59:21+03:00 sync : ggml
dd8e59f4435342eda93c5b0cf4109e21c9c7d0eb bbe98d27840453c8787d18470963530fdc27d89f Daniel Bevenius daniel.bevenius@gmail.com 2025-06-13T15:06:42+02:00 Georgi Gerganov ggerganov@gmail.com 2025-06-18T09:59:21+03:00 ggml : disable warnings for tests when using MSVC (ggml/1273)
bbe98d27840453c8787d18470963530fdc27d89f c2056ed6d461e6d5432f04f221e221ab795dc652 Daniel Bevenius daniel.bevenius@gmail.com 2025-06-13T09:05:44+02:00 Georgi Gerganov ggerganov@gmail.com 2025-06-18T09:59:21+03:00 ggml : remove unused ggml_context_container (ggml/1272)
c2056ed6d461e6d5432f04f221e221ab795dc652 c46503014db0d63fa7b1b28c58adfb51054e2dec Daniel Bevenius daniel.bevenius@gmail.com 2025-06-12T12:27:09+02:00 Georgi Gerganov ggerganov@gmail.com 2025-06-18T09:59:21+03:00 examples : include examples in msvc disable warn (ggml/1270)
c46503014db0d63fa7b1b28c58adfb51054e2dec 860a9e4eeff3eb2e7bd1cc38f65787cc6c8177af bandoti 141645996+bandoti@users.noreply.github.com 2025-06-17T17:33:25-03:00 GitHub noreply@github.com 2025-06-17T22:33:25+02:00 cmake: remove shader-gen step-targets from ggml-vulkan (#14226)
860a9e4eeff3eb2e7bd1cc38f65787cc6c8177af fe9d60e74a6cb71bcaed2029377bfa2872b4abb0 xctan xc-tan@outlook.com 2025-06-17T17:58:32+08:00 GitHub noreply@github.com 2025-06-17T12:58:32+03:00 ggml-cpu : remove the weak alias trick (#14221)
fe9d60e74a6cb71bcaed2029377bfa2872b4abb0 e434e69183fd9e1031f4445002083178c331a28b R0CKSTAR yeahdongcn@gmail.com 2025-06-17T17:48:08+08:00 GitHub noreply@github.com 2025-06-17T17:48:08+08:00 musa: fix build warning (unused variable) (#14231)
e434e69183fd9e1031f4445002083178c331a28b 89fea80d298184d1cd93564f48e060d9f541f4b4 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-16T21:58:42+02:00 GitHub noreply@github.com 2025-06-16T21:58:42+02:00 common : suggest --jinja when autodetection fails (#14222)
89fea80d298184d1cd93564f48e060d9f541f4b4 6adc3c3ebc029af058ac950a8e2a825fdf18ecc6 Georgi Gerganov ggerganov@gmail.com 2025-06-16T22:33:27+03:00 GitHub noreply@github.com 2025-06-16T22:33:27+03:00 server : fix incorrect usage of llama_get_embeddings() (#14225)
6adc3c3ebc029af058ac950a8e2a825fdf18ecc6 0dbcabde8c006d5cf781ca0fe071c41559572a72 Diego Devesa slarengh@gmail.com 2025-06-16T08:11:43-07:00 GitHub noreply@github.com 2025-06-16T08:11:43-07:00 llama : add thread safety test (#14035)
0dbcabde8c006d5cf781ca0fe071c41559572a72 ad590be98c83217fcf1a101d78d9ab389fd5dc0b bandoti 141645996+bandoti@users.noreply.github.com 2025-06-16T10:32:13-03:00 GitHub noreply@github.com 2025-06-16T10:32:13-03:00 cmake: clean up external project logic for vulkan-shaders-gen (#14179)
ad590be98c83217fcf1a101d78d9ab389fd5dc0b 7d6d91babfa129906b39c9099eca4234c44f4f1e Đinh Trọng Huy 77562200+huydt84@users.noreply.github.com 2025-06-16T21:53:41+09:00 GitHub noreply@github.com 2025-06-16T14:53:41+02:00 model : add NeoBERT (#14164)
7d6d91babfa129906b39c9099eca4234c44f4f1e d3e64b9f490cee41b7b9aa275dae2f6568ae3054 uvos philipp@uvos.xyz 2025-06-16T13:47:38+02:00 GitHub noreply@github.com 2025-06-16T13:47:38+02:00 HIP: disable rocwmma on gfx12 by default until rocm 7.0 (#14202)
d3e64b9f490cee41b7b9aa275dae2f6568ae3054 3ba0d843c6bd3faea5cf5e53dc7f3c82be20bffb Georgi Gerganov ggerganov@gmail.com 2025-06-16T14:14:00+03:00 GitHub noreply@github.com 2025-06-16T14:14:00+03:00 llama : rework embeddings logic (#14208)
3ba0d843c6bd3faea5cf5e53dc7f3c82be20bffb 0bf49eb668bb95b50e41583e22aaf60ddade1fbe Charles Xu charles.xu@arm.com 2025-06-16T11:47:57+02:00 GitHub noreply@github.com 2025-06-16T11:47:57+02:00 ggml: Add Android support for GGML_CPU_ALL_VARIANTS (#14206)
0bf49eb668bb95b50e41583e22aaf60ddade1fbe 4ad243677bca6c97f14dbc187b2116b51fcb7ffd Bartowski 3266127+bartowski1182@users.noreply.github.com 2025-06-16T09:16:06+01:00 GitHub noreply@github.com 2025-06-16T10:16:06+02:00 convert : remove arcee change in convert_hf_to_gguf_update.py (#14207)
4ad243677bca6c97f14dbc187b2116b51fcb7ffd c89c2d1ab94b11845240b7d3313c87691ea18d88 Đinh Trọng Huy 77562200+huydt84@users.noreply.github.com 2025-06-16T16:20:59+09:00 GitHub noreply@github.com 2025-06-16T09:20:59+02:00 gguf-py : allow key override when adding value to GGUFWriter (#14194)
c89c2d1ab94b11845240b7d3313c87691ea18d88 3555b3004ba7687be3d734acade52a3345758aa4 Jeff Bolz jbolz@nvidia.com 2025-06-16T00:21:08-06:00 GitHub noreply@github.com 2025-06-16T08:21:08+02:00 vulkan: mutex around vkQueueSubmit (#14127)
3555b3004ba7687be3d734acade52a3345758aa4 d7da8dc83a03b30e1ec10317080082ea76840c38 xctan xc-tan@outlook.com 2025-06-16T13:54:15+08:00 GitHub noreply@github.com 2025-06-16T13:54:15+08:00 ggml-cpu : rework weak alias on apple targets (#14146)
d7da8dc83a03b30e1ec10317080082ea76840c38 cd355eda7df1898d25d433b4bdaa4b4b479e0bad Bartowski 3266127+bartowski1182@users.noreply.github.com 2025-06-16T00:04:06+01:00 GitHub noreply@github.com 2025-06-16T01:04:06+02:00 model : Add support for Arcee AI's upcoming AFM model (#14185)
cd355eda7df1898d25d433b4bdaa4b4b479e0bad 30e5b01de2a0bcddc7c063c8ef0802703a958417 Eric Curtin ecurtin@redhat.com 2025-06-15T23:36:22+02:00 GitHub noreply@github.com 2025-06-15T23:36:22+02:00 server : When listening on a unix domain socket don't print http:// and port (#14180)
30e5b01de2a0bcddc7c063c8ef0802703a958417 e54b394082de242be4ee2e692b11fcc8d4eba371 Ed Addario 29247825+EAddario@users.noreply.github.com 2025-06-15T17:53:45+01:00 GitHub noreply@github.com 2025-06-15T18:53:45+02:00 quantize : change int to unsigned int for KV overrides (#14197)
e54b394082de242be4ee2e692b11fcc8d4eba371 2c2caa444341d99c87ff153f142c2d4762a776a2 uvos philipp@uvos.xyz 2025-06-15T17:30:13+02:00 GitHub noreply@github.com 2025-06-15T17:30:13+02:00 CUDA/HIP: fix ssm_scan on devices where warp size is not 32 (#14196)
2c2caa444341d99c87ff153f142c2d4762a776a2 5fce5f948df8f189a5401a8ecaa9753106e75abb uvos philipp@uvos.xyz 2025-06-15T15:45:27+02:00 GitHub noreply@github.com 2025-06-15T15:45:27+02:00 HIP: Replace usage of depricated preprocessor macro __AMDGCN_WAVEFRONT_SIZE__ (#14183)
5fce5f948df8f189a5401a8ecaa9753106e75abb 9ae4143bc6ecb4c2f0f0301578f619f6c201b857 Georgi Gerganov ggerganov@gmail.com 2025-06-15T10:52:11+03:00 GitHub noreply@github.com 2025-06-15T10:52:11+03:00 kv-cache : fix use-after-move of defrag info (#14189)
9ae4143bc6ecb4c2f0f0301578f619f6c201b857 c311ac664d68d10781a3e7b9f02d9d9520837d80 Mikko Juola mikjuo@gmail.com 2025-06-15T00:52:06-07:00 GitHub noreply@github.com 2025-06-15T09:52:06+02:00 model : add dots.llm1 architecture support (#14044) (#14118)
c311ac664d68d10781a3e7b9f02d9d9520837d80 b9912ac570de8945ae9383c9ca8291027bf287dd Georgi Gerganov ggerganov@gmail.com 2025-06-15T10:08:58+03:00 GitHub noreply@github.com 2025-06-15T10:08:58+03:00 cparams : rename LLAMA_MAX_PARALLEL_SEQUENCES to LLAMA_MAX_SEQ (#14188)
b9912ac570de8945ae9383c9ca8291027bf287dd 00ba7726100d7e1941d9f5a06f56a7559945b33c Georgi Gerganov ggerganov@gmail.com 2025-06-15T09:18:37+03:00 GitHub noreply@github.com 2025-06-15T09:18:37+03:00 batch : auto-gen positions + verify multi-sequence input (#14177)
00ba7726100d7e1941d9f5a06f56a7559945b33c 3cb203c89f60483e349f841684173446ed23c28f Pepijn de Vos me@pepijndevos.nl 2025-06-15T08:06:37+02:00 GitHub noreply@github.com 2025-06-15T08:06:37+02:00 docs : remove WIP since PR has been merged (#13912)
3cb203c89f60483e349f841684173446ed23c28f 2e42be42bd6bf1dcc643d6ac4e77419bfe5dd24f Piotr piotr.stankiewicz@docker.com 2025-06-14T18:25:15+02:00 GitHub noreply@github.com 2025-06-14T17:25:15+01:00 llama-chat : Do not throw when tool parsing fails (#14012)
2e42be42bd6bf1dcc643d6ac4e77419bfe5dd24f fb85a288d72abbd5e5daa8de96e6f8bfa7b5ab46 Aman Gupta amangupta052@gmail.com 2025-06-14T16:34:20+08:00 GitHub noreply@github.com 2025-06-14T10:34:20+02:00 compare-llama-bench: add option to plot (#14169)
fb85a288d72abbd5e5daa8de96e6f8bfa7b5ab46 40643edb86eb10b471b0f57d4f3f7eb0e06a0df7 Georgi Gerganov ggerganov@gmail.com 2025-06-13T20:03:05+03:00 GitHub noreply@github.com 2025-06-13T20:03:05+03:00 vocab : fix build (#14175)
40643edb86eb10b471b0f57d4f3f7eb0e06a0df7 3cfbbdb44e08fd19429fed6cc85b982a91f0efd5 Svetlozar Georgiev 55534064+sgeor255@users.noreply.github.com 2025-06-13T17:32:56+01:00 GitHub noreply@github.com 2025-06-13T18:32:56+02:00 sycl: fix docker image (#14144)
3cfbbdb44e08fd19429fed6cc85b982a91f0efd5 80709b70a2f87c13ccaf1480b799393109996789 Guy Goldenberg guy110698@gmail.com 2025-06-13T19:20:25+03:00 GitHub noreply@github.com 2025-06-13T19:20:25+03:00 Merge commit from fork
80709b70a2f87c13ccaf1480b799393109996789 26ff3685bfbaa4c8838d7afd988b17dd5eb99f92 Georgi Gerganov ggerganov@gmail.com 2025-06-13T18:35:00+03:00 GitHub noreply@github.com 2025-06-13T18:35:00+03:00 batch : add LLAMA_BATCH_DEBUG environment variable (#14172)
26ff3685bfbaa4c8838d7afd988b17dd5eb99f92 60c666347becacff81cd4bc9a52038ba71038e41 ddpasa 112642920+ddpasa@users.noreply.github.com 2025-06-13T15:17:53+02:00 GitHub noreply@github.com 2025-06-13T15:17:53+02:00 docs : Update multimodal.md (#14122)
60c666347becacff81cd4bc9a52038ba71038e41 b7cc7745e38141060845145af0a1fd489d8e3e33 Georgi Gerganov ggerganov@gmail.com 2025-06-13T13:47:55+03:00 GitHub noreply@github.com 2025-06-13T13:47:55+03:00 batch : rework llama_batch_allocr (#14153)
b7cc7745e38141060845145af0a1fd489d8e3e33 cc8d08187918c6f643c3ffabb7b1ac21aa19f3d1 Georgi Gerganov ggerganov@gmail.com 2025-06-13T11:55:44+03:00 GitHub noreply@github.com 2025-06-13T11:55:44+03:00 readme : remove survey link (#14168)
cc8d08187918c6f643c3ffabb7b1ac21aa19f3d1 d714dadb57d8feaa03d13b79345a4c3382172d61 Christian Kastner ckk@kvr.at 2025-06-13T08:38:52Z GitHub noreply@github.com 2025-06-13T10:38:52+02:00 cmake: Add ability to pass in LLAMA_BUILD_NUMBER/COMMIT (#14167)
d714dadb57d8feaa03d13b79345a4c3382172d61 ffad04397399ea1650fda6560c7c753059804876 Đinh Trọng Huy 77562200+huydt84@users.noreply.github.com 2025-06-13T17:34:08+09:00 GitHub noreply@github.com 2025-06-13T11:34:08+03:00 pooling : make cls_b and cls_out_b optional (#14165)
ffad04397399ea1650fda6560c7c753059804876 0889eba570126f8a2f5a0e88fde776bbc91cca66 Georgi Gerganov ggerganov@gmail.com 2025-06-13T11:18:25+03:00 GitHub noreply@github.com 2025-06-13T11:18:25+03:00 server : fix SWA condition for full context reprocess (#14163)
0889eba570126f8a2f5a0e88fde776bbc91cca66 c61285e7396c8e526fe7794c19e8d4f1c99bfc51 Anton Mitkov anton.mitkov@codeplay.com 2025-06-13T08:51:39+01:00 GitHub noreply@github.com 2025-06-13T08:51:39+01:00 sycl: Adding additional cpy dbg print output (#14034)
c61285e7396c8e526fe7794c19e8d4f1c99bfc51 09cf2c7c655c90e53e100f29b830a788bab0653d Ewan Crawford ewan@codeplay.com 2025-06-13T08:45:37+01:00 GitHub noreply@github.com 2025-06-13T08:45:37+01:00 SYCL: Bump oneMath commit (#14152)
09cf2c7c655c90e53e100f29b830a788bab0653d c33fe8b8c4427202706b1434e9fc8ab5752c9cac Christian Kastner ckk@kvr.at 2025-06-13T06:51:34Z GitHub noreply@github.com 2025-06-13T09:51:34+03:00 cmake : Improve build-info.cpp generation (#14156)
c33fe8b8c4427202706b1434e9fc8ab5752c9cac ed52f3668e633423054a4eab61bb7efee47025ab Georgi Gerganov ggerganov@gmail.com 2025-06-13T08:03:54+03:00 GitHub noreply@github.com 2025-06-13T08:03:54+03:00 vocab : prevent heap overflow when vocab is too small (#14145)
6391ee7ffd97e723dfbdcc32f001e1b95a63cba6 b9850050a5b5147cb2e6720b98ed8c1e8e02eff9 Yunzhe Jia yunzhe@calculet.tech 2025-06-13T10:58:06+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-13T10:58:06+08:00 update .gitignore to update cal_test.gguf
b9850050a5b5147cb2e6720b98ed8c1e8e02eff9 933b205c32551848589b8339437d778cb4f2a46d Yunzhe Jia yunzhe@calculet.tech 2025-06-06T17:25:52+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-13T09:20:23+08:00 calrt: add call chain for using calrt::infer
ed52f3668e633423054a4eab61bb7efee47025ab a681b4ba83a61dce71a4f24e558efe7278d8b1a9 Anton Mitkov anton.mitkov@codeplay.com 2025-06-12T14:15:11+01:00 GitHub noreply@github.com 2025-06-12T15:15:11+02:00 sycl: Remove not needed copy f16->f32 for dnnl mul mat (#14125)
a681b4ba83a61dce71a4f24e558efe7278d8b1a9 7d516443dd7766569110b38e6374649bee6eb1c4 Georgi Gerganov ggerganov@gmail.com 2025-06-12T14:43:09+03:00 GitHub noreply@github.com 2025-06-12T14:43:09+03:00 readme : remove project status link (#14149)
933b205c32551848589b8339437d778cb4f2a46d 33fea9a8592fd338abe279b89350a3af2e0ff1f7 Yunzhe Jia yunzhe@calculet.tech 2025-06-12T17:47:19+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-12T17:50:26+08:00 gguf-py: add cal_test.py
7d516443dd7766569110b38e6374649bee6eb1c4 f6e1a7aa8787b5c00acba6370cb70a0beff48b1e Georgi Gerganov ggerganov@gmail.com 2025-06-12T11:51:38+03:00 GitHub noreply@github.com 2025-06-12T11:51:38+03:00 server : re-enable SWA speculative decoding (#14131)
f6e1a7aa8787b5c00acba6370cb70a0beff48b1e c3ee46fab49a765d2e32e171e9ed7a5fa121dd9c Georgi Gerganov ggerganov@gmail.com 2025-06-12T11:50:01+03:00 GitHub noreply@github.com 2025-06-12T11:50:01+03:00 context : simplify output counting logic during decode (#14142)
c3ee46fab49a765d2e32e171e9ed7a5fa121dd9c e2c0b6e46a5596665569ae765f0993cea2619af6 Georgi Gerganov ggerganov@gmail.com 2025-06-12T11:49:26+03:00 GitHub noreply@github.com 2025-06-12T11:49:26+03:00 batch : remove logits_all flag (#14141)
e2c0b6e46a5596665569ae765f0993cea2619af6 9596506965f65be5d802ecef6a315fe43d2391a8 Georgi Gerganov ggerganov@gmail.com 2025-06-12T10:14:24+03:00 GitHub noreply@github.com 2025-06-12T10:14:24+03:00 cmake : handle whitepsaces in path during metal build (#14126)
9596506965f65be5d802ecef6a315fe43d2391a8 a20b2b05bce6622c585459ebf46f142f113d021c Georgi Gerganov ggerganov@gmail.com 2025-06-12T10:02:15+03:00 GitHub noreply@github.com 2025-06-12T10:02:15+03:00 kv-cache : fix split_equal handling in unified implementation (#14130)
a20b2b05bce6622c585459ebf46f142f113d021c 2e89f76b7af2c0b827be785e445f2e2b3e52e1ca compilade git@compilade.net 2025-06-12T02:56:04-04:00 GitHub noreply@github.com 2025-06-12T02:56:04-04:00 context : round n_tokens to next multiple of n_seqs when reserving (#14140)
2e89f76b7af2c0b827be785e445f2e2b3e52e1ca 532802f938c6a18cc6a704057ab571f253fd77ed bandoti 141645996+bandoti@users.noreply.github.com 2025-06-11T17:19:44-03:00 GitHub noreply@github.com 2025-06-11T17:19:44-03:00 common: fix issue with regex_escape routine on windows (#14133)
532802f938c6a18cc6a704057ab571f253fd77ed d4e0d95cf581f50c9a21d06eaecae2dd580076bd Christian Kastner ckk@kvr.at 2025-06-11T19:07:44Z GitHub noreply@github.com 2025-06-11T21:07:44+02:00 Implement GGML_CPU_ALL_VARIANTS for ARM (#14080)
d4e0d95cf581f50c9a21d06eaecae2dd580076bd cc66a7f78f95dcb5208420f4dd1abc3fc6aec0cc Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-11T19:04:23+02:00 GitHub noreply@github.com 2025-06-11T19:04:23+02:00 chore : clean up relative source dir paths (#14128)
cc66a7f78f95dcb5208420f4dd1abc3fc6aec0cc bd248d4dc7265437e1918dc53ae3f49a0c592e5f Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-11T17:16:32+02:00 GitHub noreply@github.com 2025-06-11T17:16:32+02:00 tests : add test-tokenizers-repo (#14017)
bd248d4dc7265437e1918dc53ae3f49a0c592e5f 7781e5fe99f2a4fc1c8af0a8488eedac4644cb72 Jeff Bolz jbolz@nvidia.com 2025-06-11T09:48:52-05:00 GitHub noreply@github.com 2025-06-11T09:48:52-05:00 vulkan: Better thread-safety for command pools/buffers (#14116)
7781e5fe99f2a4fc1c8af0a8488eedac4644cb72 89a184fa7125b7c3e2fb567337cc2bd7bc2d376c Aman amangupta052@gmail.com 2025-06-11T22:42:25+08:00 GitHub noreply@github.com 2025-06-11T16:42:25+02:00 webui: Wrap long numbers instead of infinite horizontal scroll (#14062)
89a184fa7125b7c3e2fb567337cc2bd7bc2d376c 2baf07727f921d9a4a1b63a2eff941e95d0488ed Georgi Gerganov ggerganov@gmail.com 2025-06-11T16:48:45+03:00 GitHub noreply@github.com 2025-06-11T16:48:45+03:00 kv-cache : relax SWA masking condition (#14119)
2baf07727f921d9a4a1b63a2eff941e95d0488ed 7ae2932116a4de3141cbc8c488f7f6e4e06d8171 Taylor quantumtraveling@gmail.com 2025-06-11T06:43:43-04:00 GitHub noreply@github.com 2025-06-11T13:43:43+03:00 server : pass default --keep argument (#14120)
7ae2932116a4de3141cbc8c488f7f6e4e06d8171 1f7d50b2936023b26eb218e944e62834b80a2ce0 Georgi Gerganov ggerganov@gmail.com 2025-06-11T12:52:45+03:00 GitHub noreply@github.com 2025-06-11T12:52:45+03:00 kv-cache : add LLAMA_KV_CACHE_DEBUG environment variable (#14121)
1f7d50b2936023b26eb218e944e62834b80a2ce0 4c763c8d1b4d4de20bf364ec1837430783cba984 Jeff Bolz jbolz@nvidia.com 2025-06-11T00:19:25-05:00 GitHub noreply@github.com 2025-06-11T07:19:25+02:00 vulkan: Track descriptor pools/sets per-context (#14109)
4c763c8d1b4d4de20bf364ec1837430783cba984 dad5c44398b78467943ed0a603ea427fe9f6fc62 lhez quic_lih@quicinc.com 2025-06-10T16:55:58-07:00 GitHub noreply@github.com 2025-06-10T16:55:58-07:00 opencl: add `mul_mv_id_q4_0_f32_8x_flat` (#14003)
dad5c44398b78467943ed0a603ea427fe9f6fc62 55f6b9fa6563f6ae49113f9abdc980c12348cc1c compilade git@compilade.net 2025-06-10T18:20:14-04:00 GitHub noreply@github.com 2025-06-10T18:20:14-04:00 kv-cache : avoid modifying recurrent cells when setting inputs (#13834)
55f6b9fa6563f6ae49113f9abdc980c12348cc1c 3678b838bb71eaccbaeb479ff38c2e12bfd2f960 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-10T23:29:52+02:00 GitHub noreply@github.com 2025-06-10T23:29:52+02:00 convert : fix duplicate key DeepSeek-R1 conversion error (#14103)
3678b838bb71eaccbaeb479ff38c2e12bfd2f960 652b70e6678d503626f3c9d33831ba604b473401 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-10T18:02:08+02:00 GitHub noreply@github.com 2025-06-10T18:02:08+02:00 llama : support GEGLU for jina-bert-v2 (#14090)
652b70e6678d503626f3c9d33831ba604b473401 3a12db23b6918682ccb70ab15d897f11fe5fc320 Jeff Bolz jbolz@nvidia.com 2025-06-10T10:53:47-05:00 GitHub noreply@github.com 2025-06-10T10:53:47-05:00 vulkan: force device 0 in CI (#14106)
3a12db23b6918682ccb70ab15d897f11fe5fc320 ae92c1855b1a5b604fa1bfcced19a556ab3e78c5 Juk Armstrong 69222624+jukofyork@users.noreply.github.com 2025-06-10T16:48:07+01:00 GitHub noreply@github.com 2025-06-10T16:48:07+01:00 Fixed spec timings to: accepted/tested instead of accepted/drafted (#14104)
ae92c1855b1a5b604fa1bfcced19a556ab3e78c5 b7ce1ad1e332da5909772d173a7e6748fbd1887a Georgi Gerganov ggerganov@gmail.com 2025-06-10T17:37:45+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-10T18:39:33+03:00 sync : ggml
b7ce1ad1e332da5909772d173a7e6748fbd1887a 97340b4c9924be86704dbf155e97c8319849ee19 Georgi Gerganov ggerganov@gmail.com 2025-06-10T11:34:10+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-10T18:39:33+03:00 ggml : fix weak alias win32 (whisper/0)
97340b4c9924be86704dbf155e97c8319849ee19 2bb0467043258bdc58dbaefb33786f1731b38937 0cc4m picard12@live.de 2025-06-10T14:01:33+02:00 GitHub noreply@github.com 2025-06-10T13:01:33+01:00 Vulkan: Don't default to CPU device (like llvmpipe), even if no other device is available, to allow fallback to CPU backend (#14099)
2bb0467043258bdc58dbaefb33786f1731b38937 b8e2194efc529378be45ab9b27d6648a5b81458a Isaac McFadyen isaac@imcf.me 2025-06-10T02:41:01-04:00 GitHub noreply@github.com 2025-06-10T09:41:01+03:00 rpc : nicer error messages for RPC server crash (#14076)
b8e2194efc529378be45ab9b27d6648a5b81458a 1a3b5e80f77c51e6d9fb7762f19b9e61fceca3de Georgi Gerganov ggerganov@gmail.com 2025-06-10T09:20:51+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-10T09:21:56+03:00 sync : ggml
1a3b5e80f77c51e6d9fb7762f19b9e61fceca3de 1f63e75f3b5dc7f44dbe63c8a41d23958fe95bc0 Kai Pastor dg0yt@darc.de 2025-06-03T12:33:28+02:00 Georgi Gerganov ggerganov@gmail.com 2025-06-10T09:21:56+03:00 Add in-build ggml::ggml ALIAS library (ggml/1260)
1f63e75f3b5dc7f44dbe63c8a41d23958fe95bc0 40cbf571c9210031340f022d104317e89a1a6bb2 Georgi Gerganov ggerganov@gmail.com 2025-06-09T23:05:02+03:00 GitHub noreply@github.com 2025-06-09T23:05:02+03:00 metal : use less stack memory in FA kernel (#14088)
40cbf571c9210031340f022d104317e89a1a6bb2 7f4fbe5183b23b6b2e25fd1ccc5d1fa8bb010cb7 Georgi Gerganov ggerganov@gmail.com 2025-06-09T23:04:35+03:00 GitHub noreply@github.com 2025-06-09T23:04:35+03:00 kv-cache : fix shift and defrag logic (#14081)
7f4fbe5183b23b6b2e25fd1ccc5d1fa8bb010cb7 f470bc36bed4d836b9de5a483fa0dfaee176d6f5 Diego Devesa slarengh@gmail.com 2025-06-09T11:03:09-07:00 GitHub noreply@github.com 2025-06-09T20:03:09+02:00 llama : allow building all tests on windows when not using shared libs (#13980)
f470bc36bed4d836b9de5a483fa0dfaee176d6f5 8f47e25f56e9792093b7497c68e9f80bab82ed19 xctan axunlei@gmail.com 2025-06-09T22:47:13+08:00 GitHub noreply@github.com 2025-06-09T16:47:13+02:00 ggml-cpu : split arch-specific implementations (#13892)
8f47e25f56e9792093b7497c68e9f80bab82ed19 201b31dc2e6fef3de598280b53fd3b69420a4e49 Diego Devesa slarengh@gmail.com 2025-06-09T07:36:26-07:00 GitHub noreply@github.com 2025-06-09T16:36:26+02:00 cuda : fix device sync on buffer clear (#14033)
201b31dc2e6fef3de598280b53fd3b69420a4e49 e21d2d4ae26d6c5daf1602a1061aa4f8e722ca57 Georgi Gerganov ggerganov@gmail.com 2025-06-09T17:17:31+03:00 GitHub noreply@github.com 2025-06-09T17:17:31+03:00 graph : fix geglu (#14077)
e21d2d4ae26d6c5daf1602a1061aa4f8e722ca57 dc0623fddb661926e0998538895155e4f081ff09 Xinpeng Dou 15529241576@163.com 2025-06-09T19:47:39+08:00 GitHub noreply@github.com 2025-06-09T19:47:39+08:00 CANN: Simplify the environment variable setting(#13104)
dc0623fddb661926e0998538895155e4f081ff09 87d34b381d5868e75586210ec17b5ef5deddc276 R0CKSTAR yeahdongcn@gmail.com 2025-06-09T18:01:17+08:00 GitHub noreply@github.com 2025-06-09T12:01:17+02:00 webui: fix sidebar being covered by main content (#14082)
87d34b381d5868e75586210ec17b5ef5deddc276 b460d16ae858c6624fd37aec316622a4060ca325 Georgi Gerganov ggerganov@gmail.com 2025-06-09T12:57:58+03:00 GitHub noreply@github.com 2025-06-09T12:57:58+03:00 server : fix LRU check (#14079)
b460d16ae858c6624fd37aec316622a4060ca325 91a8ee6a6f1f4c8547ff7b745ef95c6edc1d2af6 Nicolò Scipione nicolo.scipione@codeplay.com 2025-06-09T11:47:07+02:00 GitHub noreply@github.com 2025-06-09T11:47:07+02:00 sycl: Add reorder to Q6_K mmvq implementation (#13885)
91a8ee6a6f1f4c8547ff7b745ef95c6edc1d2af6 056eb74534ffd3efc50a9b854156eb6876a52a44 Đinh Trọng Huy 77562200+huydt84@users.noreply.github.com 2025-06-09T13:15:31+09:00 GitHub noreply@github.com 2025-06-09T05:15:31+01:00 add geglu activation function (#14074)
056eb74534ffd3efc50a9b854156eb6876a52a44 247e5c6e447707bb4539bdf1913d206088a8fc69 Yuanhao Ji jiyuanhao@apache.org 2025-06-09T11:20:06+08:00 GitHub noreply@github.com 2025-06-09T11:20:06+08:00 CANN: Enable labeler for Ascend NPU (#13914)
247e5c6e447707bb4539bdf1913d206088a8fc69 5787b5da57e54dba760c2deeac1edf892e8fc450 Diego Devesa slarengh@gmail.com 2025-06-08T11:39:56-07:00 GitHub noreply@github.com 2025-06-08T11:39:56-07:00 cuda : fix buffer type check with integrated GPUs (#14069)
5787b5da57e54dba760c2deeac1edf892e8fc450 228f34c9ceefa3ea4f4d6933edd858121e8106cb 吴小白 296015668@qq.com 2025-06-07T21:39:11+08:00 GitHub noreply@github.com 2025-06-07T10:39:11-03:00 ci: add LoongArch cross-compile build (#13944)
228f34c9ceefa3ea4f4d6933edd858121e8106cb 0974ad7a7cd4bca846b15c484ff3be890135a52c Akarshan Biswas akarshan@menlo.ai 2025-06-07T18:58:20+05:30 GitHub noreply@github.com 2025-06-07T18:58:20+05:30 SYCL: Implement few same quantized type copy kernels (#13739)
0974ad7a7cd4bca846b15c484ff3be890135a52c 745aa5319b9930068aff5e87cf5e9eef7227339b Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-07T14:13:12+02:00 GitHub noreply@github.com 2025-06-07T14:13:12+02:00 llama : fix llama_model_chat_template with template name (LLM_KV with suffix) (#14050)
745aa5319b9930068aff5e87cf5e9eef7227339b 487a5e0401423bba02cd6e97e4d45131bb20b22b Georgi Gerganov ggerganov@gmail.com 2025-06-06T14:11:15+03:00 GitHub noreply@github.com 2025-06-06T14:11:15+03:00 llama : deprecate llama_kv_self_ API (#14030)
487a5e0401423bba02cd6e97e4d45131bb20b22b d17a809ef0af09b16625e991a76f6fe80d9c332e Georgi Gerganov ggerganov@gmail.com 2025-06-06T13:29:18+03:00 GitHub noreply@github.com 2025-06-06T13:29:18+03:00 context : fix SWA-related warning for multiple sequences (#14045)
33fea9a8592fd338abe279b89350a3af2e0ff1f7 806f05a353d322e97d97e02f910de2b408a8b4ab Yunzhe Jia yunzhe@calculet.tech 2025-06-06T17:25:52+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-06T17:35:48+08:00 calrt: add load calbin
d17a809ef0af09b16625e991a76f6fe80d9c332e 1caae7fc6c77551cb1066515e0f414713eebb367 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-06T09:03:25+02:00 GitHub noreply@github.com 2025-06-06T09:03:25+02:00 llama : support multiple classifier outputs and labels (#13940)
806f05a353d322e97d97e02f910de2b408a8b4ab ec9e0301fef6476df83e94842c3b625501c95566 Yunzhe Jia yunzhe@calculet.tech 2025-05-30T09:45:37+08:00 Yunzhe Jia yunzhe@calculet.tech 2025-06-06T09:24:26+08:00 add calrt demo
1caae7fc6c77551cb1066515e0f414713eebb367 669c13e0f67edfba891c5bd7105eb4376bcf8626 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-05T17:42:31+02:00 GitHub noreply@github.com 2025-06-05T17:42:31+02:00 gguf-py : add add_classifier_output_labels method to writer (#14031)
669c13e0f67edfba891c5bd7105eb4376bcf8626 146b88e8b3e16d22cea22f8be982a4d45e913b1e Masato Nakasaka rillomas@gmail.com 2025-06-05T23:00:29+09:00 GitHub noreply@github.com 2025-06-05T16:00:29+02:00 vulkan: Enable VK_KHR_cooperative_matrix extension for Intel Xe2 GPUs (#14001)
146b88e8b3e16d22cea22f8be982a4d45e913b1e 7f37b6cf1e2c1b90bf0d9c8d91904b4b6c512748 pockers21 134406831+pockers21@users.noreply.github.com 2025-06-05T06:25:29-07:00 GitHub noreply@github.com 2025-06-05T16:25:29+03:00 ci: fix CUDA build failure on autodl cloud machines (#14005)
7f37b6cf1e2c1b90bf0d9c8d91904b4b6c512748 3a077146a4761fdbd24bdd8eb098f46b8adc4dda Georgi Gerganov ggerganov@gmail.com 2025-06-05T15:29:22+03:00 GitHub noreply@github.com 2025-06-05T15:29:22+03:00 memory : migrate from llama_kv_cache to more generic llama_memory (#14006)
3a077146a4761fdbd24bdd8eb098f46b8adc4dda d01d112abb055549d33bb8aac1755eb0c40918ad Diego Devesa slarengh@gmail.com 2025-06-05T02:57:42-07:00 GitHub noreply@github.com 2025-06-05T11:57:42+02:00 llama : allow using mmap without PrefetchVirtualMemory, apply GGML_WIN_VER to llama.cpp sources (#14013)
d01d112abb055549d33bb8aac1755eb0c40918ad 9f47fa5792bae5312615439e68dbf1826913f7ac Olexandr88 radole1203@gmail.com 2025-06-05T10:50:55+03:00 GitHub noreply@github.com 2025-06-05T10:50:55+03:00 readme : add badge (#13938)
9f47fa5792bae5312615439e68dbf1826913f7ac 9e31bec4fd53634c9e5b04650488a09a055f5dab Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-05T09:29:18+02:00 GitHub noreply@github.com 2025-06-05T09:29:18+02:00 vocab : warn about missing mask token (#14022)
9e31bec4fd53634c9e5b04650488a09a055f5dab 5a8ae3053ced350ed300ba91600519fcad1c6ba7 Georgi Gerganov ggerganov@gmail.com 2025-06-05T09:06:29+03:00 GitHub noreply@github.com 2025-06-05T09:06:29+03:00 context : fix pos_min initialization upon error decode (#14008)
5a8ae3053ced350ed300ba91600519fcad1c6ba7 0d3984424f2973c49c4bcabe4cc0153b4f90c601 Jeff Bolz jbolz@nvidia.com 2025-06-05T00:17:58-05:00 GitHub noreply@github.com 2025-06-05T07:17:58+02:00 vulkan: automatically deduce size of push constants (#13936)
0d3984424f2973c49c4bcabe4cc0153b4f90c601 3e63a58ef7addec35408e2eb67850d7cdc935dc3 Ervin Áron Tasnádi etasnadi@protonmail.com 2025-06-04T22:02:00+02:00 GitHub noreply@github.com 2025-06-04T22:02:00+02:00 ggml-vulkan: adds support for op CONV_TRANSPOSE_1D (#13813)
3e63a58ef7addec35408e2eb67850d7cdc935dc3 2589ad3704559f4dd860f5f303b19349c688a28a Georgi Gerganov ggerganov@gmail.com 2025-06-04T18:58:20+03:00 GitHub noreply@github.com 2025-06-04T18:58:20+03:00 kv-cache : refactor the update/defrag mechanism (#13988)
2589ad3704559f4dd860f5f303b19349c688a28a 482548716f664f76e325ded58c9e8b7563e5e23a Diego Devesa slarengh@gmail.com 2025-06-04T06:37:40-07:00 GitHub noreply@github.com 2025-06-04T15:37:40+02:00 ci : remove cuda 11.7 releases, switch runner to windows 2022 (#13997)
482548716f664f76e325ded58c9e8b7563e5e23a 3ac67535c86e2fc43e4eddf594412acc370bbb04 Diego Devesa slarengh@gmail.com 2025-06-04T04:15:54-07:00 GitHub noreply@github.com 2025-06-04T13:15:54+02:00 releases : use dl backend for linux release, remove arm64 linux release (#13996)
3ac67535c86e2fc43e4eddf594412acc370bbb04 0b4be4c435849b00dbd98b109cf7a22298d27b69 Xuan-Son Nguyen son@huggingface.co 2025-06-04T10:11:26+02:00 GitHub noreply@github.com 2025-06-04T10:11:26+02:00 llama-graph : use ggml_repeat_4d (#13998)
0b4be4c435849b00dbd98b109cf7a22298d27b69 e0e806f52ebcd0ee285c994fe8fd8b8787d2cb0a Johannes Gäßler johannesg@5d6.de 2025-06-04T08:57:05+02:00 GitHub noreply@github.com 2025-06-04T08:57:05+02:00 CUDA: fix FTZ in FA for Gemma 3 (#13991)
e0e806f52ebcd0ee285c994fe8fd8b8787d2cb0a 7e00e60ef86645a01fda738fef85b74afa016a34 Georgi Gerganov ggerganov@gmail.com 2025-06-04T09:50:32+03:00 GitHub noreply@github.com 2025-06-04T09:50:32+03:00 kv-cache : fix unified::seq_rm to work with seq_id < 0 (#13985)
7e00e60ef86645a01fda738fef85b74afa016a34 ea1431b0fa3a8108aac1e0a94a13ccc4a749963e Jeff Bolz jbolz@nvidia.com 2025-06-03T13:30:22-05:00 GitHub noreply@github.com 2025-06-03T20:30:22+02:00 vulkan: fix warnings in perf logger querypool code (#13937)
ea1431b0fa3a8108aac1e0a94a13ccc4a749963e 71e74a3ac929b8af91f16f73f3c2b9b2f796d207 Xuan-Son Nguyen son@huggingface.co 2025-06-03T13:09:36+02:00 GitHub noreply@github.com 2025-06-03T13:09:36+02:00 docs : add "Quick start" section for new users (#13862)
71e74a3ac929b8af91f16f73f3c2b9b2f796d207 bfb1e012a0b7658e8f00ed4333d059943ea9d648 lhez quic_lih@quicinc.com 2025-06-02T16:54:58-07:00 GitHub noreply@github.com 2025-06-02T16:54:58-07:00 opencl: add `backend_synchronize` (#13939)
bfb1e012a0b7658e8f00ed4333d059943ea9d648 363757628848a27a435bbf22ff9476e9aeda5f40 rmatif 66360289+rmatif@users.noreply.github.com 2025-06-02T23:53:36Z GitHub noreply@github.com 2025-06-02T16:53:36-07:00 OpenCL: Add concat, tsembd, upscale, tanh, pad and repeat (#13840)
363757628848a27a435bbf22ff9476e9aeda5f40 ea394d7ab1f8101716d48ce9421c94c71b93a00f Georgi Gerganov ggerganov@gmail.com 2025-06-02T21:34:40+03:00 GitHub noreply@github.com 2025-06-02T21:34:40+03:00 server : disable speculative decoding for SWA models (#13970)
ea394d7ab1f8101716d48ce9421c94c71b93a00f 5582c49c3961269eca96822abfb87528e942dd07 Georgi Gerganov ggerganov@gmail.com 2025-06-02T21:33:40+03:00 GitHub noreply@github.com 2025-06-02T21:33:40+03:00 metal : use F32 accumulators in FA kernels (#13975)
5582c49c3961269eca96822abfb87528e942dd07 c9bbc77931d223ed7e7cbcf1cb057bc02fd0db19 Georgi Gerganov ggerganov@gmail.com 2025-06-02T20:54:26+03:00 GitHub noreply@github.com 2025-06-02T20:54:26+03:00 gemma : more consistent attention scaling for v2 and v3 (#13951)
c9bbc77931d223ed7e7cbcf1cb057bc02fd0db19 bfd322796cd838f906535ff3352624fc46338894 Olivier Chafik olivier.chafik@gmail.com 2025-06-02T10:15:44-07:00 GitHub noreply@github.com 2025-06-02T10:15:44-07:00 `server`: update deepseek reasoning format (pass reasoning_content as diffs) (#13933)
bfd322796cd838f906535ff3352624fc46338894 093e3f1feb16e25e58f7d61e01266c830dd424b8 Xuan-Son Nguyen son@huggingface.co 2025-06-02T16:29:28+02:00 GitHub noreply@github.com 2025-06-02T16:29:28+02:00 mtmd : fix memory leak in mtmd_helper_eval_chunk_single (#13961)
093e3f1feb16e25e58f7d61e01266c830dd424b8 663445b0deb21fb602176da030d4154197a4fca6 shalinib-ibm Shalini.Salomi.Bodapati@ibm.com 2025-06-02T17:48:36+05:30 GitHub noreply@github.com 2025-06-02T15:18:36+03:00 cmake : Handle mixed-case 'Power' strings in POWER CPU detection (#13966)
663445b0deb21fb602176da030d4154197a4fca6 7675c555a13c9f473249e59a54db35032ce8e0fc Atharva Dubey atharva.dubey@codeplay.com 2025-06-02T10:12:20+01:00 GitHub noreply@github.com 2025-06-02T10:12:20+01:00 sycl: quantize and reorder the input to q8_1 when reorder is enabled (#13826)
7675c555a13c9f473249e59a54db35032ce8e0fc 5e1c3aed4074480f63e914d6c44c93536ed1452a Johannes Gäßler johannesg@5d6.de 2025-06-01T18:08:05+02:00 GitHub noreply@github.com 2025-06-01T18:08:05+02:00 gguf: fix failure on version == 0 (#13956)
5e1c3aed4074480f63e914d6c44c93536ed1452a c496fe0b1da28618dd17bda8b0a6bf5004554080 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-01T18:07:21+02:00 GitHub noreply@github.com 2025-06-01T18:07:21+02:00 convert : fix nomic-bert-moe mask token (#13757)
c496fe0b1da28618dd17bda8b0a6bf5004554080 e57bb87cede38341963a7a884630dbfb09c7dc00 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-06-01T17:23:11+02:00 GitHub noreply@github.com 2025-06-01T17:23:11+02:00 convert : fix vocab padding code for bert models (#13954)
e57bb87cede38341963a7a884630dbfb09c7dc00 f3a4b1659ccc94ebdf3590d472b518377bfecc7a Aaron Teo aaron.teo1@ibm.com 2025-06-01T22:53:57+08:00 GitHub noreply@github.com 2025-06-01T16:53:57+02:00 ggml: check if non-native endian model is being loaded (#13943)
f3a4b1659ccc94ebdf3590d472b518377bfecc7a 108009f5c7267b77292c11f788f602d6d7ae8fcd Georgi Gerganov ggerganov@gmail.com 2025-06-01T12:23:14+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-01T13:43:57+03:00 sync : ggml
108009f5c7267b77292c11f788f602d6d7ae8fcd d337252acf14a91a685c355fa4f3f599a8068207 Kai Pastor dg0yt@darc.de 2025-05-31T12:49:55+02:00 Georgi Gerganov ggerganov@gmail.com 2025-06-01T13:43:57+03:00 vulkan : Remove unexpected ; (ggml/1253)
d337252acf14a91a685c355fa4f3f599a8068207 af6f91db470da543dc32bc00c057aad8f060dfdb Kai Pastor dg0yt@darc.de 2025-05-31T12:39:19+02:00 Georgi Gerganov ggerganov@gmail.com 2025-06-01T13:43:57+03:00 cmake : Fix broken CMake error messages (ggml/1252)
af6f91db470da543dc32bc00c057aad8f060dfdb a7b8d35f780ab3e7639ae5345442fb1ad10f0163 Radoslav Gerganov rgerganov@gmail.com 2025-05-30T09:11:09+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-01T13:43:57+03:00 ggml : remove ggml_graph_import and ggml_graph_export declarations (ggml/1247)
a7b8d35f780ab3e7639ae5345442fb1ad10f0163 6eba72b71c677ce9aae33c422a6e67008137987a Georgi Gerganov ggerganov@gmail.com 2025-05-29T13:29:50+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-01T13:43:57+03:00 sync : whisper.cpp (ggml/1250)
6eba72b71c677ce9aae33c422a6e67008137987a fedf034a98378059274e4243d2a370a243335e73 Radoslav Gerganov rgerganov@gmail.com 2025-05-29T08:34:46+03:00 Georgi Gerganov ggerganov@gmail.com 2025-06-01T13:43:57+03:00 ggml : install dynamic backends (ggml/1240)
fedf034a98378059274e4243d2a370a243335e73 8726392d3d927fe3e504868d3548d694496ee37e Daniel Tang danielzgtg.opensource@gmail.com 2025-05-27T20:58:46-04:00 Georgi Gerganov ggerganov@gmail.com 2025-06-01T13:43:57+03:00 ggml : Print backtrace on uncaught C++ exceptions (ggml/1232)
8726392d3d927fe3e504868d3548d694496ee37e c04621711a893cbd09cff6c927cb005bc6749e36 ddh0 chemist-mulches-39@icloud.com 2025-06-01T03:44:30-05:00 GitHub noreply@github.com 2025-06-01T11:44:30+03:00 readme : update bindings (#13950)
c04621711a893cbd09cff6c927cb005bc6749e36 0fc16b42e8793364918e830c234c1f3caec29dae Georgi Gerganov ggerganov@gmail.com 2025-06-01T11:42:16+03:00 GitHub noreply@github.com 2025-06-01T11:42:16+03:00 parallel : fix n_junk == 0 (#13952)
0fc16b42e8793364918e830c234c1f3caec29dae 053b1539c02617eff744f89525ee57497c3c1fbe Georgi Gerganov ggerganov@gmail.com 2025-06-01T11:39:27+03:00 GitHub noreply@github.com 2025-06-01T11:39:27+03:00 kv-cache : split implementation in separate sources (#13920)
053b1539c02617eff744f89525ee57497c3c1fbe b3a89c3d9e34c28c5be70d8b687a84775746d4a0 Max Krasnyansky quic_maxk@quicinc.com 2025-05-31T15:39:19-07:00 GitHub noreply@github.com 2025-05-31T15:39:19-07:00 threading: support for GGML_SCHED_PRIO_LOW, update thread info on Windows to avoid throttling (#12995)
b3a89c3d9e34c28c5be70d8b687a84775746d4a0 e15898d1c7e87f1b3d14e0a3c385eeb424b085fe Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2025-05-31T18:58:35+02:00 GitHub noreply@github.com 2025-05-31T18:58:35+02:00 docs : Note about necessity of having libcurl installed for standard build. (#13945)
e15898d1c7e87f1b3d14e0a3c385eeb424b085fe 803f8baf4f741d2f0465c46c33f285886b97a071 Olivier Chafik olivier.chafik@gmail.com 2025-05-31T08:26:10-07:00 GitHub noreply@github.com 2025-05-31T08:26:10-07:00 server: allow unclosed thinking tags (#13931)
803f8baf4f741d2f0465c46c33f285886b97a071 3600cc2886956fc0a07ef6ad2f4128ccfdbc8c6f Georgi Gerganov ggerganov@gmail.com 2025-05-31T15:58:33+03:00 GitHub noreply@github.com 2025-05-31T15:58:33+03:00 llama : deprecate explicit kv_self defrag/update calls (#13921)
3600cc2886956fc0a07ef6ad2f4128ccfdbc8c6f c7e0a2054b908c28bf93bb18d4b63ccbff2c4127 Georgi Gerganov ggerganov@gmail.com 2025-05-31T15:57:44+03:00 GitHub noreply@github.com 2025-05-31T15:57:44+03:00 llama : use n_swa + n_ubatch cells for SWA cache (#13833)
c7e0a2054b908c28bf93bb18d4b63ccbff2c4127 3f55f781f1da9241f209648636fa0426fe62a495 igardev 49397134+igardev@users.noreply.github.com 2025-05-31T12:56:08+03:00 GitHub noreply@github.com 2025-05-31T11:56:08+02:00 webui : Replace alert and confirm with custom modals. (#13711)
3f55f781f1da9241f209648636fa0426fe62a495 51fa76f172957c9916e43aeb581f82c533e12394 Georgi Gerganov ggerganov@gmail.com 2025-05-31T12:55:57+03:00 GitHub noreply@github.com 2025-05-31T12:55:57+03:00 llama : auto-batch preparation (#13845)
51fa76f172957c9916e43aeb581f82c533e12394 12d0188c0dc6146ffde6d277a93f232ccbe699f8 Xuan-Son Nguyen son@huggingface.co 2025-05-31T10:14:29+02:00 GitHub noreply@github.com 2025-05-31T10:14:29+02:00 mtmd : drop `_shared` from `libmtmd` name, merge helpers into libmtmd (⚠️ breaking change) (#13917)
12d0188c0dc6146ffde6d277a93f232ccbe699f8 eb3949938e82a128855bc0676220bb2ce6e4228d Georgi Gerganov ggerganov@gmail.com 2025-05-31T10:24:04+03:00 GitHub noreply@github.com 2025-05-31T10:24:04+03:00 kv-cache : refactor + add llama_memory_state_i (#13746)
eb3949938e82a128855bc0676220bb2ce6e4228d e562eece7cb476276bfc4cbb18deb7c0369b2233 Shawn yang 137684499+Yangxiaoz@users.noreply.github.com 2025-05-31T14:48:04+08:00 GitHub noreply@github.com 2025-05-31T08:48:04+02:00 CUDA: add a prop in ggml_cuda_device_infor for distinguish iGPU or dGPU in cuda (#13856) (#13895)
e562eece7cb476276bfc4cbb18deb7c0369b2233 b47ab7b8e9c4f6154eb6abb6234866ab117d64c7 Johannes Gäßler johannesg@5d6.de 2025-05-30T21:22:03+02:00 GitHub noreply@github.com 2025-05-30T21:22:03+02:00 CUDA: fix typo in FlashAttention code (#13926)
b47ab7b8e9c4f6154eb6abb6234866ab117d64c7 dd665cc9d4b378626cde11ea0c4c91f514fdc994 Diego Devesa slarengh@gmail.com 2025-05-30T09:56:19-07:00 GitHub noreply@github.com 2025-05-30T18:56:19+02:00 sched : avoid changing cur_copy when a graph is already allocated (#13922)
dd665cc9d4b378626cde11ea0c4c91f514fdc994 df0c0c7d02f951dc639d48fd536f79200460ac83 Georgi Gerganov ggerganov@gmail.com 2025-05-30T19:38:07+03:00 GitHub noreply@github.com 2025-05-30T19:38:07+03:00 parallel : increase the variability of the prompt lengths (#13927)
df0c0c7d02f951dc639d48fd536f79200460ac83 b49a8ff96b769b8a4c36d89fb783ec0135be582b Diego Devesa slarengh@gmail.com 2025-05-30T07:37:18-07:00 GitHub noreply@github.com 2025-05-30T16:37:18+02:00 cuda : prevent using split buffers with 3d/4d matrices (#13919)
b49a8ff96b769b8a4c36d89fb783ec0135be582b 53f925074de02c5304b00c14b4d6d8910c58667d Akarshan Biswas akarshan@menlo.ai 2025-05-30T19:40:57+05:30 GitHub noreply@github.com 2025-05-30T19:40:57+05:30 SYCL: Add mrope kernel (#13755)
53f925074de02c5304b00c14b4d6d8910c58667d db38704f0133be7832123495fa8fc2601ea999d4 Georgi Gerganov ggerganov@gmail.com 2025-05-30T16:25:45+03:00 GitHub noreply@github.com 2025-05-30T16:25:45+03:00 sync : vendor (#13901)
db38704f0133be7832123495fa8fc2601ea999d4 07e4351ce663a7802b31f92fd7bc0e555c2044b6 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-30T14:50:43+02:00 GitHub noreply@github.com 2025-05-30T14:50:43+02:00 convert : fix rwkv bos/eos token (#13844)
07e4351ce663a7802b31f92fd7bc0e555c2044b6 291f2b6913c7ef8350dbf0e77da38f7af131a08e Xuan-Son Nguyen son@huggingface.co 2025-05-30T12:24:37+02:00 GitHub noreply@github.com 2025-05-30T12:24:37+02:00 convert : allow partial update to the chkhsh pre-tokenizer list (#13847)
291f2b6913c7ef8350dbf0e77da38f7af131a08e 2c90da4c7ec694797f524042aaafbb047a7e65ff Đinh Trọng Huy 77562200+huydt84@users.noreply.github.com 2025-05-30T18:56:02+09:00 GitHub noreply@github.com 2025-05-30T11:56:02+02:00 llama : add support for DistilBert (#13907)
2c90da4c7ec694797f524042aaafbb047a7e65ff ec9e0301fef6476df83e94842c3b625501c95566 zhangkaihuo zhangkaihuo@gmail.com 2025-05-30T16:31:48+08:00 GitHub noreply@github.com 2025-05-30T10:31:48+02:00 llama : use llm_build_granite for minicpm (#13911)
ec9e0301fef6476df83e94842c3b625501c95566 e83ba3e460651b20a594e9f2f0f0bffb998d3ce1 Christian Kastner ckk@kvr.at 2025-05-30T01:28:54+02:00 GitHub noreply@github.com 2025-05-30T01:28:54+02:00 cmake: Guard GGML_CPU_ALL_VARIANTS by architecture (#13890)
e83ba3e460651b20a594e9f2f0f0bffb998d3ce1 2b131621e60d8ec2cc961201beb6773ab37b6b69 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-29T21:42:31+02:00 GitHub noreply@github.com 2025-05-29T21:42:31+02:00 llama : add support for jina-reranker-v2 (#13900)
2b131621e60d8ec2cc961201beb6773ab37b6b69 54a2c7a8cd8a32b44e3a98c2999b0f5c9114be5c Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-29T15:36:05+02:00 GitHub noreply@github.com 2025-05-29T15:36:05+02:00 gguf-py : add support for sub_type (in arrays) in GGUFWriter add_key_value method (#13561)
54a2c7a8cd8a32b44e3a98c2999b0f5c9114be5c 21fcc21ad5e5de2daa5da8d08dbbcc86b8d815d7 Yibo Cai yibo.cai@arm.com 2025-05-29T19:39:20+08:00 GitHub noreply@github.com 2025-05-29T14:39:20+03:00 arm64: optimize q4_k_q8_k kernel with i8mm (#13886)
21fcc21ad5e5de2daa5da8d08dbbcc86b8d815d7 dd8ba93416f492c168f7e20f0b1434c08ebeaeb5 Christian Kastner ckk@kvr.at 2025-05-29T12:50:25+02:00 GitHub noreply@github.com 2025-05-29T12:50:25+02:00 cmake: Factor out CPU architecture detection (#13883)
dd8ba93416f492c168f7e20f0b1434c08ebeaeb5 66c92061f5c34d2f9a630b7b50cca5ce3ebb4b16 Vineel Abhinav 131174187+vineelabhinav@users.noreply.github.com 2025-05-29T14:48:43+05:30 GitHub noreply@github.com 2025-05-29T12:18:43+03:00 ggml: aarch64: Implement SVE F32 kernels for Mamba Sequential Scan Algorithm (#13882)
66c92061f5c34d2f9a630b7b50cca5ce3ebb4b16 5ca82fc1d7f8cb27f3cbb3019e944a80d5219828 Georgi Gerganov ggerganov@gmail.com 2025-05-29T12:17:16+03:00 GitHub noreply@github.com 2025-05-29T12:17:16+03:00 tests : remove json.hpp from a test (#13880)
5ca82fc1d7f8cb27f3cbb3019e944a80d5219828 6385b843a8dc8e15b8362196039720c58dd79fa2 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-29T10:00:57+02:00 GitHub noreply@github.com 2025-05-29T10:00:57+02:00 convert : workaround for AutoConfig dummy labels (#13881)
6385b843a8dc8e15b8362196039720c58dd79fa2 1b8fb8152d0f3357a9c1d9b4c02f6cc5b9cb7232 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-29T08:15:01+02:00 GitHub noreply@github.com 2025-05-29T08:15:01+02:00 llama : add RobertaForSequenceClassification reranker support (#13875)
1b8fb8152d0f3357a9c1d9b4c02f6cc5b9cb7232 53ae30640e131082d8d19bd80485b47c4553d551 Vineel Abhinav 131174187+vineelabhinav@users.noreply.github.com 2025-05-29T11:31:33+05:30 GitHub noreply@github.com 2025-05-29T09:01:33+03:00 ggml: aarch64: Implement SVE F32 kernels for vector functions (#13843)
53ae30640e131082d8d19bd80485b47c4553d551 763d06edb7dd5094ea58bad1d81e2e8d35033e34 Beinsezii 39478211+Beinsezii@users.noreply.github.com 2025-05-28T14:50:20-07:00 GitHub noreply@github.com 2025-05-28T23:50:20+02:00 gguf-py : fix SafetensorRemote return on undefined size (< 0) (#13841)
763d06edb7dd5094ea58bad1d81e2e8d35033e34 10961339b26bd2eff01d5479e8879f435da261b7 Xuan-Son Nguyen son@huggingface.co 2025-05-28T22:35:31+02:00 GitHub noreply@github.com 2025-05-28T22:35:31+02:00 llama : fix KV shift for qwen2vl (#13870)
10961339b26bd2eff01d5479e8879f435da261b7 d98f2a35fcf4a8d3e660ad48cd19e2a1f3d5b2ef Xuan-Son Nguyen son@huggingface.co 2025-05-28T22:35:22+02:00 GitHub noreply@github.com 2025-05-28T22:35:22+02:00 mtmd : move helpers to dedicated library (⚠️ breaking change) (#13866)
d98f2a35fcf4a8d3e660ad48cd19e2a1f3d5b2ef e0e3aa231d899720c2864d09cdb89d4c400eeb55 bandoti 141645996+bandoti@users.noreply.github.com 2025-05-28T15:46:47-03:00 GitHub noreply@github.com 2025-05-28T15:46:47-03:00 ci: disable LLAMA_CURL for Linux cross-builds (#13871)
e0e3aa231d899720c2864d09cdb89d4c400eeb55 aa6dff05be25709bb218bf648951d690029c4b19 Đinh Trọng Huy 77562200+huydt84@users.noreply.github.com 2025-05-29T02:01:58+09:00 GitHub noreply@github.com 2025-05-28T19:01:58+02:00 llama : add support for BertForSequenceClassification reranker (#13858)
aa6dff05be25709bb218bf648951d690029c4b19 c962ae3382a1e759c8517a229549ee53685313a1 Đinh Trọng Huy 77562200+huydt84@users.noreply.github.com 2025-05-28T23:34:18+09:00 GitHub noreply@github.com 2025-05-28T16:34:18+02:00 convert: small addition to support LlamaModel (#13838)
c962ae3382a1e759c8517a229549ee53685313a1 a3938fb53d0b5183db4f5a6db21fd9122a0e4780 Sky Iflyinskyin2013@gmail.com 2025-05-28T22:33:54+08:00 GitHub noreply@github.com 2025-05-28T16:33:54+02:00 server: fix remove 'image_url'/'input_audio' json-object effectlly for 'llama_params' in multimodal-model-mode (#13853)
a3938fb53d0b5183db4f5a6db21fd9122a0e4780 f7873fc698c09047e2873630ab7e7730a0bfb224 Xuan-Son Nguyen son@huggingface.co 2025-05-28T16:12:35+02:00 GitHub noreply@github.com 2025-05-28T16:12:35+02:00 convert : fix qwen omni conversion (#13859)
f7873fc698c09047e2873630ab7e7730a0bfb224 a68247439bd6fb756cc93ad2817e55a02aa0b100 Alex Fanthome xfanth@gmail.com 2025-05-28T14:49:28+01:00 GitHub noreply@github.com 2025-05-28T15:49:28+02:00 tests : change umlaut test (#11600)
a68247439bd6fb756cc93ad2817e55a02aa0b100 26b79b6cb3e7840ff15729350e95907e19f9f480 Johannes Gäßler johannesg@5d6.de 2025-05-28T13:33:37+02:00 GitHub noreply@github.com 2025-05-28T13:33:37+02:00 CUDA: fix FA tg at long context for CC >= 8.9 (#13852)
26b79b6cb3e7840ff15729350e95907e19f9f480 1e8659e65ad0f640674d2785d02b556ab938728c Xuan-Son Nguyen son@huggingface.co 2025-05-28T10:05:54+02:00 GitHub noreply@github.com 2025-05-28T10:05:54+02:00 convert : fix tensor naming conflict for llama 4 vision (#13836)
1e8659e65ad0f640674d2785d02b556ab938728c a3c30846e410c91c11d7bf80978795a03bb03dee leo-pony nengjunma@outlook.com 2025-05-28T11:54:20+08:00 GitHub noreply@github.com 2025-05-28T11:54:20+08:00 CANN: Add SOC TYPE printing in cmake configuration (#13837)
a3c30846e410c91c11d7bf80978795a03bb03dee 1701d4c54f93c0d203bf92ea7202a94b037c2338 lhez quic_lih@quicinc.com 2025-05-27T12:56:08-07:00 GitHub noreply@github.com 2025-05-27T12:56:08-07:00 opencl: add new ops - `argsort`, `div`, `sub`, `addrows`, `sigmoid`, `group_norm` (#13787)
1701d4c54f93c0d203bf92ea7202a94b037c2338 bef817638780dcbd8c0c80c97b7a4f8e92c8fe74 lhez quic_lih@quicinc.com 2025-05-27T12:53:14-07:00 GitHub noreply@github.com 2025-05-27T12:53:14-07:00 opencl: mark `mul_mat` `f32f32` as supporting non-contiguous tensors (#13790)
bef817638780dcbd8c0c80c97b7a4f8e92c8fe74 34b7c0439ed0f98575cc4689dfecd98991dee8be Jeff Bolz jbolz@nvidia.com 2025-05-27T11:39:07-05:00 GitHub noreply@github.com 2025-05-27T18:39:07+02:00 vulkan: use timestamp queries for GGML_VULKAN_PERF (#13817)
34b7c0439ed0f98575cc4689dfecd98991dee8be f3101a8cc665f73217c752a10a7042889275cfbc Georgi Gerganov ggerganov@gmail.com 2025-05-27T19:08:44+03:00 GitHub noreply@github.com 2025-05-27T19:08:44+03:00 cmake : add llama-cparams.cpp to build (#13832)
f3101a8cc665f73217c752a10a7042889275cfbc 1c49c70d07ef87635daa5e8fdd0b5bfd88493dd3 Akarshan Biswas akarshan@menlo.ai 2025-05-27T20:52:59+05:30 GitHub noreply@github.com 2025-05-27T20:52:59+05:30 SYCL: add gelu_erf kernel (#13749)
1c49c70d07ef87635daa5e8fdd0b5bfd88493dd3 a8ea03d8ad9c984fe6cfafead183ab188f8cbeb0 Georgi Gerganov ggerganov@gmail.com 2025-05-27T18:04:38+03:00 Georgi Gerganov ggerganov@gmail.com 2025-05-27T18:05:33+03:00 sync : ggml
a8ea03d8ad9c984fe6cfafead183ab188f8cbeb0 05f6ac6283a7859a03cd59405504262c85c4bf06 Xuan-Son Nguyen son@huggingface.co 2025-05-27T15:53:55+02:00 GitHub noreply@github.com 2025-05-27T15:53:55+02:00 ggml : add ggml_repeat_4d (#13824)
05f6ac6283a7859a03cd59405504262c85c4bf06 bc583e3c63c04a11d287c108ea9e6a515ead0423 xctan xc-tan@outlook.com 2025-05-27T21:21:36+08:00 GitHub noreply@github.com 2025-05-27T16:21:36+03:00 ggml : riscv: add xtheadvector support (#13720)
bc583e3c63c04a11d287c108ea9e6a515ead0423 72b090da2c50e540143fd312a2f9aa5f151e6136 Xuan-Son Nguyen son@huggingface.co 2025-05-27T14:06:10+02:00 GitHub noreply@github.com 2025-05-27T14:06:10+02:00 mtmd : support Qwen 2.5 Omni (input audio+vision, no audio output) (#13784)
72b090da2c50e540143fd312a2f9aa5f151e6136 7fe03e7446ed4388539d3bcc013ae4e851b8d1e2 bandoti 141645996+bandoti@users.noreply.github.com 2025-05-27T08:52:40-03:00 GitHub noreply@github.com 2025-05-27T08:52:40-03:00 docs: remove link for llama-cli function calling (#13810)
7fe03e7446ed4388539d3bcc013ae4e851b8d1e2 952f3953c1b61cc70e79e536c42ddce6a5ea5ea7 Christian Kastner ckk@kvr.at 2025-05-27T13:18:39+02:00 GitHub noreply@github.com 2025-05-27T13:18:39+02:00 ggml-cpu: x86 feature detection is specific to x86 (#13811)
952f3953c1b61cc70e79e536c42ddce6a5ea5ea7 81713121ee56755ba4a147d1a1e3fa248ec31a66 Diego Devesa slarengh@gmail.com 2025-05-27T04:05:18-07:00 GitHub noreply@github.com 2025-05-27T13:05:18+02:00 ggml : allow CUDA graphs when using pipeline parallelism (#13814)
81713121ee56755ba4a147d1a1e3fa248ec31a66 f9cd68398baf2ba8af4725ca9ed00bef205e6706 Georgi Gerganov ggerganov@gmail.com 2025-05-27T13:49:41+03:00 GitHub noreply@github.com 2025-05-27T13:49:41+03:00 kv-cells : track min/max used cells and per-sequence positions (#13808)
f9cd68398baf2ba8af4725ca9ed00bef205e6706 4f81b33e324b1669b282eb81104e4a1131be7dce Georgi Gerganov ggerganov@gmail.com 2025-05-27T12:07:52+03:00 GitHub noreply@github.com 2025-05-27T12:07:52+03:00 sampling : make sure samplers return at least 1 token (#13822)
4f81b33e324b1669b282eb81104e4a1131be7dce cdf94a18023c92f41808ec874ba577d914674717 Georgi Gerganov ggerganov@gmail.com 2025-05-27T09:40:59+03:00 GitHub noreply@github.com 2025-05-27T09:40:59+03:00 llama : validate seq id batch input (#13809)
cdf94a18023c92f41808ec874ba577d914674717 a26c4cc11ec7c6574e3691e90ecdbd67deeea35b Olivier Chafik olivier.chafik@gmail.com 2025-05-26T14:34:27-07:00 GitHub noreply@github.com 2025-05-26T22:34:27+01:00 server: --offline mode (#13804)
a26c4cc11ec7c6574e3691e90ecdbd67deeea35b 4265a87b59ebfc25f35adbf4db3b608995b0a78a Georgi Gerganov ggerganov@gmail.com 2025-05-26T22:24:01+03:00 GitHub noreply@github.com 2025-05-26T22:24:01+03:00 scripts : add option to compare commits in Debug (#13806)
4265a87b59ebfc25f35adbf4db3b608995b0a78a 6f180b915c9ed9ec0c240b5dcd64644988fb5e82 Georgi Gerganov ggerganov@gmail.com 2025-05-26T22:14:52+03:00 GitHub noreply@github.com 2025-05-26T22:14:52+03:00 cuda : avoid cuGetErrorString (#13791)
6f180b915c9ed9ec0c240b5dcd64644988fb5e82 03f582ae8fccecff225c30a2802461b44761e822 Akarshan Biswas akarshan@menlo.ai 2025-05-26T21:10:36+05:30 GitHub noreply@github.com 2025-05-26T21:10:36+05:30 SYCL: Add non contiguous support in RMS_NORM and NORM kernels (#13611)
03f582ae8fccecff225c30a2802461b44761e822 88c125f2acce0e25e5fc8481ab0681415fc64a10 Olivier Chafik olivier.chafik@gmail.com 2025-05-26T08:03:57-07:00 GitHub noreply@github.com 2025-05-26T16:03:57+01:00 server: fix streaming crashes (#13786)
88c125f2acce0e25e5fc8481ab0681415fc64a10 d74e94c1b3ac02db01e47008e1f8d371029d81ac standby24x7 standby24x7@gmail.com 2025-05-26T23:55:24+09:00 GitHub noreply@github.com 2025-05-26T16:55:24+02:00 examples/training: Fix file name in README (#13803)
d74e94c1b3ac02db01e47008e1f8d371029d81ac f13847cfb560d92492b480cca2c5d6aa9473cde3 Olivier Chafik olivier.chafik@gmail.com 2025-05-26T06:56:49-07:00 GitHub noreply@github.com 2025-05-26T14:56:49+01:00 `server`: fix format of streamed tool call deltas (diff name, fix id location) (#13800)
f13847cfb560d92492b480cca2c5d6aa9473cde3 79c137f77677b3c8ee3c60a7da033721b938399a Olivier Chafik olivier.chafik@gmail.com 2025-05-26T06:16:37-07:00 GitHub noreply@github.com 2025-05-26T14:16:37+01:00 server: fix regression on streamed non-chat completion w/ stops (#13785)
79c137f77677b3c8ee3c60a7da033721b938399a 22229314fc46b2f741bb21b12cde71f6c6a60b52 Georgi Gerganov ggerganov@gmail.com 2025-05-26T14:03:54+03:00 GitHub noreply@github.com 2025-05-26T14:03:54+03:00 examples : allow extracting embeddings from decoder contexts (#13797)
22229314fc46b2f741bb21b12cde71f6c6a60b52 9012eb9b454a82eaa4cd77ae904c0ea391e4db42 Georgi Gerganov ggerganov@gmail.com 2025-05-26T12:57:50+03:00 GitHub noreply@github.com 2025-05-26T12:57:50+03:00 llama : clarify deprecation message (#13794)
9012eb9b454a82eaa4cd77ae904c0ea391e4db42 fef693dc6b959a8e8ba11558fbeaad0b264dd457 Romain Biessy romain.biessy@codeplay.com 2025-05-26T10:28:53+02:00 GitHub noreply@github.com 2025-05-26T10:28:53+02:00 sycl: Add more debug prints (#13640)
fef693dc6b959a8e8ba11558fbeaad0b264dd457 2d38b6e4004fb1c341723e657fb1e71a4d3fb473 Jeff Bolz jbolz@nvidia.com 2025-05-25T23:02:07-05:00 GitHub noreply@github.com 2025-05-26T06:02:07+02:00 vulkan: mark IM2COL as supporting non-contig (#13783)
2d38b6e4004fb1c341723e657fb1e71a4d3fb473 e121edc4324a640be11b7e567edd39b721b0f8e4 Bizhao Shi 37729561+shibizhao@users.noreply.github.com 2025-05-26T10:20:18+08:00 GitHub noreply@github.com 2025-05-26T10:20:18+08:00 CANN: Add the basic supports of Flash Attention kernel (#13627)
e121edc4324a640be11b7e567edd39b721b0f8e4 2f099b510f460374acd52742b494595e3e3442d3 Olivier Chafik olivier.chafik@gmail.com 2025-05-26T00:30:51+01:00 GitHub noreply@github.com 2025-05-26T00:30:51+01:00 `server`: add `--reasoning-budget 0` to disable thinking (incl. qwen3 w/ enable_thinking:false) (#13771)
2f099b510f460374acd52742b494595e3e3442d3 aa50ba462f63759e1731227f20f5009cfc2a0f16 Xuan-Son Nguyen son@huggingface.co 2025-05-25T19:02:18+02:00 GitHub noreply@github.com 2025-05-25T18:02:18+01:00 webui : bump max upload file size to 500MB (#13779)
aa50ba462f63759e1731227f20f5009cfc2a0f16 de2ef53a4b2c0d703749a309d19fe68fd8f1b9ac Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-25T16:22:29+02:00 GitHub noreply@github.com 2025-05-25T16:22:29+02:00 tests : improve UGM tokenizer test coverage (#13773)
de2ef53a4b2c0d703749a309d19fe68fd8f1b9ac c508256db2de2b032e19c8ed833f4683c827c9a1 Georgi Gerganov ggerganov@gmail.com 2025-05-25T16:34:36+03:00 GitHub noreply@github.com 2025-05-25T16:34:36+03:00 kv-cache : rework kv_cell (#13706)
c508256db2de2b032e19c8ed833f4683c827c9a1 40aaa8a403df5dcfb515eb2fd7a817fd99779526 Percy Piper piper.percy@googlemail.com 2025-05-25T13:35:53+01:00 GitHub noreply@github.com 2025-05-25T15:35:53+03:00 rpc : Fix build on OpenBSD (#13541)
40aaa8a403df5dcfb515eb2fd7a817fd99779526 a08c1d2845dc279d58d826ef3c3ecad97cbbcef7 Xuan-Son Nguyen son@huggingface.co 2025-05-25T14:06:32+02:00 GitHub noreply@github.com 2025-05-25T14:06:32+02:00 mtmd : add support for Qwen2-Audio and SeaLLM-Audio (#13760)
a08c1d2845dc279d58d826ef3c3ecad97cbbcef7 d785f9c1fd1a1929c6d0e2a0b12cae5db867908b ddpasa 112642920+ddpasa@users.noreply.github.com 2025-05-25T14:04:49+02:00 GitHub noreply@github.com 2025-05-25T14:04:49+02:00 docs : add Moondream2 pre-quantized link (#13745)
d785f9c1fd1a1929c6d0e2a0b12cae5db867908b 4032ca406632212b075e3c61c2a4476128321410 Olivier Chafik olivier.chafik@gmail.com 2025-05-25T10:45:49+01:00 GitHub noreply@github.com 2025-05-25T10:45:49+01:00 server: fix/test add_generation_prompt (#13770)
4032ca406632212b075e3c61c2a4476128321410 515fdbf7ed839dfe6a24aeb6225936609a7f6d6d Piotr Jasiukajtis estibi@me.com 2025-05-25T10:29:43+02:00 GitHub noreply@github.com 2025-05-25T10:29:43+02:00 llama : add support for Qwen3 MoE tied word embeddings (#13768)
515fdbf7ed839dfe6a24aeb6225936609a7f6d6d f5cd27b71da3ac375a04a41643d14fc779a8057b Akarshan Biswas akarshan@menlo.ai 2025-05-25T12:38:37+05:30 GitHub noreply@github.com 2025-05-25T10:08:37+03:00 SYCL: revert "sycl: simplify bin_bcast_kernel (#13383)" (#13752)
f5cd27b71da3ac375a04a41643d14fc779a8057b a2d02d5793fd9af7a7224773456501691b95fd02 Olivier Chafik olivier.chafik@gmail.com 2025-05-25T01:48:08+01:00 GitHub noreply@github.com 2025-05-25T01:48:08+01:00 `server`: streaming of tool calls and thoughts when `--jinja` is on (#12379)
a2d02d5793fd9af7a7224773456501691b95fd02 17fc817b58942569c43aa63299cedde217b61f68 Diego Devesa slarengh@gmail.com 2025-05-24T15:55:16-07:00 GitHub noreply@github.com 2025-05-25T00:55:16+02:00 releases : bundle llvm omp library in windows release (#13763)
17fc817b58942569c43aa63299cedde217b61f68 2bd1b30f6979235ec67b95c183b9b77baa7ab9ce Diego Devesa slarengh@gmail.com 2025-05-24T13:27:03-07:00 GitHub noreply@github.com 2025-05-24T22:27:03+02:00 releases : enable openmp in windows cpu backend build (#13756)
2bd1b30f6979235ec67b95c183b9b77baa7ab9ce 259469c4b57c1a32606353bcac52ba683424a990 Diego Devesa slarengh@gmail.com 2025-05-24T13:26:47-07:00 GitHub noreply@github.com 2025-05-24T22:26:47+02:00 ggml-cpu : set openmp wait time if not set (#13758)
259469c4b57c1a32606353bcac52ba683424a990 4c32832c59e97d96c19349fdc92763e8949fda83 0cc4m picard12@live.de 2025-05-24T16:49:12+02:00 GitHub noreply@github.com 2025-05-24T16:49:12+02:00 Move GLM4 f32 attention fix to the correct function (#13750)
4c32832c59e97d96c19349fdc92763e8949fda83 c3a2624339187e89c4f65fd72a5fe7103968b5ad Xuan-Son Nguyen son@huggingface.co 2025-05-24T13:06:47+02:00 GitHub noreply@github.com 2025-05-24T13:06:47+02:00 ggml : add ggml_gelu_erf() CUDA kernel (#13719)
c3a2624339187e89c4f65fd72a5fe7103968b5ad ffd0eae60b7642e942c8245b89642527e36099e6 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-24T12:29:09+02:00 GitHub noreply@github.com 2025-05-24T12:29:09+02:00 vocab : fix ugm tokenizer precision (#13743)
ffd0eae60b7642e942c8245b89642527e36099e6 b775345d788ac16260e7eef49e11fe57ee5677f7 Johannes Gäßler johannesg@5d6.de 2025-05-24T11:46:19+02:00 GitHub noreply@github.com 2025-05-24T11:46:19+02:00 CUDA: fix race condition in FA vector kernels (#13742)
b775345d788ac16260e7eef49e11fe57ee5677f7 a70a8a69c2ad3de8d5525ad2b185b098011be2e8 Diego Devesa slarengh@gmail.com 2025-05-23T13:14:00-07:00 GitHub noreply@github.com 2025-05-23T23:14:00+03:00 ci : enable winget package updates (#13734)
a70a8a69c2ad3de8d5525ad2b185b098011be2e8 d13d0f6135803822ec1cd7e3efb49360b88a1bdf Diego Devesa slarengh@gmail.com 2025-05-23T13:09:38-07:00 GitHub noreply@github.com 2025-05-23T22:09:38+02:00 ci : add winget package updater (#13732)
d13d0f6135803822ec1cd7e3efb49360b88a1bdf 8a2afb7520bbc8f9fa1bbe314d5f2807eb0116b2 Georgi Gerganov ggerganov@gmail.com 2025-05-23T20:16:13+03:00 GitHub noreply@github.com 2025-05-23T20:16:13+03:00 hparams : initialize arrays (#13728)
8a2afb7520bbc8f9fa1bbe314d5f2807eb0116b2 9ecf3e66a38f20e41d221f62f05b17f70d040839 Xuan-Son Nguyen son@huggingface.co 2025-05-23T17:07:04+02:00 GitHub noreply@github.com 2025-05-23T17:07:04+02:00 llama : allow custom list of swa_layers (#13726)
9ecf3e66a38f20e41d221f62f05b17f70d040839 faaaff5f947064d13ef8b98659d81a1384c3e57b Xuan-Son Nguyen son@huggingface.co 2025-05-23T11:03:47+02:00 GitHub noreply@github.com 2025-05-23T11:03:47+02:00 server : support audio input (#13714)
faaaff5f947064d13ef8b98659d81a1384c3e57b e16c4731c7a6bfa8f61b5b39c77245a37e7fc232 Chenguang Li 757486878@qq.com 2025-05-23T16:47:53+08:00 GitHub noreply@github.com 2025-05-23T16:47:53+08:00 CANN: Support MUL_MAT_ID for q8_0 and q4_0 (#13705)
e16c4731c7a6bfa8f61b5b39c77245a37e7fc232 1dcd01960c33f52afb782b3d850fc2149a08cc6b Xuan-Son Nguyen son@huggingface.co 2025-05-23T08:12:48+02:00 GitHub noreply@github.com 2025-05-23T08:12:48+02:00 ggml : fix the order of ggml_unary_op (#13718)
1dcd01960c33f52afb782b3d850fc2149a08cc6b c10ed6cbcc3904b7d6bbcd137589eebd899aa38f Jeff Bolz jbolz@nvidia.com 2025-05-23T00:45:02-04:00 GitHub noreply@github.com 2025-05-23T06:45:02+02:00 vulkan: support CPY from any type to itself (#13695)
c10ed6cbcc3904b7d6bbcd137589eebd899aa38f a127ff17800452075bb323277e3b9f8db8612ced Jeff Bolz jbolz@nvidia.com 2025-05-23T00:33:45-04:00 GitHub noreply@github.com 2025-05-23T06:33:45+02:00 vulkan: Disable coopmat/coopmat2/bfloat extensions if glslc doesn't support it (#13696)
a127ff17800452075bb323277e3b9f8db8612ced 3079e9ac8e04ef6eddeb0c164d72edb6b6fd2df5 Judd 4046440+foldl@users.noreply.github.com 2025-05-23T12:33:08+08:00 GitHub noreply@github.com 2025-05-23T06:33:08+02:00 use LOG_WARN to replace `std::cerr` (#13657)
3079e9ac8e04ef6eddeb0c164d72edb6b6fd2df5 8a1d206f1d2b4e45918b589f3165b4be232f7ba8 Diego Devesa slarengh@gmail.com 2025-05-22T15:21:37-07:00 GitHub noreply@github.com 2025-05-23T00:21:37+02:00 release : fix windows hip release (#13707)
8a1d206f1d2b4e45918b589f3165b4be232f7ba8 797990c4bca0dca5be295c63e3fb2800dc0a69c2 Georgi Gerganov ggerganov@gmail.com 2025-05-22T22:21:07+03:00 GitHub noreply@github.com 2025-05-22T22:21:07+03:00 tts : fix n_ubatch + make WavTokenizer cache-less (#13713)
797990c4bca0dca5be295c63e3fb2800dc0a69c2 ab86335760ebb441574eb47f886fa1ee302e2131 Xuan-Son Nguyen son@huggingface.co 2025-05-22T20:42:48+02:00 GitHub noreply@github.com 2025-05-22T20:42:48+02:00 mtmd : add ultravox audio input (#13623)
ab86335760ebb441574eb47f886fa1ee302e2131 cc74d5be990e37f201591fd868a92e64abdbf902 Aaron Teo aaron.teo1@ibm.com 2025-05-23T02:31:29+08:00 GitHub noreply@github.com 2025-05-22T21:31:29+03:00 common: Include torch package for s390x (#13699)
cc74d5be990e37f201591fd868a92e64abdbf902 5be24af73d77ea23d399726f1d2a01a70ee86331 Georgi Gerganov ggerganov@gmail.com 2025-05-22T16:33:39+03:00 GitHub noreply@github.com 2025-05-22T16:33:39+03:00 server : pad small embedding batches (#13692)
5be24af73d77ea23d399726f1d2a01a70ee86331 d394a9aedc50a13b7f6373416f7c1ccabfe79c32 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-22T14:25:05+02:00 GitHub noreply@github.com 2025-05-22T14:25:05+02:00 gguf-py : correct charsmap parameter typing (#13701)
d394a9aedc50a13b7f6373416f7c1ccabfe79c32 6b56a64690a318fcabcd7739ac7e314d44785ea8 Nicolò Scipione nicolo.scipione@codeplay.com 2025-05-22T13:54:43+02:00 GitHub noreply@github.com 2025-05-22T12:54:43+01:00 sycl : Remove waits from function calls (#13702)
6b56a64690a318fcabcd7739ac7e314d44785ea8 a4e8912dfd4604be1e39bc86ba4c0b02969967ef Ewan Crawford ewan@codeplay.com 2025-05-22T09:24:09+01:00 GitHub noreply@github.com 2025-05-22T16:24:09+08:00 SYCL: Avoid using with SYCL-Graph for unsupported nodes (#13587)
a4e8912dfd4604be1e39bc86ba4c0b02969967ef edbf42edfdabb9cea72ae12137570cf48f5d8076 Henry Linjamäki henry.mikael.linjamaki@intel.com 2025-05-22T02:21:45+03:00 GitHub noreply@github.com 2025-05-21T16:21:45-07:00 opencl: Add support for multiple devices (#12622)
edbf42edfdabb9cea72ae12137570cf48f5d8076 d643bb2c798df9c2cd61067d2692b1cd417df402 Henry Linjamäki henry.mikael.linjamaki@intel.com 2025-05-21T23:21:17+03:00 GitHub noreply@github.com 2025-05-21T13:21:17-07:00 opencl: fix couple crashes (#12795)
d643bb2c798df9c2cd61067d2692b1cd417df402 8e186ef0e764c7a620e402d1f76ebad60bf31c49 Diego Devesa slarengh@gmail.com 2025-05-21T13:09:57-07:00 GitHub noreply@github.com 2025-05-21T22:09:57+02:00 releases : build CPU backend separately (windows) (#13642)
8e186ef0e764c7a620e402d1f76ebad60bf31c49 5fbfe384d4659f81c47a477eb8ee97692c7ffef9 Georgi Gerganov ggerganov@gmail.com 2025-05-21T20:00:49+03:00 GitHub noreply@github.com 2025-05-21T20:00:49+03:00 hparams : support models for which all layers use SWA (#13682)
5fbfe384d4659f81c47a477eb8ee97692c7ffef9 c76532e7ba128bb097bf6836bf0f5592e1b56b76 Georgi Gerganov ggerganov@gmail.com 2025-05-21T19:46:56+03:00 GitHub noreply@github.com 2025-05-21T19:46:56+03:00 server : improve error reporting (#13680)
c76532e7ba128bb097bf6836bf0f5592e1b56b76 2aa777d86d3f7bb80b93e226f1c25e47825f6a83 antichristHater 142441588+antichristHater@users.noreply.github.com 2025-05-21T19:40:35+03:00 GitHub noreply@github.com 2025-05-21T18:40:35+02:00 convert : add qwen2vl support for unsloth merges (#13686)
2aa777d86d3f7bb80b93e226f1c25e47825f6a83 eb0f5c28d37126baa756117d5bdaadc62e03344e Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-21T16:57:38+02:00 GitHub noreply@github.com 2025-05-21T16:57:38+02:00 examples : switch retrieval to llama_encode (#13685)
eb0f5c28d37126baa756117d5bdaadc62e03344e cf4cb59e64d72a1b4c781f71a74de5756a4e2376 Emmanuel Ferdman emmanuelferdman@gmail.com 2025-05-21T17:33:54+03:00 GitHub noreply@github.com 2025-05-21T16:33:54+02:00 gguf-py : display the invalid gguf type (#13687)
cf4cb59e64d72a1b4c781f71a74de5756a4e2376 0d5c74216170ef97e5e7511563837263f2d1a496 Xuan-Son Nguyen son@huggingface.co 2025-05-21T16:26:33+02:00 GitHub noreply@github.com 2025-05-21T16:26:33+02:00 ggml : add ggml_gelu_erf() (#13667)
0d5c74216170ef97e5e7511563837263f2d1a496 42158ae2e8ead667a83f07247321ce85f32ace66 Robin Davidsson 40024429+R-Dson@users.noreply.github.com 2025-05-21T15:15:27+02:00 GitHub noreply@github.com 2025-05-21T15:15:27+02:00 server : Add the endpoints /api/tags and /api/chat (#13659)
42158ae2e8ead667a83f07247321ce85f32ace66 797f2ac0625b22edeff03cc30e0f988da6b6b068 Dorin-Andrei Geman doringeman@gmail.com 2025-05-21T16:07:57+03:00 GitHub noreply@github.com 2025-05-21T15:07:57+02:00 server : fix first message identification (#13634)
797f2ac0625b22edeff03cc30e0f988da6b6b068 b44890df2e4fad0ece1d5366dcbc8bedae23b658 Georgi Gerganov ggerganov@gmail.com 2025-05-21T15:11:13+03:00 GitHub noreply@github.com 2025-05-21T15:11:13+03:00 kv-cache : simplify the interface (#13660)
b44890df2e4fad0ece1d5366dcbc8bedae23b658 33983057d0f578aca74ba15eccc3de9c267a5ff6 Georgi Gerganov ggerganov@gmail.com 2025-05-21T13:09:21+03:00 GitHub noreply@github.com 2025-05-21T13:09:21+03:00 model : disable SWA for Phi models (#13676)
33983057d0f578aca74ba15eccc3de9c267a5ff6 fb1cab201c6c4cd36731f100df74eece2f4706fa R0CKSTAR yeahdongcn@gmail.com 2025-05-21T09:58:49+08:00 GitHub noreply@github.com 2025-05-21T09:58:49+08:00 musa: Upgrade MUSA SDK version to rc4.0.1 and use mudnn::Unary::IDENTITY op to accelerate D2D memory copy (#13647)
fb1cab201c6c4cd36731f100df74eece2f4706fa b7a17463ec190aeee7b9077c606c910fb4688b84 Eve 139727413+netrunnereve@users.noreply.github.com 2025-05-20T21:35:16Z GitHub noreply@github.com 2025-05-20T21:35:16Z vulkan: fix warnings (#13626)
b7a17463ec190aeee7b9077c606c910fb4688b84 be0239693c1530a18496086331fc18d8a9adbad1 l3utterfly gc.pthzfoldr@gmail.com 2025-05-21T00:55:30+08:00 GitHub noreply@github.com 2025-05-20T18:55:30+02:00 mtmd-helper : bug fix to token batching in mtmd (#13650)
be0239693c1530a18496086331fc18d8a9adbad1 a4090d1174aed22dde5cacce2a4c27656b987a2f Georgi Gerganov ggerganov@gmail.com 2025-05-20T19:21:04+03:00 GitHub noreply@github.com 2025-05-20T19:21:04+03:00 model : fix llama4 graph (#13663)
a4090d1174aed22dde5cacce2a4c27656b987a2f b69f1647f9953cb3773266d2c83a92fd0e7e6d66 Georgi Gerganov ggerganov@gmail.com 2025-05-20T16:13:16+03:00 GitHub noreply@github.com 2025-05-20T16:13:16+03:00 llama : remove llama_kv_cache_view API + remove deprecated (#13653)
b69f1647f9953cb3773266d2c83a92fd0e7e6d66 759e37b0d89bc4bd1bce860dc5f3c3052e08575c Johannes Gäßler johannesg@5d6.de 2025-05-20T14:45:07+02:00 GitHub noreply@github.com 2025-05-20T14:45:07+02:00 CUDA: skip fully masked-out KV in FA vec kernel (#13584)
759e37b0d89bc4bd1bce860dc5f3c3052e08575c 4245e622e0cc3af1ca3056104e465dc4d303bd7d Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-20T12:03:17+02:00 GitHub noreply@github.com 2025-05-20T12:03:17+02:00 tests : avoid github urls due to throttling (#13654)
4245e622e0cc3af1ca3056104e465dc4d303bd7d c9c64dee572c5eedf073a6c6eb5d92d8283f7639 Svetlozar Georgiev 55534064+sgeor255@users.noreply.github.com 2025-05-20T10:34:15+01:00 GitHub noreply@github.com 2025-05-20T11:34:15+02:00 sycl: disable reorder for sycl mulmat (#13536)
c9c64dee572c5eedf073a6c6eb5d92d8283f7639 c00a2634bec49805c6b31438b1a6006a2bd793cb 0cc4m picard12@live.de 2025-05-20T10:11:56+02:00 GitHub noreply@github.com 2025-05-20T10:11:56+02:00 Set GLM4 blk.*.attn_output.weight, kqv_out-* matmul to GGML_PREC_F32 to fix infinity values in output (#13639)
c00a2634bec49805c6b31438b1a6006a2bd793cb e298d2fbd082a52c0f6ed02729f94e9bf630cf17 Georgi Gerganov ggerganov@gmail.com 2025-05-20T10:41:40+03:00 GitHub noreply@github.com 2025-05-20T10:41:40+03:00 metal : fix typo in FA kernel comments (#13651)
e298d2fbd082a52c0f6ed02729f94e9bf630cf17 f0adb80bf7c2c0d80abb04f4533b5513622d9964 Georgi Gerganov ggerganov@gmail.com 2025-05-20T08:05:46+03:00 GitHub noreply@github.com 2025-05-20T08:05:46+03:00 kv-cache : add SWA support (#13194)
f0adb80bf7c2c0d80abb04f4533b5513622d9964 f7c9429c85748cde9599499601ba48d0057722e6 Xinpeng Dou 15529241576@163.com 2025-05-20T11:43:43+08:00 GitHub noreply@github.com 2025-05-20T11:43:43+08:00 CANN: Update CANN model support (#13162)
f7c9429c85748cde9599499601ba48d0057722e6 1dfbf2cf3a9f15193dd893396d07762bbd2c4785 Nicolò Scipione nicolo.scipione@codeplay.com 2025-05-20T02:54:43+02:00 GitHub noreply@github.com 2025-05-20T08:54:43+08:00 sycl : Overcoming workaround for mmap() allocation on Windows (#13482)
1dfbf2cf3a9f15193dd893396d07762bbd2c4785 8960efd0a65e5a4830a14f6937c10703534f2569 psocolovsky 50770545+psocolovsky@users.noreply.github.com 2025-05-19T21:17:36+02:00 GitHub noreply@github.com 2025-05-19T21:17:36+02:00 common : add load_progress_callback (#13617)
8960efd0a65e5a4830a14f6937c10703534f2569 725f23f1f3f0d3adf49f95d8dfa6e7c74adff149 0cc4m picard12@live.de 2025-05-19T17:54:08+02:00 GitHub noreply@github.com 2025-05-19T17:54:08+02:00 Vulkan: Add f32 accumulator support to quantized mul mat to fix GLM4 32B incoherence (#13607)
725f23f1f3f0d3adf49f95d8dfa6e7c74adff149 92ecdcc06a4c405a415bcaa0cb772bc560aa23b1 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-05-19T14:38:20+01:00 GitHub noreply@github.com 2025-05-19T14:38:20+01:00 sycl : backend documentation review (#13544)
92ecdcc06a4c405a415bcaa0cb772bc560aa23b1 f71f40a2847d4c9f57b86cd206e0a27b2bfb6d1c Xuan-Son Nguyen son@huggingface.co 2025-05-19T13:04:14+02:00 GitHub noreply@github.com 2025-05-19T13:04:14+02:00 mtmd : add vision support for llama 4 (#13282)
f71f40a2847d4c9f57b86cd206e0a27b2bfb6d1c d30cb5a7fa17362c47e94a023276f169916e0d03 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-05-19T11:46:09+01:00 GitHub noreply@github.com 2025-05-19T11:46:09+01:00 ci : upgraded oneAPI version in SYCL workflows and dockerfile (#13532)
d30cb5a7fa17362c47e94a023276f169916e0d03 6c35981a643d4f74a286268b62f65bfa156af2e6 Georgi Gerganov ggerganov@gmail.com 2025-05-19T12:50:29+03:00 Georgi Gerganov ggerganov@gmail.com 2025-05-19T13:29:56+03:00 sync : ggml
6c35981a643d4f74a286268b62f65bfa156af2e6 8b5e19aea6ce9fe4452598663924373234041440 Johannes Gäßler johannesg@5d6.de 2025-05-19T09:33:35+02:00 Georgi Gerganov ggerganov@gmail.com 2025-05-19T13:29:56+03:00 mnist: fix segmentation fault (ggml/1227)
8b5e19aea6ce9fe4452598663924373234041440 60aea028b575ab54e0dfbb31db641bb93d2ae8c4 Diego Devesa slarengh@gmail.com 2025-05-18T18:30:13-07:00 Georgi Gerganov ggerganov@gmail.com 2025-05-19T13:29:56+03:00 ggml : fix apple OS check in ggml_print_backtrace (ggml/1229)
60aea028b575ab54e0dfbb31db641bb93d2ae8c4 9c55e5c5c24990dd17fd6c8f4f2159052d2b06f1 Daniel Tang danielzgtg.opensource@gmail.com 2025-05-17T19:06:26-04:00 Georgi Gerganov ggerganov@gmail.com 2025-05-19T13:29:56+03:00 ggml : Fix missing backtrace on Linux (ggml/1228)
9c55e5c5c24990dd17fd6c8f4f2159052d2b06f1 33d7aed4a83e7bbecac4af535208d4ed3e1ca8fd Nick 0x0b4ac@gmail.com 2025-05-19T18:25:41+08:00 GitHub noreply@github.com 2025-05-19T13:25:41+03:00 fix: check model pointer validity before use (#13631)
33d7aed4a83e7bbecac4af535208d4ed3e1ca8fd 6a2bc8bfb7cd502e5ebc72e36c97a6f848c21c2c Chenguang Li 757486878@qq.com 2025-05-19T14:21:17+08:00 GitHub noreply@github.com 2025-05-19T14:21:17+08:00 CANN: Support MOE Model MUL_MAT_ID (#13042)
6a2bc8bfb7cd502e5ebc72e36c97a6f848c21c2c e3a7cf6c5bf6a0a24217f88607b06e4405a2b5d9 Isaac McFadyen isaac@imcf.me 2025-05-17T17:59:48-04:00 GitHub noreply@github.com 2025-05-17T23:59:48+02:00 server : added --no-prefill-assistant flag (#13608)
e3a7cf6c5bf6a0a24217f88607b06e4405a2b5d9 518329b2d4434d0683c30b741b4f4cf5734b5f99 Gilad S. 7817232+giladgd@users.noreply.github.com 2025-05-17T21:26:43+03:00 GitHub noreply@github.com 2025-05-17T15:26:43-03:00 cmake: use the current build config for vulkan-shaders-gen (#13595)
518329b2d4434d0683c30b741b4f4cf5734b5f99 2f5a4e1e09567e09be8b84a5f976334de9539d17 Georgi Gerganov ggerganov@gmail.com 2025-05-17T12:58:55+03:00 GitHub noreply@github.com 2025-05-17T12:58:55+03:00 parallel : add option for non-shared and larger prompts (#13598)
2f5a4e1e09567e09be8b84a5f976334de9539d17 4f41ee11d6a4ddb02e4644922bf0b56880ee6f55 Jeff Bolz jbolz@nvidia.com 2025-05-17T16:14:55+09:00 GitHub noreply@github.com 2025-05-17T09:14:55+02:00 vulkan: move common FA code to flash_attn_base.comp (#13556)
4f41ee11d6a4ddb02e4644922bf0b56880ee6f55 3e0be1cacef290c99cbb99ceaa433b4344f87355 Jeff Bolz jbolz@nvidia.com 2025-05-17T15:35:47+09:00 GitHub noreply@github.com 2025-05-17T08:35:47+02:00 vulkan: use scalar FA rather than coopmat2 when N==1 (#13554)
3e0be1cacef290c99cbb99ceaa433b4344f87355 6aa892ec2aa7fe0c93e87c4b970d83a942fb9454 Z coffeevampirebusiness@gmail.com 2025-05-16T14:56:28-06:00 GitHub noreply@github.com 2025-05-16T22:56:28+02:00 llguidance : official v0.7.20 release (no actual changes) [noci] (#13594)
6aa892ec2aa7fe0c93e87c4b970d83a942fb9454 aea9f8b4e73876739bf88fb705502f291294e469 Xuan-Son Nguyen son@huggingface.co 2025-05-16T21:50:00+02:00 GitHub noreply@github.com 2025-05-16T21:50:00+02:00 server : do not return error out of context (with ctx shift disabled) (#13577)
aea9f8b4e73876739bf88fb705502f291294e469 06c1e4abc1e50ef6dcf6369f9685ca8fe82d21fe Xuan-Son Nguyen son@huggingface.co 2025-05-16T21:49:01+02:00 GitHub noreply@github.com 2025-05-16T21:49:01+02:00 webui : improve accessibility for visually impaired people (#13551)
06c1e4abc1e50ef6dcf6369f9685ca8fe82d21fe 415e40a357002d5d9b7a97ef20ccea9f4ae04c80 Xuan-Son Nguyen son@huggingface.co 2025-05-16T20:04:18+02:00 GitHub noreply@github.com 2025-05-16T20:04:18+02:00 readme : add list of dependencies and their license (#13591)
415e40a357002d5d9b7a97ef20ccea9f4ae04c80 654a67794f7a420c6931de2f4c145eaaade5dd16 Diego Devesa slarengh@gmail.com 2025-05-16T10:36:51-07:00 GitHub noreply@github.com 2025-05-16T19:36:51+02:00 releases : use arm version of curl for arm releases (#13592)
654a67794f7a420c6931de2f4c145eaaade5dd16 5364ae4ba53cc6367b8c8bf78876839122ca4e57 Georgi Gerganov ggerganov@gmail.com 2025-05-16T20:32:58+03:00 GitHub noreply@github.com 2025-05-16T20:32:58+03:00 metal : add FA-vec kernel for head size 64 (#13583)
5364ae4ba53cc6367b8c8bf78876839122ca4e57 7c07ac244d59c833bf209582f6df019a77cdda59 Diego Devesa slarengh@gmail.com 2025-05-16T07:38:07-07:00 GitHub noreply@github.com 2025-05-16T16:38:07+02:00 llama : print hint when loading a model when no backends are loaded (#13589)
7c07ac244d59c833bf209582f6df019a77cdda59 0a338ed013c23aecdce6449af736a35a465fa60f Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-16T14:54:23+02:00 GitHub noreply@github.com 2025-05-16T14:54:23+02:00 ci : add ppc64el to build-linux-cross (#13575)
0a338ed013c23aecdce6449af736a35a465fa60f bc098c3cf0aac93e57e9bda9d95e2456acf88894 Łukasz Ślusarczyk 112692748+lslusarczyk@users.noreply.github.com 2025-05-16T12:15:29+02:00 GitHub noreply@github.com 2025-05-16T18:15:29+08:00 sycl : fixed compilation warnings (#13582)
bc098c3cf0aac93e57e9bda9d95e2456acf88894 c6a2c9e7411f54b0770b319740561bbd6a2ebd27 Olivier Chafik olivier.chafik@gmail.com 2025-05-15T23:29:10+01:00 GitHub noreply@github.com 2025-05-15T23:29:10+01:00 minja: sync (qwen3) (#13573)
c6a2c9e7411f54b0770b319740561bbd6a2ebd27 07ad2b6db3c117868728e2cdfe3b711473d66e14 Diego Devesa slarengh@gmail.com 2025-05-15T10:13:11-07:00 GitHub noreply@github.com 2025-05-15T19:13:11+02:00 gguf : use ggml log system (#13571)
07ad2b6db3c117868728e2cdfe3b711473d66e14 c531edfa34fec8074d0b424396cdd450df23b612 Daniel Tang danielzgtg.opensource@gmail.com 2025-05-15T12:47:10-04:00 GitHub noreply@github.com 2025-05-15T18:47:10+02:00 gguf-py : fix disconnect-before-connect in editor-gui (#13569)
c531edfa34fec8074d0b424396cdd450df23b612 02cdd2d8b092b5a4bb18e013c6887ce49ba20ac5 Xuan-Son Nguyen son@huggingface.co 2025-05-15T17:40:07+02:00 GitHub noreply@github.com 2025-05-15T17:40:07+02:00 convert : fix conversion for llama 4 (#13567)
02cdd2d8b092b5a4bb18e013c6887ce49ba20ac5 64bb51cf90d3eede8c150a23d59a0c718b78065b Atharva Dubey atharva.dubey@codeplay.com 2025-05-15T16:39:52+01:00 GitHub noreply@github.com 2025-05-15T17:39:52+02:00 sycl: simplify bin_bcast_kernel (#13383)
64bb51cf90d3eede8c150a23d59a0c718b78065b 9c404ed54c3c8d8d2aa3153313766c8286739387 Svetlozar Georgiev 55534064+sgeor255@users.noreply.github.com 2025-05-15T16:35:44+01:00 GitHub noreply@github.com 2025-05-15T17:35:44+02:00 sycl: reordered Q4_K MMVQ (#13109)
9c404ed54c3c8d8d2aa3153313766c8286739387 6c8b91500e75df6664278d1e9af3e39e8a2fb0d0 Łukasz Ślusarczyk 112692748+lslusarczyk@users.noreply.github.com 2025-05-15T16:53:41+02:00 GitHub noreply@github.com 2025-05-15T16:53:41+02:00 sycl: use oneDNN for matrices multiplication (#12972)
6c8b91500e75df6664278d1e9af3e39e8a2fb0d0 3cc1f1f1d24472a6558c942b1c78989ff4b0e569 Diego Devesa slarengh@gmail.com 2025-05-15T06:46:55-07:00 GitHub noreply@github.com 2025-05-15T15:46:55+02:00 llama-bench : fix -ot with dl backends (#13563)
3cc1f1f1d24472a6558c942b1c78989ff4b0e569 c753d7bed0dc2cc2d5c42dfa9806cba91748392e Xuan-Son Nguyen son@huggingface.co 2025-05-15T14:24:50+02:00 GitHub noreply@github.com 2025-05-15T14:24:50+02:00 webui : handle PDF input (as text or image) + convert pasted long content to file (#13562)
c753d7bed0dc2cc2d5c42dfa9806cba91748392e b2838049ccf50859cea4c390e81405c2fb01e820 Piotr Wilkin (ilintar) piotr.wilkin@syndatis.com 2025-05-15T08:40:58+02:00 GitHub noreply@github.com 2025-05-15T08:40:58+02:00 server : proper error handling for missing elements in messages array (OpenAI compatible backend) (#13540)
b2838049ccf50859cea4c390e81405c2fb01e820 aa48e373f256df9608395ee6881e7010840d6202 Georgi Gerganov ggerganov@gmail.com 2025-05-15T05:57:02+03:00 GitHub noreply@github.com 2025-05-15T05:57:02+03:00 bench : handle decode errors (#13548)
aa48e373f256df9608395ee6881e7010840d6202 e3a9421b78da5c810d812e6348a405f5acc39f34 Olivier Chafik ochafik@users.noreply.github.com 2025-05-15T02:39:51+01:00 GitHub noreply@github.com 2025-05-15T02:39:51+01:00 `server`: inject date_string in llama 3.x template + fix date for firefunction v2 (#12802)
e3a9421b78da5c810d812e6348a405f5acc39f34 5ab5d5fb256aa23f8ab4de8463515ea627f43006 Georgi Gerganov ggerganov@gmail.com 2025-05-14T23:15:15+03:00 GitHub noreply@github.com 2025-05-14T23:15:15+03:00 kv-cache : fix out-of-bounds view during reserve graph (#13547)
5ab5d5fb256aa23f8ab4de8463515ea627f43006 3198405e98530c683d12fd123e3920f2bd2aafa5 Yibo Cai cyb70289@gmail.com 2025-05-15T03:53:52+08:00 GitHub noreply@github.com 2025-05-14T21:53:52+02:00 arm64: optimize q6_k_q8_k kernel with i8mm (#13519)
3198405e98530c683d12fd123e3920f2bd2aafa5 f5170c1d7a66222ca7c75d2022fec3ed87257e0b Olivier Chafik ochafik@users.noreply.github.com 2025-05-14T19:50:57+01:00 GitHub noreply@github.com 2025-05-14T19:50:57+01:00 `common`: add partial regex support (#12808)
f5170c1d7a66222ca7c75d2022fec3ed87257e0b 017f10b5fa630a013ec4f9936e410a60d4f460d5 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-14T20:22:49+02:00 GitHub noreply@github.com 2025-05-14T21:22:49+03:00 editorconfig : fix trailing whitespace from #13542 (#13546)
017f10b5fa630a013ec4f9936e410a60d4f460d5 4696d5674999dc10a7fb8c27b33406a929f7463a Gilad S. 7817232+giladgd@users.noreply.github.com 2025-05-14T19:18:18+03:00 GitHub noreply@github.com 2025-05-14T19:18:18+03:00 fix: crash when calling `llama_state_get_size` on a context without a KV cache (#13542)
4696d5674999dc10a7fb8c27b33406a929f7463a b7d26720821823e23e2273a99e38398d511242e9 Johannes Gäßler johannesg@5d6.de 2025-05-14T16:41:02+02:00 GitHub noreply@github.com 2025-05-14T16:41:02+02:00 CUDA: fix crash on large batch size for quant. MoE (#13537)
b7d26720821823e23e2273a99e38398d511242e9 6da34fa27620fa56e3334172e023f4f2533df51f Diego Devesa slarengh@gmail.com 2025-05-14T07:12:36-07:00 GitHub noreply@github.com 2025-05-14T16:12:36+02:00 llama : fix quantize with dl backends (#13539)
6da34fa27620fa56e3334172e023f4f2533df51f 5e7d95e22e386d316f7f659b74c9c34b65507912 Johannes Gäßler johannesg@5d6.de 2025-05-14T16:08:20+02:00 GitHub noreply@github.com 2025-05-14T16:08:20+02:00 CUDA: faster Deepseek FA, add Turing support (#13435)
5e7d95e22e386d316f7f659b74c9c34b65507912 053174436f3b3cbb01077f26ff17809537b832c7 Gabe Goodhart ghart@us.ibm.com 2025-05-14T06:53:59-06:00 GitHub noreply@github.com 2025-05-14T15:53:59+03:00 fix: Move build_inp_pos to the top of the graph section for build_granite (#13538)
053174436f3b3cbb01077f26ff17809537b832c7 360a9c98e13d35f322b4c5b1309aab0cc90ed82b Georgi Gerganov ggerganov@gmail.com 2025-05-14T15:42:10+03:00 GitHub noreply@github.com 2025-05-14T15:42:10+03:00 server : passthrough the /models endpoint during loading (#13535)
360a9c98e13d35f322b4c5b1309aab0cc90ed82b 09d13d94fb169093095416ac6323551c37b75339 Xuan-Son Nguyen son@huggingface.co 2025-05-14T13:35:07+02:00 GitHub noreply@github.com 2025-05-14T13:35:07+02:00 server : fix cache_tokens bug with no cache_prompt (#13533)
09d13d94fb169093095416ac6323551c37b75339 24e86cae7219b0f3ede1d5abdf5bf3ad515cccb8 bandoti 141645996+bandoti@users.noreply.github.com 2025-05-14T07:53:57-03:00 GitHub noreply@github.com 2025-05-14T07:53:57-03:00 cmake: simplify vulkan shader test logic (#13263)
24e86cae7219b0f3ede1d5abdf5bf3ad515cccb8 bb1681fbd532eba26ae4c14cd8be884c8afeb31c Jeff Bolz jbolz@nvidia.com 2025-05-14T18:55:26+09:00 GitHub noreply@github.com 2025-05-14T11:55:26+02:00 vulkan: KHR_coopmat flash attention (#13506)
bb1681fbd532eba26ae4c14cd8be884c8afeb31c d486dd3e8ecfba0170f8632a1530540de560f4a3 Xuan-Son Nguyen son@huggingface.co 2025-05-14T10:26:12+02:00 GitHub noreply@github.com 2025-05-14T10:26:12+02:00 webui : use fflate for more deterministic gzip compress (#13525)
d486dd3e8ecfba0170f8632a1530540de560f4a3 21ca987fba504d273ea28ebc4d3e5b3736a11c8e Luca Stefani luca.stefani.ge1@gmail.com 2025-05-14T10:07:31+02:00 GitHub noreply@github.com 2025-05-14T10:07:31+02:00 webui: Allow pasting file from clipboard (#13526)
21ca987fba504d273ea28ebc4d3e5b3736a11c8e be1d4a13db26750fac702ceb3af88ae4f39dc9f4 ddpasa 112642920+ddpasa@users.noreply.github.com 2025-05-14T09:59:12+02:00 GitHub noreply@github.com 2025-05-14T09:59:12+02:00 docs: Update link to ggml-org in multimodal.md (#13513)
be1d4a13db26750fac702ceb3af88ae4f39dc9f4 ab3971f2a0a526564bfde0e4cc8a5b90f9d33ad2 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-14T08:41:01+02:00 GitHub noreply@github.com 2025-05-14T08:41:01+02:00 scripts : fix compare-llama-bench.py show parameter (#13514)
ab3971f2a0a526564bfde0e4cc8a5b90f9d33ad2 e5c834f718a32b7584f142799bbf508fddb9021c Jeff Bolz jbolz@nvidia.com 2025-05-14T13:15:50+09:00 GitHub noreply@github.com 2025-05-14T06:15:50+02:00 vulkan: workaround FA compile failures on macos (#13517)
e5c834f718a32b7584f142799bbf508fddb9021c 71bdbdb58757d508557e6d8b387f666cdfb25c5e Ed Addario 29247825+EAddario@users.noreply.github.com 2025-05-13T18:12:31+01:00 GitHub noreply@github.com 2025-05-13T19:12:31+02:00 quantize : improve tensor-type pattern matching (#13033)
71bdbdb58757d508557e6d8b387f666cdfb25c5e f0995d28ce3d15095b6845d94ce4465e46575873 Xuan-Son Nguyen son@huggingface.co 2025-05-13T17:07:21+02:00 GitHub noreply@github.com 2025-05-13T17:07:21+02:00 clip : clip.h become private API (⚠️ breaking change) (#13510)
f0995d28ce3d15095b6845d94ce4465e46575873 c252e0c4097b34666e5a81db9d0450d71fa3098f Georgi Gerganov ggerganov@gmail.com 2025-05-13T18:04:39+03:00 GitHub noreply@github.com 2025-05-13T18:04:39+03:00 metal : use FA-vec kernel up to batch size 20 (#13496)
c252e0c4097b34666e5a81db9d0450d71fa3098f 4f711afed5e7ef4304b567c8888ee1aa60e868eb Georgi Gerganov ggerganov@gmail.com 2025-05-13T18:04:00+03:00 GitHub noreply@github.com 2025-05-13T18:04:00+03:00 metal : optimize multi-sequence FA vec kernel (#13493)
4f711afed5e7ef4304b567c8888ee1aa60e868eb b89d605a91dee0a518ecd582f8991a07d523e2fa Dan Johansson dan.johansson@arm.com 2025-05-13T17:02:28+02:00 GitHub noreply@github.com 2025-05-13T18:02:28+03:00 ggml-cpu: Update KleidiAI to v1.6 and fix include directives (#13509)
b89d605a91dee0a518ecd582f8991a07d523e2fa b4726345aca49e2ad62d615e6e370b3dbad6434f Georgi Gerganov ggerganov@gmail.com 2025-05-13T18:01:53+03:00 GitHub noreply@github.com 2025-05-13T18:01:53+03:00 batched-bench : fix pp batch contents (#13492)
b4726345aca49e2ad62d615e6e370b3dbad6434f bf7937112058f2815fc3825a9ff7b536ecafa3bb Xuan-Son Nguyen son@huggingface.co 2025-05-13T15:33:58+02:00 GitHub noreply@github.com 2025-05-13T15:33:58+02:00 mtmd : remove libllava, remove clip-quantize-cli (⚠️ breaking change) (#13460)
bf7937112058f2815fc3825a9ff7b536ecafa3bb d590cd4c244e5f260c42c290b83a358b9d86d763 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-13T15:31:12+02:00 GitHub noreply@github.com 2025-05-13T15:31:12+02:00 scripts : support arbitrary input file formats in compare-llama-bench.py (#13455)
d590cd4c244e5f260c42c290b83a358b9d86d763 1e2809bc4b5d8db2a9ed12ac872eca832c53f5fd Gabe Goodhart ghart@us.ibm.com 2025-05-13T07:12:01-06:00 GitHub noreply@github.com 2025-05-13T15:12:01+02:00 model : Granite MoE shared (#13269)
1e2809bc4b5d8db2a9ed12ac872eca832c53f5fd cf0a43bb6490bd49344775abb22ba26f8047cb54 Georgi Gerganov ggerganov@gmail.com 2025-05-13T14:01:45+03:00 Georgi Gerganov ggerganov@gmail.com 2025-05-13T14:02:28+03:00 sync : ggml
cf0a43bb6490bd49344775abb22ba26f8047cb54 f0d46ef15717cd609a7b69cf6190edde64d466c8 Diego Devesa slarengh@gmail.com 2025-05-12T15:31:37-07:00 GitHub noreply@github.com 2025-05-13T00:31:37+02:00 llama-bench : add defrag-thold, check for invalid ranges (#13487)
f0d46ef15717cd609a7b69cf6190edde64d466c8 de4c07f93783a1a96456a44dc16b9db538ee1618 lhez quic_lih@quicinc.com 2025-05-12T13:13:49-07:00 GitHub noreply@github.com 2025-05-12T13:13:49-07:00 opencl: remove unnecessary assert for `add` (#13257)
de4c07f93783a1a96456a44dc16b9db538ee1618 10d2af0eaa0aafd7c6577b279dfa5221ff44a63f Xuan-Son Nguyen son@huggingface.co 2025-05-12T15:06:51+02:00 GitHub noreply@github.com 2025-05-12T15:06:51+02:00 clip : cap max image size 1024 for qwen vl model (#13478)
10d2af0eaa0aafd7c6577b279dfa5221ff44a63f 064cc596ac44308dc326a17c9e3163c34a6f29d1 Johannes Gäßler johannesg@5d6.de 2025-05-12T14:44:49+02:00 GitHub noreply@github.com 2025-05-12T14:44:49+02:00 llama/ggml: add LLM training support (#10544)
064cc596ac44308dc326a17c9e3163c34a6f29d1 91159ee9df84edb72ccf874e353d2414f3a8c60d Georgi Gerganov ggerganov@gmail.com 2025-05-12T15:12:27+03:00 GitHub noreply@github.com 2025-05-12T15:12:27+03:00 context : fix state io for memory-less contexts (#13470)
91159ee9df84edb72ccf874e353d2414f3a8c60d 22cdab343b63edd0906f1c132616746085f81983 Anudit Nagar nagaranudit@gmail.com 2025-05-12T18:56:42+07:00 GitHub noreply@github.com 2025-05-12T13:56:42+02:00 server : allow content to be null in oaicompat_completion_params_parse (#13477)
22cdab343b63edd0906f1c132616746085f81983 a71a4075cdb8e81eaaa834e48ffda88b038286bc Diego Devesa slarengh@gmail.com 2025-05-12T13:08:22+02:00 GitHub noreply@github.com 2025-05-12T13:08:22+02:00 llama-bench : accept ranges for integer parameters (#13410)
a71a4075cdb8e81eaaa834e48ffda88b038286bc 95e18884fc7ea4031f70f1a518d5d1df616e5717 Dan Johansson dan.johansson@arm.com 2025-05-12T13:06:19+02:00 GitHub noreply@github.com 2025-05-12T13:06:19+02:00 ggml-cpu: Integrate fp32=bf16xbf16 SME KleidiAI kernel (#13053)
95e18884fc7ea4031f70f1a518d5d1df616e5717 df8491922f0ea4e032338186350dff2fb6b2b4e3 Johannes Gäßler johannesg@5d6.de 2025-05-12T10:51:21+02:00 GitHub noreply@github.com 2025-05-12T10:51:21+02:00 CUDA: fix misaligned synchronization in FA (#13469)
df8491922f0ea4e032338186350dff2fb6b2b4e3 14492144c286bbf38bb1903128403d9e2ebad54c Xuan-Son Nguyen son@huggingface.co 2025-05-12T10:29:13+02:00 GitHub noreply@github.com 2025-05-12T10:29:13+02:00 ggml : add mrope kernel for metal (#13457)
14492144c286bbf38bb1903128403d9e2ebad54c c104023994d36a8e791fc6a43789b84fd552cefc Atharva Dubey atharva.dubey@codeplay.com 2025-05-12T06:15:32+01:00 GitHub noreply@github.com 2025-05-12T13:15:32+08:00 enable dpcpp nightly builds with libraries (#13406)
c104023994d36a8e791fc6a43789b84fd552cefc 9a390c4829cd3058d26a2e2c09d16e3fd12bf1b1 City 125218114+city96@users.noreply.github.com 2025-05-12T00:39:06+02:00 GitHub noreply@github.com 2025-05-12T00:39:06+02:00 mtmd : Use RMS norm for InternVL 3 38B and 78B mmproj (#13459)
9a390c4829cd3058d26a2e2c09d16e3fd12bf1b1 09232370fc6426aa5dd9be01a8271b9c28f5af3a Anthony Umfer aumfer@gmail.com 2025-05-11T11:08:26-04:00 GitHub noreply@github.com 2025-05-11T17:08:26+02:00 tools : fix uninitialized llama_batch in server (#13436)
09232370fc6426aa5dd9be01a8271b9c28f5af3a 7474e00b34629e9cd8b06bc87ad935584ea30f8e Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-11T16:20:39+02:00 GitHub noreply@github.com 2025-05-11T16:20:39+02:00 scripts : exit compare-llama-bench.py gracefully when there's nothing to compare (#13451)
7474e00b34629e9cd8b06bc87ad935584ea30f8e 7f323a589f8684c0eb722e7309074cb5eac0c8b5 Johannes Gäßler johannesg@5d6.de 2025-05-11T16:09:33+02:00 GitHub noreply@github.com 2025-05-11T16:09:33+02:00 CUDA: fix crash with partial offloading of MoE (#13439)
7f323a589f8684c0eb722e7309074cb5eac0c8b5 3eac209319a6726fd9687c6188fc6b916b65953d David Huang 1969802+hjc4869@users.noreply.github.com 2025-05-11T20:18:39+08:00 GitHub noreply@github.com 2025-05-11T14:18:39+02:00 Add `--no-op-offload` to improve `-ot` pp perf in MoE models like llama4 400B (#13386)
3eac209319a6726fd9687c6188fc6b916b65953d a634d75d1bb4034b8c945042ceea268b5b895730 City 125218114+city96@users.noreply.github.com 2025-05-11T11:35:52+02:00 GitHub noreply@github.com 2025-05-11T11:35:52+02:00 mtmd : support InternVL 3 38B and 78B mmproj (#13443)
a634d75d1bb4034b8c945042ceea268b5b895730 62d4250e52917dc4d634f054b1e2183ed7dee944 Xuan-Son Nguyen son@huggingface.co 2025-05-11T11:34:23+02:00 GitHub noreply@github.com 2025-05-11T11:34:23+02:00 mtmd : move helpers to dedicated file (#13442)
62d4250e52917dc4d634f054b1e2183ed7dee944 0208355f42bdab88a08507ead4a6302790a08323 Thomas Germer 99991@users.noreply.github.com 2025-05-10T22:26:46+02:00 GitHub noreply@github.com 2025-05-10T22:26:46+02:00 docs : Fix typo in InternVL3 model name (#13440)
0208355f42bdab88a08507ead4a6302790a08323 d2a4ef05c60506ee48e7375eb36f2257de7ab0d2 Johannes Gäßler johannesg@5d6.de 2025-05-10T22:22:48+02:00 GitHub noreply@github.com 2025-05-10T22:22:48+02:00 CUDA: fix race conditions FlashAttention kernels (#13438)
d2a4ef05c60506ee48e7375eb36f2257de7ab0d2 15e6125a397f6086c1dfdf7584acdb7c730313dc Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-10T22:08:07+02:00 GitHub noreply@github.com 2025-05-10T22:08:07+02:00 vocab : add ByteDance-Seed/Seed-Coder (#13423)
15e6125a397f6086c1dfdf7584acdb7c730313dc 3b24d26c22b9d92b5aa39930988700c84e524cf4 Xuan-Son Nguyen son@huggingface.co 2025-05-10T19:57:54+02:00 GitHub noreply@github.com 2025-05-10T19:57:54+02:00 mtmd : add hard limit on image resolution for qwen2vl / qwen2.5vl (#13434)
3b24d26c22b9d92b5aa39930988700c84e524cf4 43dfd741a5da77982e0a94d1e522a2481dfb914d Xuan-Son Nguyen son@huggingface.co 2025-05-10T18:44:49+02:00 GitHub noreply@github.com 2025-05-10T18:44:49+02:00 server : update docs (#13432)
43dfd741a5da77982e0a94d1e522a2481dfb914d b064a51a4e7246fa43a3307f58e59f65e6be170a Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-10T17:19:52+02:00 GitHub noreply@github.com 2025-05-10T17:19:52+02:00 llguidance : set tokenizer slices to default (#13424)
b064a51a4e7246fa43a3307f58e59f65e6be170a 053367d149f778cdabc356ee3024494e0dd53223 Thammachart Chinvarapon 1731496+Thammachart@users.noreply.github.com 2025-05-10T21:34:48+07:00 GitHub noreply@github.com 2025-05-10T16:34:48+02:00 ci: free_disk_space flag enabled for intel variant (#13426)
053367d149f778cdabc356ee3024494e0dd53223 d8919424f1dee7dc1638349c616f2ef5d2ee16fb Xuan-Son Nguyen son@huggingface.co 2025-05-10T16:26:42+02:00 GitHub noreply@github.com 2025-05-10T16:26:42+02:00 mtmd : support InternVL 2.5 and 3 (#13422)
d8919424f1dee7dc1638349c616f2ef5d2ee16fb 7fef11766cdeb9fa7bbbe3db13580616b7d3d599 Johannes Gäßler johannesg@5d6.de 2025-05-10T09:16:52+02:00 GitHub noreply@github.com 2025-05-10T09:16:52+02:00 CUDA: fix FlashAttention on Turing (#13415)
7fef11766cdeb9fa7bbbe3db13580616b7d3d599 dc1d2adfc0f4de84da7923866d00781ec5c4e666 Xuan-Son Nguyen son@huggingface.co 2025-05-10T08:16:29+02:00 GitHub noreply@github.com 2025-05-10T08:16:29+02:00 arg : add env var to control mmproj (#13416)
dc1d2adfc0f4de84da7923866d00781ec5c4e666 7c28a74e0783f4bb74a246fb9f19bf212139e365 Jeff Bolz jbolz@nvidia.com 2025-05-09T23:07:07-07:00 GitHub noreply@github.com 2025-05-10T08:07:07+02:00 vulkan: scalar flash attention implementation (#13324)
7c28a74e0783f4bb74a246fb9f19bf212139e365 33eff4024084d1f0c8441b79f7208a52fad79858 Helton Reis 47722840+HRKings@users.noreply.github.com 2025-05-09T17:15:39-03:00 GitHub noreply@github.com 2025-05-09T23:15:39+03:00 chore(llguidance): use tagged version that does not break the build (#13413)
33eff4024084d1f0c8441b79f7208a52fad79858 17512a94d636c4b6c1332370acb3e5af3ca70918 Xuan-Son Nguyen son@huggingface.co 2025-05-09T19:29:37+02:00 GitHub noreply@github.com 2025-05-09T19:29:37+02:00 server : vision support via libmtmd (#12898)
17512a94d636c4b6c1332370acb3e5af3ca70918 611aa914ef4231fab5d1ad04773c42e119ae2d2e Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-05-09T16:34:08+01:00 GitHub noreply@github.com 2025-05-09T16:34:08+01:00 sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs (#12858)
611aa914ef4231fab5d1ad04773c42e119ae2d2e 0cf6725e9f9a164c39f7a87214d60342f7f946d8 Georgi Gerganov ggerganov@gmail.com 2025-05-09T15:14:56+03:00 GitHub noreply@github.com 2025-05-09T15:14:56+03:00 metal : optimize MoE for large batches (#13388)
0cf6725e9f9a164c39f7a87214d60342f7f946d8 27ebfcacbaadc6104e2b18acd8f13515cbf63dce Johannes Gäßler johannesg@5d6.de 2025-05-09T13:34:58+02:00 GitHub noreply@github.com 2025-05-09T13:34:58+02:00 CUDA: FA support for Deepseek (Ampere or newer) (#13306)
27ebfcacbaadc6104e2b18acd8f13515cbf63dce 5c86c9ed3ef1cc7307fdce05f0f0e2e45253cf90 Diego Devesa slarengh@gmail.com 2025-05-09T13:02:07+02:00 GitHub noreply@github.com 2025-05-09T13:02:07+02:00 llama : do not crash if there is no CPU backend (#13395)
5c86c9ed3ef1cc7307fdce05f0f0e2e45253cf90 efb8b47eda78ea8ae570d4fece3953aae499289e Johannes Gäßler johannesg@5d6.de 2025-05-09T12:14:04+02:00 GitHub noreply@github.com 2025-05-09T12:14:04+02:00 CUDA: fix crash on large batch size for MoE models (#13384)
efb8b47eda78ea8ae570d4fece3953aae499289e 0527771dd80bd18479dfaaa0a98be297fc3592bf Bartowski 3266127+bartowski1182@users.noreply.github.com 2025-05-09T05:53:58-04:00 GitHub noreply@github.com 2025-05-09T11:53:58+02:00 imatrix : Add --parse-special for enabling parsing of special tokens in imatrix calculation (#13389)
0527771dd80bd18479dfaaa0a98be297fc3592bf 2189fd3b6327a1d17893694125da8edcf74a6468 R0CKSTAR xiaodong.ye@mthreads.com 2025-05-09T17:25:50+08:00 GitHub noreply@github.com 2025-05-09T10:25:50+01:00 llama-run: add support for downloading models from ModelScope (#13370)
2189fd3b6327a1d17893694125da8edcf74a6468 3f96aeff394e9b72bbd2fa665c3e023a70ed8648 Xuan-Son Nguyen son@huggingface.co 2025-05-09T11:18:02+02:00 GitHub noreply@github.com 2025-05-09T11:18:02+02:00 mtmd : fix batch_view for m-rope (#13397)
3f96aeff394e9b72bbd2fa665c3e023a70ed8648 b486ba05bf973aae3652b3fd593e0b257d3a41d4 Xuan-Son Nguyen son@huggingface.co 2025-05-09T11:17:51+02:00 GitHub noreply@github.com 2025-05-09T11:17:51+02:00 llama : one-off chat template fix for Mistral-Small-2503 (#13398)
b486ba05bf973aae3652b3fd593e0b257d3a41d4 02115dcd9a6d0c03b527d5bee8c1493ab819ddbd Radoslav Gerganov rgerganov@gmail.com 2025-05-09T10:31:07+03:00 GitHub noreply@github.com 2025-05-09T10:31:07+03:00 rpc : add rpc_msg_set_tensor_hash_req (#13353)
02115dcd9a6d0c03b527d5bee8c1493ab819ddbd d9c4accaff30926e8a5fd8a2429e549158a845e0 Jeff Bolz jbolz@nvidia.com 2025-05-09T02:23:41-05:00 GitHub noreply@github.com 2025-05-09T09:23:41+02:00 vulkan: Allow up to 4096 elements for mul_mat_id row_ids (#13326)
d9c4accaff30926e8a5fd8a2429e549158a845e0 15e03282bb432631193464100c2237a3b6bcfe4c Xuan-Son Nguyen son@huggingface.co 2025-05-09T09:06:37+02:00 GitHub noreply@github.com 2025-05-09T09:06:37+02:00 server : (webui) rename has_multimodal --> modalities (#13393)
15e03282bb432631193464100c2237a3b6bcfe4c f05a6d71a0f3dbf0730b56a1abbad41c0f42e63d Diego Devesa slarengh@gmail.com 2025-05-08T23:45:22+02:00 GitHub noreply@github.com 2025-05-08T23:45:22+02:00 ci : limit write permission to only the release step + fixes (#13392)
f05a6d71a0f3dbf0730b56a1abbad41c0f42e63d ee01d71e585f4a17ae83ec55d77acfdcf0bfa798 Matt Clayton 156335168+mattjcly@users.noreply.github.com 2025-05-08T14:25:39-04:00 GitHub noreply@github.com 2025-05-08T20:25:39+02:00 mtmd : Expose helper_decode_image_chunk (#13366)
ee01d71e585f4a17ae83ec55d77acfdcf0bfa798 8c83449cb780c201839653812681c3a4cf17feed Xuan-Son Nguyen son@huggingface.co 2025-05-08T18:51:45+02:00 GitHub noreply@github.com 2025-05-08T18:51:45+02:00 server : (webui) fix a very small misalignment (#13387)
8c83449cb780c201839653812681c3a4cf17feed 1a844be132dbe865358e1de81136920c1d35ac73 Xuan-Son Nguyen son@huggingface.co 2025-05-08T15:37:29+02:00 GitHub noreply@github.com 2025-05-08T15:37:29+02:00 server : (webui) revamp the input area, plus many small UI improvements (#13365)
1a844be132dbe865358e1de81136920c1d35ac73 0ccc1213549e39ef4c1affb1bf5f49651ef4ce48 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-08T15:34:29+02:00 GitHub noreply@github.com 2025-05-08T15:34:29+02:00 convert : support rope_scaling type and rope_type (#13349)
0ccc1213549e39ef4c1affb1bf5f49651ef4ce48 6562e5a4d6c58326dcd79002ea396d4141f1b18e welix taichitary@gmail.com 2025-05-08T22:03:53+09:00 GitHub noreply@github.com 2025-05-08T15:03:53+02:00 mtmd : fix the calculation of n_tokens for smolvlm (#13381)
6562e5a4d6c58326dcd79002ea396d4141f1b18e 51fb96b1ff2e1cc98b2492a012b7d93531a6a9a8 Georgi Gerganov ggerganov@gmail.com 2025-05-08T14:28:33+03:00 GitHub noreply@github.com 2025-05-08T14:28:33+03:00 context : allow cache-less context for embeddings (#13108)
51fb96b1ff2e1cc98b2492a012b7d93531a6a9a8 70a6991edf1b60e7afa8962f830320583f3babb0 Georgi Gerganov ggerganov@gmail.com 2025-05-08T14:26:50+03:00 GitHub noreply@github.com 2025-05-08T14:26:50+03:00 context : remove logits_all flag (#13284)
70a6991edf1b60e7afa8962f830320583f3babb0 f0610212069929f92d428ae3b5596bfbe69c020b Diego Devesa slarengh@gmail.com 2025-05-08T13:15:28+02:00 GitHub noreply@github.com 2025-05-08T13:15:28+02:00 ci : move release workflow to a separate file (#13362)
f0610212069929f92d428ae3b5596bfbe69c020b 8733e0cf6eefc7c7752297cc22d0836706f4222c Diego Devesa slarengh@gmail.com 2025-05-08T13:15:15+02:00 GitHub noreply@github.com 2025-05-08T13:15:15+02:00 llama : print size and type of overridden tensors (#13364)
8733e0cf6eefc7c7752297cc22d0836706f4222c 814f795e063c257f33b921eab4073484238a151a Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-05-08T10:08:01+01:00 GitHub noreply@github.com 2025-05-08T10:08:01+01:00 sycl: addressing non-contiguous src1 mul_mats (nc and batched) (#13343)
814f795e063c257f33b921eab4073484238a151a d879433824ac3c16b3b5f00075895d0ee9688e34 Diego Devesa slarengh@gmail.com 2025-05-07T16:36:33+02:00 GitHub noreply@github.com 2025-05-07T16:36:33+02:00 docker : disable arm64 and intel images (#13356)
d879433824ac3c16b3b5f00075895d0ee9688e34 13b0a04597a4581cad4d9027a848f450c623801d Georgi Gerganov ggerganov@gmail.com 2025-05-07T16:39:36+03:00 Georgi Gerganov ggerganov@gmail.com 2025-05-07T17:28:36+03:00 sync : ggml
13b0a04597a4581cad4d9027a848f450c623801d bba9d945c14b93c5264b7956e00736601ca6f89a Daniel Bevenius daniel.bevenius@gmail.com 2025-05-05T13:09:35+02:00 Georgi Gerganov ggerganov@gmail.com 2025-05-07T17:28:36+03:00 whisper: remove MSVC warnings pragmas (whisper/3090)
bba9d945c14b93c5264b7956e00736601ca6f89a bc4e1128f78be0fbb4e2fa630adb6a04b969ac68 Jared Tweed jaredtwe@gmail.com 2025-05-02T02:41:35-07:00 Georgi Gerganov ggerganov@gmail.com 2025-05-07T17:28:36+03:00 cmake : removed stdc++fs (whisper/3097)
bc4e1128f78be0fbb4e2fa630adb6a04b969ac68 39e73ae0d69f882d7e29cecc6dd8f5052fca6731 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-07T12:49:27+02:00 GitHub noreply@github.com 2025-05-07T12:49:27+02:00 llama : deci : support ffn-free with attention (#13296)
39e73ae0d69f882d7e29cecc6dd8f5052fca6731 1f73301b63668d61b5f3109489050a27dc3f65be Ycros 18012+ycros@users.noreply.github.com 2025-05-07T18:23:28+10:00 GitHub noreply@github.com 2025-05-07T11:23:28+03:00 common : Add a warning when we can't match samplers from a string or char. (#13330)
1f73301b63668d61b5f3109489050a27dc3f65be 4773d7a02ffdb05ba9e673ff21ce95351836e33a R0CKSTAR xiaodong.ye@mthreads.com 2025-05-07T15:48:23+08:00 GitHub noreply@github.com 2025-05-07T09:48:23+02:00 cuda : remove nrows_x in mul_mat_q_process_tile (#13325)
4773d7a02ffdb05ba9e673ff21ce95351836e33a 6c7fd67b647a76846d1691cd181011dff4549d02 Georgi Gerganov ggerganov@gmail.com 2025-05-07T10:28:02+03:00 GitHub noreply@github.com 2025-05-07T10:28:02+03:00 examples : remove infill (#13283)
6c7fd67b647a76846d1691cd181011dff4549d02 141a908a59bbc68ceae3bf090b850e33322a2ca9 piDack 104877312+piDack@users.noreply.github.com 2025-05-07T15:23:11+08:00 GitHub noreply@github.com 2025-05-07T09:23:11+02:00 llama : support tie embedding for chatglm models (#13328)
141a908a59bbc68ceae3bf090b850e33322a2ca9 32916a49072f01c43e20df374af5f8a1f70d6963 Johannes Gäßler johannesg@5d6.de 2025-05-06T23:35:51+02:00 GitHub noreply@github.com 2025-05-06T23:35:51+02:00 CUDA: mix virt/real CUDA archs for GGML_NATIVE=OFF (#13135)
32916a49072f01c43e20df374af5f8a1f70d6963 ffc727203af1061fdeb49efef30f76171722e403 Xuan-Son Nguyen son@huggingface.co 2025-05-06T22:40:24+02:00 GitHub noreply@github.com 2025-05-06T22:40:24+02:00 clip : refactor graph builder (#13321)
ffc727203af1061fdeb49efef30f76171722e403 91a86a6f354aa73a7aab7bc3d283be410fdc93a5 DocShotgun 126566557+DocShotgun@users.noreply.github.com 2025-05-06T13:36:24-07:00 GitHub noreply@github.com 2025-05-06T22:36:24+02:00 sampling : make top_n_sigma no-op at <=0 or a single candidate (#13345)
91a86a6f354aa73a7aab7bc3d283be410fdc93a5 f4ed10b69cc38c54070a47f841827de5e8984cdf oobabooga oobabooga4@gmail.com 2025-05-06T15:24:15-03:00 GitHub noreply@github.com 2025-05-06T20:24:15+02:00 sampling : don't consider -infinity values in top_n_sigma (#13344)
f4ed10b69cc38c54070a47f841827de5e8984cdf 1e333d5bba18e99bc328bb87ac1ee6a4e6260e0e Diego Devesa slarengh@gmail.com 2025-05-06T20:15:31+02:00 GitHub noreply@github.com 2025-05-06T20:15:31+02:00 cmake : remove arm64 msvc presets (#13342)
1e333d5bba18e99bc328bb87ac1ee6a4e6260e0e 2f54e348ad2999c4e31b8777592247622b20420f Akarshan Biswas akarshan@menlo.ai 2025-05-06T20:27:06+05:30 GitHub noreply@github.com 2025-05-06T20:27:06+05:30 SYCL: Disable reorder optimize by default and stop setting tensor extras when optimize is disabled (#13254)
2f54e348ad2999c4e31b8777592247622b20420f 2356fb1d53c86d838756211010bbabfafda7cb94 Xuan-Son Nguyen son@huggingface.co 2025-05-06T14:25:40+02:00 GitHub noreply@github.com 2025-05-06T14:25:40+02:00 llama : fix build_ffn without gate (#13336)
2356fb1d53c86d838756211010bbabfafda7cb94 764b85627b46f43d7ea801867cd1b6abef484574 Johannes Gäßler johannesg@5d6.de 2025-05-06T13:58:51+02:00 GitHub noreply@github.com 2025-05-06T13:58:51+02:00 CUDA: fix bad asserts for partial offload (#13337)
764b85627b46f43d7ea801867cd1b6abef484574 15a28ec8c705b188ebe178170966d1dcc36fe151 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-06T11:12:06+02:00 GitHub noreply@github.com 2025-05-06T11:12:06+02:00 convert : qwen2/3moe : set yarn metadata if present (#13331)
15a28ec8c705b188ebe178170966d1dcc36fe151 a7366faa5bb2fff97b9fb43340d853709f52d8c9 Johannes Gäßler johannesg@5d6.de 2025-05-06T08:36:46+02:00 GitHub noreply@github.com 2025-05-06T08:36:46+02:00 CUDA: fix --split-mode row for MMQ (#13323)
a7366faa5bb2fff97b9fb43340d853709f52d8c9 907036502070ba608bdb2aaebf802092d4cfba07 compilade git@compilade.net 2025-05-05T22:27:31-04:00 GitHub noreply@github.com 2025-05-05T22:27:31-04:00 gguf-py : avoid requiring pyside6 for other scripts (#13036)
907036502070ba608bdb2aaebf802092d4cfba07 233461f8121455f957a47e6a22a77b3bc88277b0 Johannes Gäßler johannesg@5d6.de 2025-05-05T22:32:13+02:00 GitHub noreply@github.com 2025-05-05T22:32:13+02:00 CUDA: fix logic for clearing padding with -ngl 0 (#13320)
233461f8121455f957a47e6a22a77b3bc88277b0 b34c859146630dff136943abc9852ca173a7c9d6 oobabooga oobabooga4@gmail.com 2025-05-05T17:12:19-03:00 GitHub noreply@github.com 2025-05-05T22:12:19+02:00 sampling : Integrate Top-nσ into main sampling chain (and add it to the server) (#13264)
b34c859146630dff136943abc9852ca173a7c9d6 9b61acf06041dcbaff6afa5f28940e93297f8520 igardev 49397134+igardev@users.noreply.github.com 2025-05-05T17:03:31+03:00 GitHub noreply@github.com 2025-05-05T16:03:31+02:00 server : Webui - change setText command from parent window to also send the message. (#13309)
9b61acf06041dcbaff6afa5f28940e93297f8520 5215b91e9377ce23e4ccc92ec3156bf5c7f892a3 Xuan-Son Nguyen son@huggingface.co 2025-05-05T16:02:55+02:00 GitHub noreply@github.com 2025-05-05T16:02:55+02:00 mtmd : rename llava directory to mtmd (#13311)
5215b91e9377ce23e4ccc92ec3156bf5c7f892a3 ae803bfc3d0fc2d0d3e1cce22ee103a30939e104 Xuan-Son Nguyen son@huggingface.co 2025-05-05T12:54:44+02:00 GitHub noreply@github.com 2025-05-05T12:54:44+02:00 clip : fix confused naming ffn_up and ffn_down (#13290)
ae803bfc3d0fc2d0d3e1cce22ee103a30939e104 66645a5285d8c4c5f9a3b3f360d042baac2d820a Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-05T12:34:26+02:00 GitHub noreply@github.com 2025-05-05T12:34:26+02:00 convert : bailingmoe : set yarn metadata if present (#13312)
66645a5285d8c4c5f9a3b3f360d042baac2d820a 27aa2595321c4d9cc4086a8e67bdea204b8309b0 Akarshan Biswas akarshan@menlo.ai 2025-05-05T13:39:10+05:30 GitHub noreply@github.com 2025-05-05T13:39:10+05:30 SYCL: Disable mul_mat kernels for noncontiguous tensor b (#13308)
27aa2595321c4d9cc4086a8e67bdea204b8309b0 9fdfcdaeddd1ef57c6d041b89cd8fb7048a0f028 Xuan-Son Nguyen son@huggingface.co 2025-05-04T23:43:42+02:00 GitHub noreply@github.com 2025-05-04T23:43:42+02:00 mtmd : add C public API (#13184)
9fdfcdaeddd1ef57c6d041b89cd8fb7048a0f028 6eb7d25c700e29d9272064db408855178d01741b Diego Devesa slarengh@gmail.com 2025-05-04T21:25:43+02:00 GitHub noreply@github.com 2025-05-04T21:25:43+02:00 rpc : use backend registry, support dl backends (#13304)
6eb7d25c700e29d9272064db408855178d01741b 86bd60d3fe4a08b8c9d920e6defbc2412d803569 Aaron Teo aaron.teo1@ibm.com 2025-05-05T01:49:12+08:00 GitHub noreply@github.com 2025-05-04T19:49:12+02:00 ggml : activate s390x simd for Q3_K (#13301)
86bd60d3fe4a08b8c9d920e6defbc2412d803569 9f2da5871f4bbd205b8a3b952cdc76283218d595 Diego Devesa slarengh@gmail.com 2025-05-04T17:05:20+02:00 GitHub noreply@github.com 2025-05-04T17:05:20+02:00 llava/mtmd : fixes to fully support dl backends (#13303)
9f2da5871f4bbd205b8a3b952cdc76283218d595 93c4e23905987949b714b21ae918ff6bfb55fe36 Diego Devesa slarengh@gmail.com 2025-05-04T14:20:49+02:00 GitHub noreply@github.com 2025-05-04T14:20:49+02:00 llama : build windows releases with dl backends (#13220)
93c4e23905987949b714b21ae918ff6bfb55fe36 8afbd968182909cf93fb15959fc867b6dd3adb53 Johannes Gäßler johannesg@5d6.de 2025-05-04T14:16:39+02:00 GitHub noreply@github.com 2025-05-04T14:16:39+02:00 CUDA: fix race condition in MMQ stream-k fixup (#13299)
8afbd968182909cf93fb15959fc867b6dd3adb53 8ae5ebcf859b05a2ea3bbd930133a2fe4a89ed3c Johannes Gäßler johannesg@5d6.de 2025-05-04T13:58:38+02:00 GitHub noreply@github.com 2025-05-04T13:58:38+02:00 CUDA: fix race condition in MMQ ids_dst (#13294)
8ae5ebcf859b05a2ea3bbd930133a2fe4a89ed3c 3e959f09764a2bb0e64af594eab83f7fb3e08eb2 Jeff Bolz jbolz@nvidia.com 2025-05-04T00:17:16-05:00 GitHub noreply@github.com 2025-05-04T07:17:16+02:00 vulkan: Additional type support for unary, binary, and copy (#13266)
3e959f09764a2bb0e64af594eab83f7fb3e08eb2 36667c8edcded08063ed51c7d57e9e086bbfc903 Johannes Gäßler johannesg@5d6.de 2025-05-04T00:50:37+02:00 GitHub noreply@github.com 2025-05-04T00:50:37+02:00 imatrix: fix oob writes if src1 is not contiguous (#13286)
36667c8edcded08063ed51c7d57e9e086bbfc903 3bf785f3efa89ed28294fbf73054558a2b034bfb Xuan-Son Nguyen son@huggingface.co 2025-05-03T20:07:54+02:00 GitHub noreply@github.com 2025-05-03T20:07:54+02:00 clip : revert the change of BOI/EOI token for GLM-edge (⚠️ breaking change) (#13259)
3bf785f3efa89ed28294fbf73054558a2b034bfb 1d36b3670b285e69e58b9d687c770a2a0a192194 ymcki 84055651+ymcki@users.noreply.github.com 2025-05-03T23:39:51+08:00 GitHub noreply@github.com 2025-05-03T17:39:51+02:00 llama : Llama-3_1-Nemotron-Ultra-253B-v1 support (#12843)
1d36b3670b285e69e58b9d687c770a2a0a192194 b34443923cad751483cc53af2e680d595daadce7 Diego Devesa slarengh@gmail.com 2025-05-02T20:27:13+02:00 GitHub noreply@github.com 2025-05-02T20:27:13+02:00 llama : move end-user examples to tools directory (#13249)
b34443923cad751483cc53af2e680d595daadce7 a75cb30dc9e63488c3614e2d5a9fe2306eaf47cd Georgi Gerganov ggerganov@gmail.com 2025-05-02T20:54:30+03:00 GitHub noreply@github.com 2025-05-02T20:54:30+03:00 sync : ggml (#13268)
a75cb30dc9e63488c3614e2d5a9fe2306eaf47cd 3f3769ba76061a511f02f2a48da2ad2d93fce511 Georgi Gerganov ggerganov@gmail.com 2025-05-02T20:54:13+03:00 GitHub noreply@github.com 2025-05-02T20:54:13+03:00 context : fix reorder logic (#13267)
3f3769ba76061a511f02f2a48da2ad2d93fce511 2f567611c0234bbca0a4009762acb47b56866095 shalinib-ibm Shalini.Salomi.Bodapati@ibm.com 2025-05-02T22:23:12+05:30 GitHub noreply@github.com 2025-05-02T19:53:12+03:00 ggml : Enable MMA for BF16 in llamafile_sgemm (#13148)
2f567611c0234bbca0a4009762acb47b56866095 7d2123484e6ba8fcd90fff8c01661b9dbbefdc9d Jared Van Bortel jared@nomic.ai 2025-05-02T11:42:30-04:00 GitHub noreply@github.com 2025-05-02T11:42:30-04:00 llama-model : support Qwen2 embedding models and pooling_mode_lasttoken (#13245)
7d2123484e6ba8fcd90fff8c01661b9dbbefdc9d 074e42ab31de4c99aa7d9d2d239660f64b2380d6 Jared Van Bortel jared@nomic.ai 2025-05-02T11:41:54-04:00 GitHub noreply@github.com 2025-05-02T11:41:54-04:00 convert : use correct context length for nomic-embed-text-v2 (#13216)
074e42ab31de4c99aa7d9d2d239660f64b2380d6 c642bc014c105728ce45015813351dc5a37f60a2 Xuan-Son Nguyen son@huggingface.co 2025-05-02T17:17:15+02:00 GitHub noreply@github.com 2025-05-02T17:17:15+02:00 convert : converting mmproj for Qwen2/2.5VL from convert_hf_to_gguf (#13209)
c642bc014c105728ce45015813351dc5a37f60a2 cb06a3c363f50cd35113984fe8fb164aea419077 Georgi Gerganov ggerganov@gmail.com 2025-05-02T17:48:36+03:00 GitHub noreply@github.com 2025-05-02T17:48:36+03:00 kv-cache : separate recurrent vs non-recurrent impl (#12799)
cb06a3c363f50cd35113984fe8fb164aea419077 626083faf73faa54440f934bb1741bf443be91b1 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-02T12:44:24+02:00 GitHub noreply@github.com 2025-05-02T12:44:24+02:00 llama : orion rope type is neox (#13261)
626083faf73faa54440f934bb1741bf443be91b1 2af6880178b4bc2c0eced726bab68b4bf333042b Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-05-02T12:40:56+02:00 GitHub noreply@github.com 2025-05-02T12:40:56+02:00 llama : plamo rope type is neox (#13260)
2af6880178b4bc2c0eced726bab68b4bf333042b e84773ab604fe0d935d03741df003716398dc57b piDack 104877312+piDack@users.noreply.github.com 2025-05-02T17:06:09+08:00 GitHub noreply@github.com 2025-05-02T11:06:09+02:00 llama-chat : reset glmedge chat template (#13253)
e84773ab604fe0d935d03741df003716398dc57b fab647e8842c5f80da7e8f2c625dab6a0e19e5d4 Shakil Ahmed 44522075+ahmedshakill@users.noreply.github.com 2025-05-02T14:20:27+06:00 GitHub noreply@github.com 2025-05-02T10:20:27+02:00 mtmd-cli : fix out_of_range when input image path is empty (#13244)
fab647e8842c5f80da7e8f2c625dab6a0e19e5d4 dcf886007de4b8e5200f461a13233315f897fb9d Georgi Gerganov ggerganov@gmail.com 2025-05-02T09:48:31+03:00 GitHub noreply@github.com 2025-05-02T09:48:31+03:00 server : add cache reuse card link to help (#13230)
dcf886007de4b8e5200f461a13233315f897fb9d d24d5928086471063fa9d9fd45aca710fd1336ae Xuan-Son Nguyen son@huggingface.co 2025-05-02T08:45:10+02:00 GitHub noreply@github.com 2025-05-02T08:45:10+02:00 convert : explicitly disable trust_remote_code for AutoConfig (#13246)
d24d5928086471063fa9d9fd45aca710fd1336ae 8efbdadc616fa717c369059b9b388160958d886c bandoti 141645996+bandoti@users.noreply.github.com 2025-05-01T19:06:39-03:00 GitHub noreply@github.com 2025-05-01T19:06:39-03:00 ci: fix cross-compile sync issues (#12804)
8efbdadc616fa717c369059b9b388160958d886c f057808ffadce213f54c1884ab5096e60140f358 Justin Santa Barbara justinsb@google.com 2025-05-01T17:32:11-04:00 GitHub noreply@github.com 2025-05-01T23:32:11+02:00 rpc : avoid uninitialized memory in serialize_tensor (#13210)
f057808ffadce213f54c1884ab5096e60140f358 d7a14c42a1883a34a6553cbfe30da1e1b84dfd6a Jesse Gross jesse@kernel.org 2025-05-01T13:46:10-07:00 GitHub noreply@github.com 2025-05-01T22:46:10+02:00 ggml: Don't assert fail when tensor data changes (#13222)
d7a14c42a1883a34a6553cbfe30da1e1b84dfd6a b6e4ff69b8abd509647b531bd5b4e86950204f66 Diego Devesa slarengh@gmail.com 2025-05-01T21:48:08+02:00 GitHub noreply@github.com 2025-05-01T21:48:08+02:00 build : fix build info on windows (#13239)
b6e4ff69b8abd509647b531bd5b4e86950204f66 e0f572c8466e70d35cbd70ee536ad8fc83b2acac Loïc Carrère loic.carrere@gmail.com 2025-05-01T21:32:21+02:00 GitHub noreply@github.com 2025-05-01T21:32:21+02:00 clip : (minicpmv) Re-enable upscaling of images smaller than the CLIP image size (#13237)
e0f572c8466e70d35cbd70ee536ad8fc83b2acac 79f26e9e125b21760aeb016f34bfd42a93f48351 matteo matteo.serva@gmail.com 2025-05-01T21:16:38+02:00 GitHub noreply@github.com 2025-05-01T21:16:38+02:00 llama-chat : update GLM4 chat template (#13238)
79f26e9e125b21760aeb016f34bfd42a93f48351 fc727bcdd5a311c7c69a76dbf87f4784e828c7b4 Jeff Bolz jbolz@nvidia.com 2025-05-01T13:49:39-05:00 GitHub noreply@github.com 2025-05-01T20:49:39+02:00 vulkan: Add bfloat16 support (#12554)
fc727bcdd5a311c7c69a76dbf87f4784e828c7b4 b0ecbd434be024c06bf547491be444ed92e1123e Jeff Bolz jbolz@nvidia.com 2025-05-01T13:19:31-05:00 GitHub noreply@github.com 2025-05-01T20:19:31+02:00 vulkan: Handle src1 batch dimension in non-contiguous mat-vec-mul shader (#13191)
b0ecbd434be024c06bf547491be444ed92e1123e b1dd4d08e8fe40c9484422c0c5ec9bbd13b067a9 Johannes Gäßler johannesg@5d6.de 2025-05-01T20:18:56+02:00 GitHub noreply@github.com 2025-05-01T20:18:56+02:00 test: non-cont. b in test-backend-ops -o MUL_MAT (#13187)
b1dd4d08e8fe40c9484422c0c5ec9bbd13b067a9 99881f77d82efda80b21057d84f1cc2df2f1e0f6 Georgi Gerganov ggerganov@gmail.com 2025-05-01T17:07:13+03:00 Georgi Gerganov ggerganov@gmail.com 2025-05-01T20:15:34+03:00 sync : ggml
99881f77d82efda80b21057d84f1cc2df2f1e0f6 b5769d92b4510c77691ad9e3f8b643c2ba202e53 Daniel Bevenius daniel.bevenius@gmail.com 2025-05-01T10:05:24+02:00 Georgi Gerganov ggerganov@gmail.com 2025-05-01T20:15:34+03:00 whisper : add check that target name exists (whisper/3103)
b5769d92b4510c77691ad9e3f8b643c2ba202e53 8936784f7a1ec4f91637d04b77fdc90ec36ebac9 Daniel Bevenius daniel.bevenius@gmail.com 2025-04-29T15:47:55+02:00 Georgi Gerganov ggerganov@gmail.com 2025-05-01T20:15:34+03:00 ggml : suppress Windows compiler warnings (whisper/3075)
8936784f7a1ec4f91637d04b77fdc90ec36ebac9 13c9a3319b65469e883c49dd1c478abedc410157 Xuan-Son Nguyen son@huggingface.co 2025-05-01T17:05:42+02:00 GitHub noreply@github.com 2025-05-01T17:05:42+02:00 mtmd : add **vision** support for Mistral Small 3.1 (#13231)
13c9a3319b65469e883c49dd1c478abedc410157 a70183eb0079fdf6ebeaed12966338a039461b5d Xuan-Son Nguyen son@huggingface.co 2025-05-01T10:23:25+02:00 GitHub noreply@github.com 2025-05-01T10:23:25+02:00 arg : remove CURLINFO_EFFECTIVE_METHOD (#13228)
a70183eb0079fdf6ebeaed12966338a039461b5d 8d33d740c308fe0049515668aac0e561269afe9e Jared Van Bortel jared@nomic.ai 2025-05-01T03:09:41-04:00 GitHub noreply@github.com 2025-05-01T10:09:41+03:00 llama-model : fix the reported size class for nomic-embed-text-v2-moe (#13223)
8d33d740c308fe0049515668aac0e561269afe9e 4254bb49518e0f920f14c9aadd96eefdfd38b429 Georgi Gerganov ggerganov@gmail.com 2025-05-01T09:59:02+03:00 Georgi Gerganov ggerganov@gmail.com 2025-05-01T10:00:39+03:00 sync : ggml
4254bb49518e0f920f14c9aadd96eefdfd38b429 9998540149a490200894acfe595e9a1546bab723 Diego Devesa slarengh@gmail.com 2025-04-30T15:20:40+02:00 Georgi Gerganov ggerganov@gmail.com 2025-05-01T09:58:44+03:00 ggml : fix ggml_gallocr_ptr type (ggml/1205)
9998540149a490200894acfe595e9a1546bab723 e1e8e0991ffd9e99a445c6812bb519d5bac9f4b5 Georgi Gerganov ggerganov@gmail.com 2025-04-24T18:59:06+03:00 Georgi Gerganov ggerganov@gmail.com 2025-05-01T09:58:44+03:00 cuda : fix unused variable compile warning (whisper/0)
e1e8e0991ffd9e99a445c6812bb519d5bac9f4b5 6f67cf1f480926391ad75ff746e0a021214bf70c Johannes Gäßler johannesg@5d6.de 2025-04-30T23:12:59+02:00 GitHub noreply@github.com 2025-04-30T23:12:59+02:00 CUDA: batched+noncont MMQ, refactor bs>1 MoE code (#13199)
6f67cf1f480926391ad75ff746e0a021214bf70c 16a457facd996915652f6274384c87602b27d21a Xuan-Son Nguyen son@huggingface.co 2025-04-30T22:29:15+02:00 GitHub noreply@github.com 2025-04-30T21:29:15+01:00 arg : -hf do not fail if url mismatch (#13219)
16a457facd996915652f6274384c87602b27d21a 3e168bede4d27b35656ab8026015b87659ecbec2 ddh0 dylanhalladay02@icloud.com 2025-04-30T15:28:43-05:00 GitHub noreply@github.com 2025-04-30T21:28:43+01:00 fix typo: `n_ctx_pre_seq` -> `n_ctx_per_seq` (#13221)
3e168bede4d27b35656ab8026015b87659ecbec2 ceda28ef8e310a8dee60bf275077a3eedae8e36c Xuan-Son Nguyen son@huggingface.co 2025-04-30T16:56:24+02:00 GitHub noreply@github.com 2025-04-30T16:56:24+02:00 convert : improve model arch handling (#13122)
ceda28ef8e310a8dee60bf275077a3eedae8e36c 3b127c738535d95e06abd0d43da147bc13516ad0 Tatsuya Tanaka tanakasan2525@gmail.com 2025-04-30T22:25:20+09:00 GitHub noreply@github.com 2025-04-30T15:25:20+02:00 llava : remove duplicate include (#13207)
3b127c738535d95e06abd0d43da147bc13516ad0 e5007a5edf2692ef7151a81a61ce2716b83374e5 Olivier Chafik ochafik@users.noreply.github.com 2025-04-30T13:52:35+01:00 GitHub noreply@github.com 2025-04-30T14:52:35+02:00 common : add -jf / --json-schema-file flag (#12011)
e5007a5edf2692ef7151a81a61ce2716b83374e5 416313773b53585fddcafbcb914cbbfbaeb94b1f Jeff Bolz jbolz@nvidia.com 2025-04-30T07:38:37-05:00 GitHub noreply@github.com 2025-04-30T14:38:37+02:00 vulkan: use uint array index to avoid glslang bug (#13193)
416313773b53585fddcafbcb914cbbfbaeb94b1f 07c2e2f76cce9a61c110b6995fbb90ccea2c3aaa shalinib-ibm Shalini.Salomi.Bodapati@ibm.com 2025-04-30T16:47:08+05:30 GitHub noreply@github.com 2025-04-30T13:17:08+02:00 ggml : fix ppc64le build (#13176)
07c2e2f76cce9a61c110b6995fbb90ccea2c3aaa 44cd8d91ff2c9e4a0f2e3151f8d6f04c928e2571 Xuan-Son Nguyen son@huggingface.co 2025-04-30T13:06:15+02:00 GitHub noreply@github.com 2025-04-30T13:06:15+02:00 convert : correct typo image_mean --> image_std (#13208)
44cd8d91ff2c9e4a0f2e3151f8d6f04c928e2571 5933e6fdc9c051eea6c83b5a7608de12f9f15670 Aaron Teo 57927438+taronaeo@users.noreply.github.com 2025-04-30T17:47:35+08:00 GitHub noreply@github.com 2025-04-30T10:47:35+01:00 feat(ggml-cpu): enable z17 compile (#13182)
5933e6fdc9c051eea6c83b5a7608de12f9f15670 da84c04d8fa43ff92b172feb8130c74d062f956a Xuan-Son Nguyen son@huggingface.co 2025-04-30T10:46:32+02:00 GitHub noreply@github.com 2025-04-30T10:46:32+02:00 arg : allow using -hf offline (#13202)
da84c04d8fa43ff92b172feb8130c74d062f956a a0f7016d170ca4bfe24d9a9f26c024d034af69f2 Xuan-Son Nguyen son@huggingface.co 2025-04-30T10:44:07+02:00 GitHub noreply@github.com 2025-04-30T10:44:07+02:00 docker : do not build tests (#13204)
a0f7016d170ca4bfe24d9a9f26c024d034af69f2 19e899ce21a7c9ffcf8bb2b22269a75f6e078f8f xiaofei hbuxiaofei@gmail.com 2025-04-30T14:29:22+08:00 GitHub noreply@github.com 2025-04-30T09:29:22+03:00 rpc : fix cache directory initialization (#13188)
19e899ce21a7c9ffcf8bb2b22269a75f6e078f8f e2e1ddb93a01ce282e304431b37e60b3cddb6114 Johannes Gäßler johannesg@5d6.de 2025-04-29T23:32:04+02:00 GitHub noreply@github.com 2025-04-29T23:32:04+02:00 scripts: n_depth for compare-llama-bench [no ci] (#13201)
e2e1ddb93a01ce282e304431b37e60b3cddb6114 d9d398f84f96d16c308d4976f5e90222ecc2a492 matteo matteo.serva@gmail.com 2025-04-29T20:33:10+02:00 GitHub noreply@github.com 2025-04-29T20:33:10+02:00 server : Prefilling assistant message in openai compatible API (#13174)
d9d398f84f96d16c308d4976f5e90222ecc2a492 5a6398011704c31178d7b774be67856ba57647c8 Georgi Gerganov ggerganov@gmail.com 2025-04-29T20:22:57+03:00 GitHub noreply@github.com 2025-04-29T20:22:57+03:00 sampling : when top-k <= 0 -> noop (#13173)
5a6398011704c31178d7b774be67856ba57647c8 cdf76586b23c67abd3ca064ee2c084c57ae240bd Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-04-29T16:24:36+01:00 GitHub noreply@github.com 2025-04-29T17:24:36+02:00 llama-bench: fixed size of fields to correctly map to values (#13183)
cdf76586b23c67abd3ca064ee2c084c57ae240bd 7d3af70b089bb349b5d17eb01839224c99ec1952 Johannes Gäßler johannesg@5d6.de 2025-04-29T16:00:27+02:00 GitHub noreply@github.com 2025-04-29T16:00:27+02:00 CUDA: fix non-cont. inputs for batched mat mul (#13155)
7d3af70b089bb349b5d17eb01839224c99ec1952 00e3e5a194e88e604e7c91391b9e90332888fd72 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-04-29T13:25:53+02:00 GitHub noreply@github.com 2025-04-29T13:25:53+02:00 llama : llm_type order by size (#13177)
00e3e5a194e88e604e7c91391b9e90332888fd72 e98b3692be4cd8fbbd9a56fbacc2f2bf0bf26a68 Xuan-Son Nguyen son@huggingface.co 2025-04-29T11:47:04+02:00 GitHub noreply@github.com 2025-04-29T11:47:04+02:00 mtmd : add qwen2vl and qwen2.5vl (#13141)
e98b3692be4cd8fbbd9a56fbacc2f2bf0bf26a68 b6ce7430b7eb51f032152316880204e0a9c0470e Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-04-29T11:00:31+02:00 GitHub noreply@github.com 2025-04-29T11:00:31+02:00 llama : set qwen3 model type sizes (#13175)
b6ce7430b7eb51f032152316880204e0a9c0470e 5f5e39e1ba5dbea814e41f2a15e035d749a520bc Xuan-Son Nguyen son@huggingface.co 2025-04-29T08:45:49+02:00 GitHub noreply@github.com 2025-04-29T09:45:49+03:00 llama-graph : fix text position for mrope (#13159)
5f5e39e1ba5dbea814e41f2a15e035d749a520bc eaea3253244dc4bbe07f6cd81325847ccc6cf93e AT manyoso@users.noreply.github.com 2025-04-28T15:52:15-04:00 GitHub noreply@github.com 2025-04-28T22:52:15+03:00 model : Nomic Embed Text V2 with Mixture-of-Experts (MoE) architecture (#12466)
eaea3253244dc4bbe07f6cd81325847ccc6cf93e 43ddab6eeeaab5a04fe5a364af0bafb0e4d35065 Xuan-Son Nguyen son@huggingface.co 2025-04-28T21:23:19+02:00 GitHub noreply@github.com 2025-04-28T21:23:19+02:00 clip : fix model size display (#13153)
43ddab6eeeaab5a04fe5a364af0bafb0e4d35065 1831f538f720d1d99fba146f24f0a8e970838cc4 Ville Vesilehto ville@vesilehto.fi 2025-04-28T21:00:20+03:00 GitHub noreply@github.com 2025-04-28T21:00:20+03:00 fix(rpc): Improve input validation and error handling (#13069)
1831f538f720d1d99fba146f24f0a8e970838cc4 4e87962e34a4b257ec374c4baf6b1568554b81a9 Vishal Agarwal vishalagarwal.jss@gmail.com 2025-04-28T20:20:39+05:30 GitHub noreply@github.com 2025-04-28T16:50:39+02:00 llama-bench: add `-d` depth arg (#13096)
4e87962e34a4b257ec374c4baf6b1568554b81a9 fb0471d1753824e75474c24f82fbdd54c94dceda Xuan-Son Nguyen son@huggingface.co 2025-04-28T16:12:56+02:00 GitHub noreply@github.com 2025-04-28T16:12:56+02:00 mtmd : fix glm-edge redundant token count (#13139)
fb0471d1753824e75474c24f82fbdd54c94dceda d2b2031e5f11b826dcc718138642f147a2009665 pockers21 134406831+pockers21@users.noreply.github.com 2025-04-28T06:45:40-07:00 GitHub noreply@github.com 2025-04-28T16:45:40+03:00 context : do not clear output buffer on reserve (#13152)
d2b2031e5f11b826dcc718138642f147a2009665 5fa9e63be82225fb3249c76f39ddda3e5bdec0a3 Xuan-Son Nguyen son@huggingface.co 2025-04-28T14:20:56+02:00 GitHub noreply@github.com 2025-04-28T14:20:56+02:00 llama : (mrope) allow using normal 1D position for text token (#13138)
5fa9e63be82225fb3249c76f39ddda3e5bdec0a3 a4c340f974f9b7ac0c1aae897aabaa54549a97e5 Xuan-Son Nguyen son@huggingface.co 2025-04-28T12:18:59+02:00 GitHub noreply@github.com 2025-04-28T12:18:59+02:00 clip : refactor set input for cgraph + fix qwen2.5vl input (#13136)
a4c340f974f9b7ac0c1aae897aabaa54549a97e5 d0a417f3c7a5a22ef05b3b76d91dbe1d3362bf0c Akarshan Biswas akarshan@menlo.ai 2025-04-28T15:03:25+05:30 GitHub noreply@github.com 2025-04-28T11:33:25+02:00 SYCL: Add all missing unary kernels (#13074)
d0a417f3c7a5a22ef05b3b76d91dbe1d3362bf0c 43f2b07193cbcccd266734320ea9b948f5a01926 Georgi Gerganov ggerganov@gmail.com 2025-04-28T12:10:18+03:00 GitHub noreply@github.com 2025-04-28T12:10:18+03:00 readme : update hot topics (#13150)
43f2b07193cbcccd266734320ea9b948f5a01926 e5d6c2554e7597665e26991a93fa2f3d16c79ad5 Georgi Gerganov ggerganov@gmail.com 2025-04-28T11:57:19+03:00 GitHub noreply@github.com 2025-04-28T11:57:19+03:00 common : fix noreturn compile warning (#13151)
e5d6c2554e7597665e26991a93fa2f3d16c79ad5 f0dd6a1926cdb2f4183a937deee40db26ef8f1da Xuan-Son Nguyen son@huggingface.co 2025-04-28T10:11:58+02:00 GitHub noreply@github.com 2025-04-28T10:11:58+02:00 llama-chat : fix typo GML --> GLM (#13143)
f0dd6a1926cdb2f4183a937deee40db26ef8f1da 69699be48a6b94570773532850667f1591dc5bbe R0CKSTAR xiaodong.ye@mthreads.com 2025-04-28T15:33:28+08:00 GitHub noreply@github.com 2025-04-28T09:33:28+02:00 musa: fix typo in cc control (#13144)
69699be48a6b94570773532850667f1591dc5bbe 85f36e5e7173eef7c671c778db44c034e1d0ab19 Johannes Gäßler johannesg@5d6.de 2025-04-28T09:29:26+02:00 GitHub noreply@github.com 2025-04-28T09:29:26+02:00 CUDA: fix q_nope_absorbed prec for DS 2 Lite f16 (#13137)
85f36e5e7173eef7c671c778db44c034e1d0ab19 c0a97b762e5ec767dc414f0dc4979befd4c09a52 Xuan-Son Nguyen son@huggingface.co 2025-04-28T07:16:59+02:00 GitHub noreply@github.com 2025-04-28T08:16:59+03:00 arg : fix unused variable (#13142)
c0a97b762e5ec767dc414f0dc4979befd4c09a52 ced44be34290fab450f8344efa047d8a08e723b4 4onen 11580688+4onen@users.noreply.github.com 2025-04-27T14:48:26-07:00 GitHub noreply@github.com 2025-04-27T23:48:26+02:00 llama-bench : Add `--override-tensors` arg (#12922)
ced44be34290fab450f8344efa047d8a08e723b4 e291450b7602d7a36239e4ceeece37625f838373 matteo matteo.serva@gmail.com 2025-04-27T21:57:32+02:00 GitHub noreply@github.com 2025-04-27T21:57:32+02:00 llama-chat : fix wrong template in GLM4-0414 (#13140)
e291450b7602d7a36239e4ceeece37625f838373 59e991c23cc44cb5fb657e7b9358cac21fb79828 R0CKSTAR xiaodong.ye@mthreads.com 2025-04-27T19:22:49+08:00 GitHub noreply@github.com 2025-04-27T13:22:49+02:00 musa: fix build warning (#13129)
59e991c23cc44cb5fb657e7b9358cac21fb79828 ca2bb89eac2097ab4620448737e58af8452e444b LostRuins Concedo 39025047+LostRuins@users.noreply.github.com 2025-04-27T18:43:37+08:00 GitHub noreply@github.com 2025-04-27T12:43:37+02:00 Fixes Qwen2.5VL segfault during inference with https://github.com/ggml-org/llama.cpp/pull/12402 as has_qwen2vl_merger migration was incomplete (#13133)
ca2bb89eac2097ab4620448737e58af8452e444b 2d451c80590b9ac250322769ac13d3b4870dbcf7 HimariO dsfhe49854@gmail.com 2025-04-27T16:10:34+08:00 GitHub noreply@github.com 2025-04-27T10:10:34+02:00 clip : Add Qwen2.5VL support (#12402)
2d451c80590b9ac250322769ac13d3b4870dbcf7 4753791e70acd4d4e02f2098f14a03df26c992bd Xuan-Son Nguyen son@huggingface.co 2025-04-26T22:58:12+02:00 GitHub noreply@github.com 2025-04-26T22:58:12+02:00 common : add common_remote_get_content (#13123)
4753791e70acd4d4e02f2098f14a03df26c992bd 77d5e9a76a7b4a8a7c5bf9cf6ebef91860123cba Xuan-Son Nguyen son@huggingface.co 2025-04-26T22:39:47+02:00 GitHub noreply@github.com 2025-04-26T22:39:47+02:00 clip : improve projector naming (#13118)
77d5e9a76a7b4a8a7c5bf9cf6ebef91860123cba d5fe4e81bd447124836ecfb47d794f8768665b9f SXX sxx1136965276@gmail.com 2025-04-26T22:05:31+08:00 GitHub noreply@github.com 2025-04-26T16:05:31+02:00 ggml: move fp16/bf16 conversion optimizations to CPU backend + export conversion APIs (#13107)
d5fe4e81bd447124836ecfb47d794f8768665b9f 295354ea6848a77bdee204ee1c971d9b92ffcca9 frob rick+github@frob.com.au 2025-04-26T10:10:20+02:00 GitHub noreply@github.com 2025-04-26T10:10:20+02:00 grammar : handle maxItems == 0 in JSON schema (#13117)
295354ea6848a77bdee204ee1c971d9b92ffcca9 558a764713468f26f5a163d25a22100c9a04a48f Diego Devesa slarengh@gmail.com 2025-04-25T19:40:11+02:00 GitHub noreply@github.com 2025-04-25T19:40:11+02:00 llama : fix K-shift with quantized K and BLAS backend (#13113)
558a764713468f26f5a163d25a22100c9a04a48f edb18b6e8f5ea6509ad43057f8bb98fc557dbc4e City 125218114+city96@users.noreply.github.com 2025-04-25T14:38:34+02:00 GitHub noreply@github.com 2025-04-25T14:38:34+02:00 Force FP32 compute in GLM4 FFN Down (#13101)
edb18b6e8f5ea6509ad43057f8bb98fc557dbc4e 514c45608f93f66106a712dee1abe062099ce790 Xuan-Son Nguyen son@huggingface.co 2025-04-25T14:31:42+02:00 GitHub noreply@github.com 2025-04-25T14:31:42+02:00 clip : fix pixtral on some GPU backends (#13097)
514c45608f93f66106a712dee1abe062099ce790 553a5c3a9fdf771be2101bc3529937963f817457 Neo Zhang Jianyu jianyu.zhang@intel.com 2025-04-25T17:37:51+08:00 GitHub noreply@github.com 2025-04-25T17:37:51+08:00 change the reorder tensor from init to execute OP (#13003)
553a5c3a9fdf771be2101bc3529937963f817457 13be08daf992c89d5169518229b3740041c0f419 Radoslav Gerganov rgerganov@gmail.com 2025-04-25T10:08:08+03:00 GitHub noreply@github.com 2025-04-25T10:08:08+03:00 rpc : do not wait for response when sending RPC_CMD_SET_TENSOR (#12943)
13be08daf992c89d5169518229b3740041c0f419 226251ed56b85190e18a1cca963c45b888f4953c Xuan-Son Nguyen son@huggingface.co 2025-04-24T22:17:04+02:00 GitHub noreply@github.com 2025-04-24T22:17:04+02:00 clip : remove boi/eoi embeddings for GLM-edge model (#13081)
226251ed56b85190e18a1cca963c45b888f4953c 87616f0680947800ecba3e9f6bc6e101943bf8e6 Georgi Gerganov ggerganov@gmail.com 2025-04-24T22:29:22+03:00 GitHub noreply@github.com 2025-04-24T22:29:22+03:00 embeddings : fix batch sizes (#13076)
87616f0680947800ecba3e9f6bc6e101943bf8e6 63b4911494afe04778c61b9c19019341d71c99fc Georgi Gerganov ggerganov@gmail.com 2025-04-24T17:22:27+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-24T17:32:47+03:00 ggml : fix trailing whitespaces (#0)
63b4911494afe04778c61b9c19019341d71c99fc c6e8cc28c15166dba15629dba6a7366d4d5955ca Georgi Gerganov ggerganov@gmail.com 2025-04-24T16:47:43+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-24T17:32:47+03:00 sync : ggml
c6e8cc28c15166dba15629dba6a7366d4d5955ca b10d8bfdb1dac40cce34e8860ca5ec7d950c3a44 Acly aclysia@gmail.com 2025-04-17T14:16:45+02:00 Georgi Gerganov ggerganov@gmail.com 2025-04-24T17:32:47+03:00 ggml : Depthwise 2D convolution (ggml/1152)
b10d8bfdb1dac40cce34e8860ca5ec7d950c3a44 13b4548877326fdabee3e831b8cfd65d9844383c Johannes Gäßler johannesg@5d6.de 2025-04-24T15:57:10+02:00 GitHub noreply@github.com 2025-04-24T15:57:10+02:00 CUDA: use switch statements in constexpr functions (#13095)
13b4548877326fdabee3e831b8cfd65d9844383c 572b3141d343d7f947bf53b57513016e90db5680 Georgi Gerganov ggerganov@gmail.com 2025-04-24T16:00:10+03:00 GitHub noreply@github.com 2025-04-24T16:00:10+03:00 cmake : do not include ./src as public for libllama (#13062)
572b3141d343d7f947bf53b57513016e90db5680 7c727fbe39150fbe8381f4fa43fed08719ebebe6 Georgi Gerganov ggerganov@gmail.com 2025-04-24T15:44:05+03:00 GitHub noreply@github.com 2025-04-24T15:44:05+03:00 clang-tidy : disable warning about missing math parenthesis (#13091)
7c727fbe39150fbe8381f4fa43fed08719ebebe6 80982e815e67bae2442237f4e11466f44c9a2988 Xuan-Son Nguyen son@huggingface.co 2025-04-24T14:04:14+02:00 GitHub noreply@github.com 2025-04-24T14:04:14+02:00 arg : add --no-mmproj-offload (#13093)
80982e815e67bae2442237f4e11466f44c9a2988 7604a7d6b80e78eef8f275fc700d0f64820d672f Xuan-Son Nguyen son@huggingface.co 2025-04-24T12:14:13+02:00 GitHub noreply@github.com 2025-04-24T12:14:13+02:00 arg : clean up handling --mmproj with -hf (#13082)
7604a7d6b80e78eef8f275fc700d0f64820d672f b3b6d862cfdf190e1b9ad961639a25f5ebc0c7e3 Georgi Gerganov ggerganov@gmail.com 2025-04-24T10:38:30+03:00 GitHub noreply@github.com 2025-04-24T10:38:30+03:00 metal : fix floating-point range of attention scores in FA kernels (#13090)
b3b6d862cfdf190e1b9ad961639a25f5ebc0c7e3 56304069599f4dd9749d94b9bca2c2c65bb27c02 Eve 139727413+netrunnereve@users.noreply.github.com 2025-04-24T07:18:33Z GitHub noreply@github.com 2025-04-24T09:18:33+02:00 vulkan: matmul gcn tuning (#13016)
56304069599f4dd9749d94b9bca2c2c65bb27c02 ecda2ec4b347031a9b8a89ee2efc664ce63f599c pl752 pl752@mail.ru 2025-04-24T02:32:35+05:00 GitHub noreply@github.com 2025-04-23T23:32:35+02:00 llama-mtmd-cli: Sigint rework in mtmd vision example (#13080)
ecda2ec4b347031a9b8a89ee2efc664ce63f599c eb1776b15a32d832f1266deeeab75b9d255c5849 Xuan-Son Nguyen son@huggingface.co 2025-04-23T20:21:59+02:00 GitHub noreply@github.com 2025-04-23T20:21:59+02:00 mtmd : Support Pixtral 12B (#13065)
eb1776b15a32d832f1266deeeab75b9d255c5849 2cca6c01e46d2fc1124d15730273ed2acdad1016 piDack 104877312+piDack@users.noreply.github.com 2025-04-23T22:59:14+08:00 GitHub noreply@github.com 2025-04-23T16:59:14+02:00 convert : Append mult-eos,half-rope,bos to GLM4-0414 and Z (#13021)
2cca6c01e46d2fc1124d15730273ed2acdad1016 658987cfc9d752dca7758987390d5fb1a7a0a54a Radoslav Gerganov rgerganov@gmail.com 2025-04-23T10:32:49+03:00 GitHub noreply@github.com 2025-04-23T10:32:49+03:00 rpc : add command line option for number of threads for the CPU backend (#13060)
658987cfc9d752dca7758987390d5fb1a7a0a54a dc39a5e7a84815a90fa0c515ed8927870cf858c9 Johannes Gäßler johannesg@5d6.de 2025-04-22T21:27:40+02:00 GitHub noreply@github.com 2025-04-22T21:27:40+02:00 CUDA: noncont MMVQ + batched bs1 MUL_MAT_ID (#13014)
dc39a5e7a84815a90fa0c515ed8927870cf858c9 ab47dec3d37aa1927c2ec590e166b76141374ed3 Xuan-Son Nguyen son@huggingface.co 2025-04-22T16:24:54+02:00 GitHub noreply@github.com 2025-04-22T16:24:54+02:00 mtmd : support SmolVLM (version 1 and 2) (#13050)
ab47dec3d37aa1927c2ec590e166b76141374ed3 7b53389c24a507564be39e1ea82746a39749059b Georgi Gerganov ggerganov@gmail.com 2025-04-22T16:16:10+03:00 GitHub noreply@github.com 2025-04-22T16:16:10+03:00 security : add note about RPC and server functionality (#13061)
7b53389c24a507564be39e1ea82746a39749059b 243453533e029334181dda50d911d5fc5a2b2486 Georgi Gerganov ggerganov@gmail.com 2025-04-22T16:15:51+03:00 GitHub noreply@github.com 2025-04-22T16:15:51+03:00 metal : add memory pool for temp allocs (#12850)
243453533e029334181dda50d911d5fc5a2b2486 1d735c0b4fa0551c51c2f4ac888dd9a01f447985 Xuan-Son Nguyen son@huggingface.co 2025-04-22T10:37:00+02:00 GitHub noreply@github.com 2025-04-22T10:37:00+02:00 llava : update documentations (#13055)
1d735c0b4fa0551c51c2f4ac888dd9a01f447985 5368ddda7a262d195b54687a31009dcc1f8b1602 Diego Devesa slarengh@gmail.com 2025-04-21T18:13:51+02:00 GitHub noreply@github.com 2025-04-21T18:13:51+02:00 ggml : add SSE 4.2 and x64 base variant for CPUs without AVX (#12871)
5368ddda7a262d195b54687a31009dcc1f8b1602 84a9bf2fc2875205f0806fbbfbb66dc67204094c Akarshan Biswas akarshan@menlo.ai 2025-04-21T19:13:30+05:30 GitHub noreply@github.com 2025-04-21T19:13:30+05:30 SYCL: Add non-contiguous support in ROPE (#12993)
84a9bf2fc2875205f0806fbbfbb66dc67204094c 2016f07bd106c73699ecbaace80f55db5ed95dac Xuan-Son Nguyen son@huggingface.co 2025-04-21T15:32:58+02:00 GitHub noreply@github.com 2025-04-21T15:32:58+02:00 mtmd : merge llava, gemma3 and minicpmv CLI into single `llama-mtmd-cli` (#13012)
2016f07bd106c73699ecbaace80f55db5ed95dac 6602304814e679cc8c162bb760a034aceb4f8965 Xuan-Son Nguyen son@huggingface.co 2025-04-20T23:29:36+02:00 GitHub noreply@github.com 2025-04-20T23:29:36+02:00 convert : experimental support for `--mmproj` flag (#13023)
6602304814e679cc8c162bb760a034aceb4f8965 66168204be9559dc841f06f0025f3da01d9a8546 Jeffrey Morgan jmorganca@gmail.com 2025-04-20T03:15:41-07:00 GitHub noreply@github.com 2025-04-20T12:15:41+02:00 llava: fix errors in clip.h on certain compilers (#13030)
66168204be9559dc841f06f0025f3da01d9a8546 4ba9d711ba0a5bf0be00936d3258d4a5e4164d4f Jeff Bolz jbolz@nvidia.com 2025-04-20T03:50:02-05:00 GitHub noreply@github.com 2025-04-20T10:50:02+02:00 vulkan: support noncontiguous rms_norm (#13031)
4ba9d711ba0a5bf0be00936d3258d4a5e4164d4f 00137157fca3d17b90380762b4d7cc158d385bd3 Jeffrey Morgan jmorganca@gmail.com 2025-04-19T22:28:40-07:00 GitHub noreply@github.com 2025-04-20T08:28:40+03:00 metal: add neg operator (#13029)
00137157fca3d17b90380762b4d7cc158d385bd3 fb28f4f80efb5a2990a6a240433b02e259b0dbb1 bandoti 141645996+bandoti@users.noreply.github.com 2025-04-19T13:05:03-03:00 GitHub noreply@github.com 2025-04-19T18:05:03+02:00 Disable CI cross-compile builds (#13022)
fb28f4f80efb5a2990a6a240433b02e259b0dbb1 37b9f0d29d7301dd0ec5dfae8f357ccee96325d7 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-04-19T16:26:38+02:00 GitHub noreply@github.com 2025-04-19T16:26:38+02:00 gguf-py : fix upload python package workflow (#13020)
37b9f0d29d7301dd0ec5dfae8f357ccee96325d7 6408210082cc0a61b992b487be7e2ff2efbb9e36 Xuan-Son Nguyen son@huggingface.co 2025-04-19T09:15:45+02:00 GitHub noreply@github.com 2025-04-19T09:15:45+02:00 clip : refactor, add `image_manipulation` and `llava_uhd` classes (#13011)
6408210082cc0a61b992b487be7e2ff2efbb9e36 aff9d107b066c9d464f7ab1324280acfdcebf569 Daniel Tang danielzgtg.opensource@gmail.com 2025-04-18T16:02:55-04:00 GitHub noreply@github.com 2025-04-18T22:02:55+02:00 main : Fix Ctrl+D/newline handling (#12951)
aff9d107b066c9d464f7ab1324280acfdcebf569 35370ba94593c3abc9550328377d461fc4f822b7 Chris Thompson christopherthompson81@gmail.com 2025-04-18T12:30:41-06:00 GitHub noreply@github.com 2025-04-18T20:30:41+02:00 gguf-py : GGUF Editor GUI - Python + Qt6 (#12930)
35370ba94593c3abc9550328377d461fc4f822b7 8d6600576318dfc6b091fca744b0fd36a5e5255f Xuan-Son Nguyen son@huggingface.co 2025-04-18T19:58:12+02:00 GitHub noreply@github.com 2025-04-18T19:58:12+02:00 server : use std::move whenever possible (#12936)
8d6600576318dfc6b091fca744b0fd36a5e5255f b9154ecff93ff54dc554411eb844a2a654be49f2 Akarshan Biswas akarshan@menlo.ai 2025-04-18T19:27:56+05:30 GitHub noreply@github.com 2025-04-18T15:57:56+02:00 SYCL: Refactor and enable FP16 in binary broadcast OPs (#12975)
b9154ecff93ff54dc554411eb844a2a654be49f2 2db9ba1464f3de0aceb2b5289963e69fc369cb66 Xuan-Son Nguyen son@huggingface.co 2025-04-18T10:04:51+02:00 GitHub noreply@github.com 2025-04-18T10:04:51+02:00 mtmd : add methods to access `mtmd_image_tokens` (#12906)
2db9ba1464f3de0aceb2b5289963e69fc369cb66 2f74c354c0f752ed9aabf7d3a350e6edebd7e744 Radoslav Gerganov rgerganov@gmail.com 2025-04-18T10:13:42+03:00 GitHub noreply@github.com 2025-04-18T10:13:42+03:00 rpc : add RPC_CMD_HELLO (#12955)
2f74c354c0f752ed9aabf7d3a350e6edebd7e744 207c22ec2d6d793fc70830138617d1e016c5151c Georgi Gerganov ggerganov@gmail.com 2025-04-17T18:16:36+03:00 GitHub noreply@github.com 2025-04-17T18:16:36+03:00 graph : make FA compatible with MLA + add initial Metal kernels (#12953)
207c22ec2d6d793fc70830138617d1e016c5151c 7a395f67a7a02bb361d944b816d6e933889e28e1 Alan Gray agray3@users.noreply.github.com 2025-04-17T14:19:42+01:00 GitHub noreply@github.com 2025-04-17T15:19:42+02:00 ggml: Re-enable CUDA graphs in presence of CONT and DUP nodes (#12970)
7a395f67a7a02bb361d944b816d6e933889e28e1 971f245b3b5f3f55991bb779cb541b00f82eea1d hipudding huafengchun@gmail.com 2025-04-17T20:34:16+08:00 GitHub noreply@github.com 2025-04-17T20:34:16+08:00 CANN: Add support for async operator submission (#12864)
971f245b3b5f3f55991bb779cb541b00f82eea1d 12b17501e6015ffe568ac54fdf08e6580833bf1b Mikko Juola mikjuo@gmail.com 2025-04-17T01:37:05-07:00 GitHub noreply@github.com 2025-04-17T11:37:05+03:00 llama : recognize IBM Granite 3.3 FIM tokens (#12988)
12b17501e6015ffe568ac54fdf08e6580833bf1b 015022bb53387baa8b23817ac03743705c7d472b kimminsu 80271594+kimminsu38oo@users.noreply.github.com 2025-04-17T06:25:57+09:00 GitHub noreply@github.com 2025-04-16T14:25:57-07:00 opencl: fix incorrect local_size index in profiling log (#12868)
015022bb53387baa8b23817ac03743705c7d472b b43d89e311c5e7fbf62e5ec3c0401eb536677267 Jeff Bolz jbolz@nvidia.com 2025-04-16T13:37:25-05:00 GitHub noreply@github.com 2025-04-16T20:37:25+02:00 vulkan: enable coopmat2 FA gqa and split_k optimizations more often (#12931)
b43d89e311c5e7fbf62e5ec3c0401eb536677267 80f19b41869728eeb6a26569957b92a773a2b2c6 Chenguang Li 757486878@qq.com 2025-04-16T16:21:05+08:00 GitHub noreply@github.com 2025-04-16T16:21:05+08:00 CANN: Add 310P operator support check (#12962)
80f19b41869728eeb6a26569957b92a773a2b2c6 f8f820cc4dc37032d5375972ba904ce53043445d lhez quic_lih@quicinc.com 2025-04-15T12:26:00-07:00 GitHub noreply@github.com 2025-04-15T12:26:00-07:00 opencl: split `ggml-opencl.cl` into multiple files and cleanup (#12886)
f8f820cc4dc37032d5375972ba904ce53043445d 54a72720432810c89c2693f34908b60b88da1e46 Georgi Gerganov ggerganov@gmail.com 2025-04-15T14:45:05+03:00 GitHub noreply@github.com 2025-04-15T14:45:05+03:00 metal : add FA-vec kernels for head size 96 (#12952)
54a72720432810c89c2693f34908b60b88da1e46 84778e97703740d8ac5fb64e14d83b80eafa0f3c hipudding huafengchun@gmail.com 2025-04-15T19:08:55+08:00 GitHub noreply@github.com 2025-04-15T12:08:55+01:00 CANN: Add x86 build ci (#12950)
84778e97703740d8ac5fb64e14d83b80eafa0f3c 510676475f885ec064ff147af9f20ee7a9b12a50 David Huang 1969802+hjc4869@users.noreply.github.com 2025-04-15T17:20:38+08:00 GitHub noreply@github.com 2025-04-15T11:20:38+02:00 CUDA/HIP: Share the same unified memory allocation logic. (#12934)
510676475f885ec064ff147af9f20ee7a9b12a50 daa422881a0ec7944771bcc8ff8de34d11f5bd3b Akarshan Biswas akarshan@menlo.ai 2025-04-15T14:07:42+05:30 GitHub noreply@github.com 2025-04-15T10:37:42+02:00 SYCL: Add ROPE vision kernel (#12887)
daa422881a0ec7944771bcc8ff8de34d11f5bd3b eccc7a1602f0752507de4aaad1008b9618a282c8 Juk Armstrong 69222624+jukofyork@users.noreply.github.com 2025-04-15T07:49:57+01:00 GitHub noreply@github.com 2025-04-15T09:49:57+03:00 llama : DeepSeek V2/V3 MLA implementation (#12801)
eccc7a1602f0752507de4aaad1008b9618a282c8 0019279bb56f028e4eee19b59b19750736c719f7 Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2025-04-15T11:52:36+05:30 GitHub noreply@github.com 2025-04-15T09:22:36+03:00 ggml : Add AVX512 implementation of GEMM - Q4_Kx8 (#12829)
0019279bb56f028e4eee19b59b19750736c719f7 b0c75ac9f93322b45e3766e149417334b4fd1ed9 Chenguang Li 757486878@qq.com 2025-04-15T10:09:35+08:00 GitHub noreply@github.com 2025-04-15T10:09:35+08:00 CANN: Opt ROPE optimization (#12865)
b0c75ac9f93322b45e3766e149417334b4fd1ed9 d6d2c2ab8c8865784ba9fef37f2b2de3f2134d33 Xinpeng Dou 15529241576@163.com 2025-04-15T10:04:24+08:00 GitHub noreply@github.com 2025-04-15T10:04:24+08:00 CANN: Optimize CANN buffer pool memory management (#12875)
d6d2c2ab8c8865784ba9fef37f2b2de3f2134d33 75afa0ae31f0a51aaadcc5ff146eb7a32a7f9088 Russyyds 161207317+Russyyds@users.noreply.github.com 2025-04-15T01:18:20+08:00 GitHub noreply@github.com 2025-04-14T19:18:20+02:00 Add performance print for gemma3 in example (#12929)
75afa0ae31f0a51aaadcc5ff146eb7a32a7f9088 c772d549264c1be058411312a54049e0dc86a037 Akarshan Biswas akarshan@menlo.ai 2025-04-14T17:53:53+05:30 GitHub noreply@github.com 2025-04-14T14:23:53+02:00 SYCL: Fix im2col (#12910)
c772d549264c1be058411312a54049e0dc86a037 81c7e64fc239288e91a58adad9145110e0353822 Radoslav Gerganov rgerganov@gmail.com 2025-04-14T13:59:34+03:00 GitHub noreply@github.com 2025-04-14T13:59:34+03:00 rpc : use ggml_context_ptr (#12938)
81c7e64fc239288e91a58adad9145110e0353822 526739b879c9d6a702ed681138611197fa4d3f18 Neo Zhang Jianyu jianyu.zhang@intel.com 2025-04-14T18:19:07+08:00 GitHub noreply@github.com 2025-04-14T18:19:07+08:00 dsiable curl lib check, this action is missed by commit bd3f59f81289b920bcc597a208c14f55e39ed37e (#12761) (#12937)
526739b879c9d6a702ed681138611197fa4d3f18 a25355e2643b1feb2fcbbc4c00312aa86370d858 Georgi Gerganov ggerganov@gmail.com 2025-04-14T08:52:10+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-14T09:26:15+03:00 sync : ggml
a25355e2643b1feb2fcbbc4c00312aa86370d858 e959d32b1c16c02c0bc0bcdc7bc3f7077bc32f99 cmdr2 secondary.cmdr2@gmail.com 2025-04-11T12:14:19+05:30 Georgi Gerganov ggerganov@gmail.com 2025-04-14T09:26:15+03:00 cpu: fix cpu backend's supports-op for GET_ROWS_BACK. fixes a fatal when running test-backend-ops with only the CPU backend (ggml/1190)
e959d32b1c16c02c0bc0bcdc7bc3f7077bc32f99 307bfa253dea07c9270e78fa53b133504e9c3c9d SXX sxx1136965276@gmail.com 2025-04-14T13:47:55+08:00 GitHub noreply@github.com 2025-04-14T08:47:55+03:00 ggml: use _mm[512/256]_dpbusd[_avx]_epi32 to directly accumulate into the result register (#12773)
307bfa253dea07c9270e78fa53b133504e9c3c9d 71e90e8813f90097701e62f7fce137d96ddf41e2 Alan Gray agray3@users.noreply.github.com 2025-04-13T22:12:21+01:00 GitHub noreply@github.com 2025-04-13T23:12:21+02:00 ggml: disable CUDA graphs for unsupported DUP and CONT node types (#12891)
71e90e8813f90097701e62f7fce137d96ddf41e2 bc091a4dc585af25c438c8473285a8cfec5c7695 Ed Addario 29247825+EAddario@users.noreply.github.com 2025-04-13T19:29:28+01:00 GitHub noreply@github.com 2025-04-13T21:29:28+03:00 quantize: Handle user-defined quantization levels for additional tensors (#12511)
bc091a4dc585af25c438c8473285a8cfec5c7695 a4837577aae59c0ae640f3810094724bcac6bb28 Prajwal B Mehendarkar prajwal.b.mehendarkar@ibm.com 2025-04-12T21:03:39+05:30 GitHub noreply@github.com 2025-04-12T17:33:39+02:00 common : Define cache directory on AIX (#12915)
a4837577aae59c0ae640f3810094724bcac6bb28 e59ea539b83d2c7947c99bd350549364dbba450c Jeff Bolz jbolz@nvidia.com 2025-04-12T03:44:48-05:00 GitHub noreply@github.com 2025-04-12T10:44:48+02:00 vulkan: use aligned loads for flash attention mask (#12853)
e59ea539b83d2c7947c99bd350549364dbba450c c94085df2871d2cad11f6411304d7798d7dd4cdd Matt Clayton 156335168+mattjcly@users.noreply.github.com 2025-04-12T01:29:03-04:00 GitHub noreply@github.com 2025-04-12T07:29:03+02:00 llava: Fix cpu-only clip image encoding sefault (#12907)
c94085df2871d2cad11f6411304d7798d7dd4cdd e8a62631b3b05bcf2ad0e0c686881a9f3e3f03ca Georgi Gerganov ggerganov@gmail.com 2025-04-11T23:37:41+03:00 GitHub noreply@github.com 2025-04-11T23:37:41+03:00 server : add VSCode's Github Copilot Chat support (#12896)
e8a62631b3b05bcf2ad0e0c686881a9f3e3f03ca b6930ebc421e20399663bbd3bc7b7266d5236d06 yuri@FreeBSD yurivict@users.noreply.github.com 2025-04-11T13:04:14-07:00 GitHub noreply@github.com 2025-04-11T22:04:14+02:00 rpc : Set cache directory in rpc-server.cpp on FreeBSD (#12903)
b6930ebc421e20399663bbd3bc7b7266d5236d06 68b08f36d047bfa048ebd7d16262c53bf9ece701 Olivier Chafik ochafik@users.noreply.github.com 2025-04-11T12:47:52-07:00 GitHub noreply@github.com 2025-04-11T21:47:52+02:00 `tool-call`: fix non-tool-calling grammar crashes w/ Qwen / Hermes 2 templates (#12900)
68b08f36d047bfa048ebd7d16262c53bf9ece701 578754b3157d662c2fdd51eaa62b6c1f43d3172c yuri@FreeBSD yurivict@users.noreply.github.com 2025-04-11T12:45:44-07:00 GitHub noreply@github.com 2025-04-11T21:45:44+02:00 common : Define cache directory on FreeBSD (#12892)
578754b3157d662c2fdd51eaa62b6c1f43d3172c b2034c2b55b36b2192bdefb3b295db2a911370f5 Ewan Crawford ewan.cr@gmail.com 2025-04-11T15:32:14+02:00 GitHub noreply@github.com 2025-04-11T15:32:14+02:00 sycl: Support sycl_ext_oneapi_limited_graph (#12873)
b2034c2b55b36b2192bdefb3b295db2a911370f5 06bb53ad9b6e6d92b6ab6979927530080b1c990c tastelikefeet 58414341+tastelikefeet@users.noreply.github.com 2025-04-11T20:01:56+08:00 GitHub noreply@github.com 2025-04-11T14:01:56+02:00 contrib: support modelscope community (#12664)
06bb53ad9b6e6d92b6ab6979927530080b1c990c 0c509239445088f21265580b86733c5b184261d9 Yuxuan Zhang 2448370773@qq.com 2025-04-11T18:10:10+08:00 GitHub noreply@github.com 2025-04-11T12:10:10+02:00 llama-model : add Glm4Model implementation for GLM-4-0414 (#12867)
0c509239445088f21265580b86733c5b184261d9 fccf9cae83e6c6cd31a0ecb403d237638e427d0a Xuan-Son Nguyen son@huggingface.co 2025-04-11T12:09:39+02:00 GitHub noreply@github.com 2025-04-11T12:09:39+02:00 clip : use smart pointer (⚠️ breaking change) (#12869)
fccf9cae83e6c6cd31a0ecb403d237638e427d0a ec6c09d0fac1f2699c3ea1994e10482cb4b95e0f Akarshan Biswas akarshan@menlo.ai 2025-04-11T13:33:50+05:30 GitHub noreply@github.com 2025-04-11T16:03:50+08:00 SYCL: Add fp16 type support to unary op kernels (#12788)
ec6c09d0fac1f2699c3ea1994e10482cb4b95e0f 8ac9f5d765f2b1b7f821873618c0cff68ca59bc8 Daniel Han danielhanchen@gmail.com 2025-04-11T00:49:09-07:00 GitHub noreply@github.com 2025-04-11T09:49:09+02:00 convert : Llama4 RoPE fix (#12889)
8ac9f5d765f2b1b7f821873618c0cff68ca59bc8 12e9158f25cb47e97ca9b8083e1ec7415f262260 R0CKSTAR xiaodong.ye@mthreads.com 2025-04-11T15:26:17+08:00 GitHub noreply@github.com 2025-04-11T09:26:17+02:00 ci : Replace freediskspace to free_disk_space in docker.yml (#12861)
12e9158f25cb47e97ca9b8083e1ec7415f262260 5b1f13cb64dbb0de7e95dae86725928d350c188c Daniel Bevenius daniel.bevenius@gmail.com 2025-04-11T09:24:34+02:00 GitHub noreply@github.com 2025-04-11T09:24:34+02:00 xcf : add check for visionos build version (#12854)
5b1f13cb64dbb0de7e95dae86725928d350c188c 8b91d5355a84bbb5a73fd23fb658272473240efe Xuan-Son Nguyen son@huggingface.co 2025-04-11T09:23:37+02:00 GitHub noreply@github.com 2025-04-11T09:23:37+02:00 convert : proper tensor name mapping for llama4 (#12870)
8b91d5355a84bbb5a73fd23fb658272473240efe 0fed24c34787d2aa916ed204db88889591ad6e55 Xuan-Son Nguyen son@huggingface.co 2025-04-11T08:49:50+02:00 GitHub noreply@github.com 2025-04-11T08:49:50+02:00 llama : correct rms norm for llama 4 (#12882)
0fed24c34787d2aa916ed204db88889591ad6e55 47ba87d0a4f91eb181ea60f8a7fc2539b123aeaf Aaron Teo 57927438+taronaeo@users.noreply.github.com 2025-04-11T13:20:07+08:00 GitHub noreply@github.com 2025-04-11T08:20:07+03:00 ggml: fix compilation error s390x (#12848)
47ba87d0a4f91eb181ea60f8a7fc2539b123aeaf 1d2b613445bab6b179296aa63c8ee346fb947cc4 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:08:23+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:17:47+03:00 sync : ggml
1d2b613445bab6b179296aa63c8ee346fb947cc4 eb420e11484ef32cc1abedb2a26ecef1bbf1e31b Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:04:25+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:17:47+03:00 tests : fix init order (#0)
eb420e11484ef32cc1abedb2a26ecef1bbf1e31b cb79c2e7fa28e874694c14598c1fd2fde82263e1 Georgi Gerganov ggerganov@gmail.com 2025-04-10T23:59:16+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:17:47+03:00 sync : ggml
cb79c2e7fa28e874694c14598c1fd2fde82263e1 fe92821ea9ae53f3088cf2699a9e102448295fa0 cmdr2 secondary.cmdr2@gmail.com 2025-04-10T17:53:08+05:30 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:17:47+03:00 ggml: don't include arm_neon.h when using CUDA 12 with ARM Neon (ggml/1187)
fe92821ea9ae53f3088cf2699a9e102448295fa0 459895c32629e36cc3ca53bf3596d3e1322cfe09 Diego Devesa slarengh@gmail.com 2025-04-09T12:32:13+02:00 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:17:47+03:00 ggml : add bilinear upscale support (ggml/1185)
459895c32629e36cc3ca53bf3596d3e1322cfe09 e4bf72d631434b38656cb97e2411826d900433e8 Diego Devesa slarengh@gmail.com 2025-04-09T12:31:34+02:00 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:17:47+03:00 ggml : add more generic custom op, remove deprecated custom ops (ggml/1183)
e4bf72d631434b38656cb97e2411826d900433e8 8b9cc7cdd8a0dcf0176c60c755322c95b5965299 Georgi Gerganov ggerganov@gmail.com 2025-04-10T23:59:01+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-11T00:17:47+03:00 scripts : fix sync-ggml-am.sh
8b9cc7cdd8a0dcf0176c60c755322c95b5965299 64eda5deb9859e87a020e56bab5d2f9ca956f1de Xuan-Son Nguyen son@huggingface.co 2025-04-10T22:57:16+02:00 GitHub noreply@github.com 2025-04-10T22:57:16+02:00 llava : introduce libmtmd (#12849)
64eda5deb9859e87a020e56bab5d2f9ca956f1de fe5b78c89670b2f37ecb216306bed3e677b49d9f Xuan-Son Nguyen son@huggingface.co 2025-04-10T17:24:44+02:00 GitHub noreply@github.com 2025-04-10T17:24:44+02:00 convert : ability to lazy-load safetensors remotely without downloading to disk (#12820)
fe5b78c89670b2f37ecb216306bed3e677b49d9f 11d07e1e69138b46375e9267b31acd58e3813577 Chenguang Li 757486878@qq.com 2025-04-10T08:51:52+08:00 GitHub noreply@github.com 2025-04-10T08:51:52+08:00 CANN: Support more ops (#12841)
11d07e1e69138b46375e9267b31acd58e3813577 b0091ecc1e5c0f689be856fade3803a534f35c9f Prajwal B Mehendarkar prajwal.b.mehendarkar@ibm.com 2025-04-10T04:48:01+05:30 GitHub noreply@github.com 2025-04-10T01:18:01+02:00 Fixes #12823 (#12830)
b0091ecc1e5c0f689be856fade3803a534f35c9f 31f7803bc4e7c0dcc279ee04c2ecfb76b2afdd3e Rudi Servo rudiservo@gmail.com 2025-04-09T23:17:12Z GitHub noreply@github.com 2025-04-10T01:17:12+02:00 docker : added all CPU to GPU images (#12749)
31f7803bc4e7c0dcc279ee04c2ecfb76b2afdd3e 2391506ace6abb56186def40c7107fdfa694ed55 Piotr Kubaj pkubaj@anongoth.pl 2025-04-09T23:00:34Z GitHub noreply@github.com 2025-04-10T01:00:34+02:00 ggml-cpu-impl.h: do not redefine bool on POWER9 (#12856)
2391506ace6abb56186def40c7107fdfa694ed55 d3bd7193ba66c15963fd1c59448f22019a8caf6e Piotr Kubaj pkubaj@anongoth.pl 2025-04-09T23:00:25Z GitHub noreply@github.com 2025-04-10T01:00:25+02:00 ggml-impl.h: fix build on POWER9 (#12855)
d3bd7193ba66c15963fd1c59448f22019a8caf6e d9a63b2f2e91cdcb0eda211b7f49fadcdea0f664 Bo Zheng 368586905@qq.com 2025-04-09T17:47:36+08:00 GitHub noreply@github.com 2025-04-09T11:47:36+02:00 llama : Support Qwen3 and Qwen3MoE (#12828)
d9a63b2f2e91cdcb0eda211b7f49fadcdea0f664 8ed71242f464dc0a3fb3cffcfe064e55bdec72f9 R0CKSTAR xiaodong.ye@mthreads.com 2025-04-09T17:22:30+08:00 GitHub noreply@github.com 2025-04-09T11:22:30+02:00 musa: enable freediskspace for docker image build (#12839)
8ed71242f464dc0a3fb3cffcfe064e55bdec72f9 381603a77504ad1788965f694094540c1bed9ea2 Romain Biessy romain.biessy@codeplay.com 2025-04-09T11:22:04+02:00 GitHub noreply@github.com 2025-04-09T11:22:04+02:00 sycl: update documentation to use -no-cnv (#12845)
381603a77504ad1788965f694094540c1bed9ea2 65a69e6e1b6d55cd5f78f8bcdfaba8a8c59a8d96 Plamen Minev pacominev@gmail.com 2025-04-09T11:11:11+03:00 GitHub noreply@github.com 2025-04-09T10:11:11+02:00 ci: detach common from the library (#12827)
65a69e6e1b6d55cd5f78f8bcdfaba8a8c59a8d96 47277d6d1d0d515cff34292a1a78a0d1b7252350 Xuan-Son Nguyen son@huggingface.co 2025-04-09T10:09:53+02:00 GitHub noreply@github.com 2025-04-09T10:09:53+02:00 clip : do not print ftype (#12832)
47277d6d1d0d515cff34292a1a78a0d1b7252350 6e1c4cebdb697f925c523d3a969128d945161bdd Georgi Gerganov ggerganov@gmail.com 2025-04-09T10:54:42+03:00 GitHub noreply@github.com 2025-04-09T10:54:42+03:00 readme : add rpc backend (#12842)
6e1c4cebdb697f925c523d3a969128d945161bdd 0090950f679475c5ecaac2f7bca5049cca96492b Chenguang Li 757486878@qq.com 2025-04-09T14:04:14+08:00 GitHub noreply@github.com 2025-04-09T14:04:14+08:00 CANN: Support Opt CONV_TRANSPOSE_1D and ELU (#12786)
0090950f679475c5ecaac2f7bca5049cca96492b 7ecd780b1a1d5214b8d04c25ebfc194d310816ed Jeff Bolz jbolz@nvidia.com 2025-04-09T00:25:08-05:00 GitHub noreply@github.com 2025-04-09T07:25:08+02:00 vulkan: In coopmat2 mmq, load q4_k/q5_k scales through shared memory (#12833)
7ecd780b1a1d5214b8d04c25ebfc194d310816ed 7538246e7ce0606694c38055cc2fc9f60535be6c Jeff Bolz jbolz@nvidia.com 2025-04-09T00:12:57-05:00 GitHub noreply@github.com 2025-04-09T07:12:57+02:00 vulkan: Use fp16 for the flash attention P*V multiplication (#12783)
7538246e7ce0606694c38055cc2fc9f60535be6c b32efad2bc42460637c3a364c9554ea8217b3d7f Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-04-08T23:21:31+02:00 GitHub noreply@github.com 2025-04-08T23:21:31+02:00 cuda : add f32 to bf16 copy op (#12806)
b32efad2bc42460637c3a364c9554ea8217b3d7f a19b5cef16d885c44c635da4a5c97113c1577de8 Matt Clayton 156335168+mattjcly@users.noreply.github.com 2025-04-08T16:01:58-04:00 GitHub noreply@github.com 2025-04-08T22:01:58+02:00 llava: improve clip_ctx destructor to not memleak load_image_size (#12834)
a19b5cef16d885c44c635da4a5c97113c1577de8 78a1ba0a4f2bfed5b8b8e312592143d22e531698 Georgi Gerganov ggerganov@gmail.com 2025-04-08T19:54:51+03:00 GitHub noreply@github.com 2025-04-08T19:54:51+03:00 llama : fix FA when KV cache is not used (i.e. embeddings) (#12825)
78a1ba0a4f2bfed5b8b8e312592143d22e531698 2dabf759e7c8c827d38cdd18d0792070e6a4f4e1 Xuan-Son Nguyen son@huggingface.co 2025-04-08T18:37:06+02:00 GitHub noreply@github.com 2025-04-08T18:37:06+02:00 server : fix thread.join() on exit (#12831)
2dabf759e7c8c827d38cdd18d0792070e6a4f4e1 1d343b4069c74b2c7b19ae84260cd98aa2320a9a dm4 sunrisedm4@gmail.com 2025-04-08T21:49:13+08:00 GitHub noreply@github.com 2025-04-08T15:49:13+02:00 llava: add more helper functions to check projector types in clip context (#12824)
1d343b4069c74b2c7b19ae84260cd98aa2320a9a 8ca6e1c3a4deb6bb27ee294bfd5706098d94ae88 Prajwal B Mehendarkar prajwal.b.mehendarkar@ibm.com 2025-04-08T18:00:59+05:30 GitHub noreply@github.com 2025-04-08T14:30:59+02:00 arg : Including limits file on AIX (#12822)
8ca6e1c3a4deb6bb27ee294bfd5706098d94ae88 656babd6c21a3b9b3622324cfcc80a2ab78da25b characharm 123120856+characharm@users.noreply.github.com 2025-04-08T14:14:59+05:00 GitHub noreply@github.com 2025-04-08T11:14:59+02:00 server : webui : Improve Chat Input with Auto-Sizing Textarea (#12785)
656babd6c21a3b9b3622324cfcc80a2ab78da25b a226bc7a9ac50551f9f113808de0f0046837f188 Neo Zhang Jianyu jianyu.zhang@intel.com 2025-04-08T15:03:21+08:00 GitHub noreply@github.com 2025-04-08T15:03:21+08:00 Revert "sycl:remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor" (#12812)
a226bc7a9ac50551f9f113808de0f0046837f188 1466621e738779eefe1bb672e17dc55d63d166bb compilade git@compilade.net 2025-04-08T03:03:07-04:00 GitHub noreply@github.com 2025-04-08T09:03:07+02:00 gguf-py : support lazy tensor splitting (#12809)
1466621e738779eefe1bb672e17dc55d63d166bb 82974011f312057b446c27267105bd7ad3810599 Xuan-Son Nguyen son@huggingface.co 2025-04-07T23:06:44+02:00 GitHub noreply@github.com 2025-04-07T23:06:44+02:00 llama : Support llama 4 text-only (#12791)
82974011f312057b446c27267105bd7ad3810599 4ccea213bc629c4eef7b520f7f6c59ce9bbdaca0 lhez quic_lih@quicinc.com 2025-04-07T13:22:54-07:00 GitHub noreply@github.com 2025-04-07T13:22:54-07:00 opencl: better identify Adreno GPU (#12760)
4ccea213bc629c4eef7b520f7f6c59ce9bbdaca0 1a1ab7e7a4a9b6e6440c1d9965d2b9d1b7e7dafb stduhpf stephduh@live.fr 2025-04-07T17:47:08+02:00 GitHub noreply@github.com 2025-04-07T18:47:08+03:00 hellaswag: display estimated score confidence interval (#12797)
1a1ab7e7a4a9b6e6440c1d9965d2b9d1b7e7dafb a4e46e28f99d7f64f38d8328cdb6bee5a3a1cd03 Georgi Gerganov ggerganov@gmail.com 2025-04-07T13:18:07+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-07T18:44:17+03:00 cuda : fix HIP and MUSA BF16 (#0)
a4e46e28f99d7f64f38d8328cdb6bee5a3a1cd03 ff067dbcb9e6ca4ed464d3db999ff8e9c503498b Georgi Gerganov ggerganov@gmail.com 2025-04-07T12:32:39+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-07T18:44:17+03:00 sync : ggml
ff067dbcb9e6ca4ed464d3db999ff8e9c503498b 36ca8b362885e4b4984d72e348ba403568864a95 Georgi Gerganov ggerganov@gmail.com 2025-04-07T12:25:15+03:00 Georgi Gerganov ggerganov@gmail.com 2025-04-07T18:44:17+03:00 ggml : simplify Arm fp16 CPU logic (ggml/1177)
36ca8b362885e4b4984d72e348ba403568864a95 995083e4ed24933e6a289472f9de0f0b53ca5eca Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-04-04T21:05:12+02:00 Georgi Gerganov ggerganov@gmail.com 2025-04-07T18:44:17+03:00 CUDA: don't convert BF16 weights to FP32 (ggml/1174)
995083e4ed24933e6a289472f9de0f0b53ca5eca 518a01480eb3a7c80a4951b430db9dee55428310 cmdr2 secondary.cmdr2@gmail.com 2025-04-02T17:46:16+05:30 Georgi Gerganov ggerganov@gmail.com 2025-04-07T18:44:17+03:00 cpu: move all the operators into a separate c++ file (except mul_mat) (ggml/1167)
518a01480eb3a7c80a4951b430db9dee55428310 e391d3ee8ddae86be70c034de1082ad51c55e211 zhouwg zhouwg2000@gmail.com 2025-04-07T23:22:57+08:00 GitHub noreply@github.com 2025-04-07T17:22:57+02:00 sycl: remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor (#12734)
e391d3ee8ddae86be70c034de1082ad51c55e211 bd3f59f81289b920bcc597a208c14f55e39ed37e Xuan-Son Nguyen son@huggingface.co 2025-04-07T14:37:28+02:00 GitHub noreply@github.com 2025-04-07T15:37:28+03:00 ci : no curl on ggml-ci (#12796)
bd3f59f81289b920bcc597a208c14f55e39ed37e 52b3d71f128c1eeab39b918adf1f7a8f5e256619 Xuan-Son Nguyen son@huggingface.co 2025-04-07T13:35:19+02:00 GitHub noreply@github.com 2025-04-07T13:35:19+02:00 cmake : enable curl by default (#12761)
52b3d71f128c1eeab39b918adf1f7a8f5e256619 d0d5b2232b3826cac2c6e62d0244a474ba79860f zhouwg zhouwg2000@gmail.com 2025-04-07T19:34:14+08:00 GitHub noreply@github.com 2025-04-07T19:34:14+08:00 CANN: fix typo in ggml-cann (#12733)
d0d5b2232b3826cac2c6e62d0244a474ba79860f 916c83bfe7f8b08ada609c3b8e583cf5301e594b hipudding huafengchun@gmail.com 2025-04-07T17:10:36+08:00 GitHub noreply@github.com 2025-04-07T17:10:36+08:00 CANN: Refactor to reduce duplicate code (#12731)
916c83bfe7f8b08ada609c3b8e583cf5301e594b 0c74b04376b0b9efc096480fe10f866afc8d7c1c R0CKSTAR xiaodong.ye@mthreads.com 2025-04-06T21:23:54+08:00 GitHub noreply@github.com 2025-04-06T15:23:54+02:00 musa: fix compilation warnings in mp_22/31 (#12780)
0c74b04376b0b9efc096480fe10f866afc8d7c1c 80b717d493a9a5bae7167ad2384c12c60bb2ef20 Jeff Bolz jbolz@nvidia.com 2025-04-06T04:03:47-05:00 GitHub noreply@github.com 2025-04-06T11:03:47+02:00 vulkan: fix NaN issue in flash attention shader (#12776)
80b717d493a9a5bae7167ad2384c12c60bb2ef20 6bf28f0111ff9f21b3c1b1eace20c590281e7ba6 Jeff Bolz jbolz@nvidia.com 2025-04-06T03:47:13-05:00 GitHub noreply@github.com 2025-04-06T10:47:13+02:00 vulkan: Use unclamped loads for flash attention mask (#12720)
6bf28f0111ff9f21b3c1b1eace20c590281e7ba6 f1e3eb4249db68d97be352c0adf16eef7ae53795 0cc4m picard12@live.de 2025-04-05T18:04:03+02:00 GitHub noreply@github.com 2025-04-05T18:04:03+02:00 Vulkan: Tune Vulkan mmq int dot shader for performance (#12767)
f1e3eb4249db68d97be352c0adf16eef7ae53795 0364178ca2452bd241a4cdd4350d0d7a6ff8c4b4 Sergey Fedorov vital.had@gmail.com 2025-04-05T23:46:00+08:00 GitHub noreply@github.com 2025-04-05T17:46:00+02:00 common : fix includes in arg.cpp and gemma3-cli.cpp (#12766)
0364178ca2452bd241a4cdd4350d0d7a6ff8c4b4 c6ff5d2a8da2587cc78c9ede9171dfc3f076c757 Xuan-Son Nguyen son@huggingface.co 2025-04-05T17:17:40+02:00 GitHub noreply@github.com 2025-04-05T17:17:40+02:00 clip : refactor clip_init, add tests (#12757)
c6ff5d2a8da2587cc78c9ede9171dfc3f076c757 7a84777f42a9b3ba47db5d20b7662f8ddf92f652 エシュナヴァリシア 148695646+eternaphia@users.noreply.github.com 2025-04-05T21:31:42+08:00 GitHub noreply@github.com 2025-04-05T15:31:42+02:00 common: custom hf endpoint support (#12769)
7a84777f42a9b3ba47db5d20b7662f8ddf92f652 3e1d29348b5d77269f6931500dd1c1a729d429c8 Olivier Chafik ochafik@users.noreply.github.com 2025-04-04T13:16:39-07:00 GitHub noreply@github.com 2025-04-04T21:16:39+01:00 sync: minja (#12739)
3e1d29348b5d77269f6931500dd1c1a729d429c8 1be76e4620ddc99b3cbff0f206436117ae561047 Georgi Gerganov ggerganov@gmail.com 2025-04-04T21:48:10+03:00 GitHub noreply@github.com 2025-04-04T21:48:10+03:00 kv-cache : simplify + fix warning for recurrent models (#12756)
1be76e4620ddc99b3cbff0f206436117ae561047 b7723942970b70412793f1926a35cfb93d5c157c bandoti 141645996+bandoti@users.noreply.github.com 2025-04-04T14:05:12-03:00 GitHub noreply@github.com 2025-04-04T14:05:12-03:00 ci: add Linux cross-compile build (#12428)
b7723942970b70412793f1926a35cfb93d5c157c 23106f94ea2bc3da929afb7330655fd5515d08dc Nauful Shaikh nauful@gmail.com 2025-04-04T09:09:52-05:00 GitHub noreply@github.com 2025-04-04T16:09:52+02:00 server : webui : Upgrade daisyui, tailwindcss. (#12735)
23106f94ea2bc3da929afb7330655fd5515d08dc 94148ba330968bbfb8d9ecc67751bdc2218486cd nick huang nickhuang99@hotmail.com 2025-04-04T22:09:12+08:00 GitHub noreply@github.com 2025-04-04T16:09:12+02:00 gguf-split : --merge now respects --dry-run option (#12681)
94148ba330968bbfb8d9ecc67751bdc2218486cd 9ac4d611d05b03f961e627f4f399e4377bdb4ad3 Nicolò Scipione nicolo.scipione@codeplay.com 2025-04-04T16:00:46+02:00 GitHub noreply@github.com 2025-04-04T16:00:46+02:00 sycl: allow ggml-sycl configuration and compilation using Visual Studio project/solution (#12625)
9ac4d611d05b03f961e627f4f399e4377bdb4ad3 348888e0dc469a476f9e9eccb394893c513daab8 Ronny Brendel ronnyb@nvidia.com 2025-04-04T15:12:40+02:00 GitHub noreply@github.com 2025-04-04T10:12:40-03:00 cmake: fix ggml-shaders-gen compiler paths containing spaces (#12747)
348888e0dc469a476f9e9eccb394893c513daab8 74d4f5b041ad837153b0e90fc864b8290e01d8d5 Daniel Bevenius daniel.bevenius@gmail.com 2025-04-04T10:24:12+02:00 GitHub noreply@github.com 2025-04-04T10:24:12+02:00 docs : add XCFramework section to README.md [no ci] (#12746)
74d4f5b041ad837153b0e90fc864b8290e01d8d5 35e592eb30832187412360912ab8f2f5b7984df1 Jeff Bolz jbolz@nvidia.com 2025-04-04T00:54:35-05:00 GitHub noreply@github.com 2025-04-04T07:54:35+02:00 vulkan: Hybrid waitForFences/getFenceStatus to reduce fence latency (#12630)
35e592eb30832187412360912ab8f2f5b7984df1 7d7b1bafa7980603a61cb102879fe38a33f66c08 Jeff Bolz jbolz@nvidia.com 2025-04-04T00:53:20-05:00 GitHub noreply@github.com 2025-04-04T07:53:20+02:00 vulkan: set cmake minimum and project name in vulkan-shaders (#12744)
7d7b1bafa7980603a61cb102879fe38a33f66c08 c262beddf29f3f3be5bbbf167b56029a19876956 lhez quic_lih@quicinc.com 2025-04-03T22:18:17-07:00 GitHub noreply@github.com 2025-04-03T22:18:17-07:00 opencl: update doc for OpenCL (#12702)
c262beddf29f3f3be5bbbf167b56029a19876956 5dd5d1ab00d074e3b7c02ca3ae12f6bf3e86336a Gaurav Garg 52341457+gaugarg-nv@users.noreply.github.com 2025-04-03T21:50:29+05:30 GitHub noreply@github.com 2025-04-03T18:20:29+02:00 CUDA: Prefer vector flash decoding kernel for Gemma models (#12738)
5dd5d1ab00d074e3b7c02ca3ae12f6bf3e86336a 1c059995e080e37aeaacaae35850fc4c044b9dbe yumeyao yumeyao@gmail.com 2025-04-03T23:32:54+08:00 GitHub noreply@github.com 2025-04-03T18:32:54+03:00 vocab : use string_view::find() to avoid unnecessary looking up beyond the fragment range (#12706)
1c059995e080e37aeaacaae35850fc4c044b9dbe 2004644b7a5da6fe080e51861ab583480280f1d3 Jeff Bolz jbolz@nvidia.com 2025-04-03T10:08:26-05:00 GitHub noreply@github.com 2025-04-03T10:08:26-05:00 vulkan: Fix missing cmake logic for dot product extension (#12721)
2004644b7a5da6fe080e51861ab583480280f1d3 5f696e88e0eb490c8028421d3c5419df9dcf5385 Atharva Dubey atharva.dubey@codeplay.com 2025-04-03T13:12:39+01:00 GitHub noreply@github.com 2025-04-03T15:12:39+03:00 ci : add env variable in ggml-ci and document the same in SYCL.md (#12736)
5f696e88e0eb490c8028421d3c5419df9dcf5385 193c3e03a63ccda3ac3d6a2999e41e6d1414fe23 R0CKSTAR xiaodong.ye@mthreads.com 2025-04-03T19:51:35+08:00 GitHub noreply@github.com 2025-04-03T13:51:35+02:00 sync : minja (inclusionAI/Ling) and update tests (#12699)
193c3e03a63ccda3ac3d6a2999e41e6d1414fe23 65cfe136a0793b2fdf5af8bdb1ab2cf10c3e1d2e a3sh 38979186+A3shTnT@users.noreply.github.com 2025-04-03T15:32:55+08:00 GitHub noreply@github.com 2025-04-03T09:32:55+02:00 fix MUSA compiler warning (#12704)
65cfe136a0793b2fdf5af8bdb1ab2cf10c3e1d2e 3f9da22c2b21a2cef216de50006436ef1cab8764 Chenguang Li 757486878@qq.com 2025-04-03T15:18:08+08:00 GitHub noreply@github.com 2025-04-03T15:18:08+08:00 CANN: Support operator SIN COS ARGMAX (#12709)
3f9da22c2b21a2cef216de50006436ef1cab8764 2a0dc97e56eac6db0a4016f0b45da6d0a0055ef2 Alan Gray agray3@users.noreply.github.com 2025-04-03T02:31:15+01:00 GitHub noreply@github.com 2025-04-03T03:31:15+02:00 Simplify and improve CUDA graphs through use of indirect copy pointers (#9017)
2a0dc97e56eac6db0a4016f0b45da6d0a0055ef2 97a20c012be2f9bfbeee0209405a31001f93ccf9 hipudding huafengchun@gmail.com 2025-04-03T08:49:51+08:00 GitHub noreply@github.com 2025-04-03T08:49:51+08:00 CANN: Fix failed test cases (#12708)
97a20c012be2f9bfbeee0209405a31001f93ccf9 f01bd02376f919b05ee635f438311be8dfc91d7c lhez quic_lih@quicinc.com 2025-04-02T17:01:42-07:00 GitHub noreply@github.com 2025-04-02T17:01:42-07:00 opencl: use `max_alloc_size` in backend ctx instead of querying again (#12705)
f01bd02376f919b05ee635f438311be8dfc91d7c 6f3bd38640f07e4dec7f145d2fbf093ce48c9544 Jeff Bolz jbolz@nvidia.com 2025-04-02T14:25:08-05:00 GitHub noreply@github.com 2025-04-02T14:25:08-05:00 vulkan: Implement split_k for coopmat2 flash attention. (#12627)
6f3bd38640f07e4dec7f145d2fbf093ce48c9544 be0a0f8cae039e2286f757612accebfb8f21b36e bandoti 141645996+bandoti@users.noreply.github.com 2025-04-02T14:56:26-03:00 GitHub noreply@github.com 2025-04-02T14:56:26-03:00 cmake: remove caching from vulkan coopmat checks (#12719)
be0a0f8cae039e2286f757612accebfb8f21b36e 92e3006bb69dfeb656ccf5c7c1c1efadb03c88c2 Jeff Bolz jbolz@nvidia.com 2025-04-02T12:40:32-05:00 GitHub noreply@github.com 2025-04-02T19:40:32+02:00 vulkan: Implement grouped query attention in the coopmat2 FA shader (#12559)
92e3006bb69dfeb656ccf5c7c1c1efadb03c88c2 833e2b7409211a07df97716998c5002526642652 0cc4m picard12@live.de 2025-04-02T19:12:30+02:00 GitHub noreply@github.com 2025-04-02T19:12:30+02:00 Vulkan: Fix mmq int dot float cache size (#12722)
833e2b7409211a07df97716998c5002526642652 e0e912f49b3195ef9d0c51378629ba03c9b972da Georgi Gerganov ggerganov@gmail.com 2025-04-02T16:38:54+03:00 GitHub noreply@github.com 2025-04-02T16:38:54+03:00 model : print tensor size during load (#12711)
e0e912f49b3195ef9d0c51378629ba03c9b972da a10b36c91a091f4606710fba4e9327fd71e0e738 Diego Devesa slarengh@gmail.com 2025-04-02T14:52:01+02:00 GitHub noreply@github.com 2025-04-02T14:52:01+02:00 llama : add option to override model tensor buffers (#11397)
a10b36c91a091f4606710fba4e9327fd71e0e738 83a88bd6affbe148a622ac730952ac5b8b585979 Georgi Gerganov ggerganov@gmail.com 2025-04-02T14:32:59+03:00 GitHub noreply@github.com 2025-04-02T14:32:59+03:00 llama : refactor kv cache guard (#12695)
83a88bd6affbe148a622ac730952ac5b8b585979 42eb248f46e1175349e553b6eda6cb63027d74d1 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-04-02T11:21:48+02:00 GitHub noreply@github.com 2025-04-02T11:21:48+02:00 vocab : BailingMoE : change possessive quantifiers to greedy (#12677)
42eb248f46e1175349e553b6eda6cb63027d74d1 9bacd6b37461608385360fd64326c13247ccf18e Xuan-Son Nguyen son@huggingface.co 2025-04-02T09:58:34+02:00 GitHub noreply@github.com 2025-04-02T09:58:34+02:00 common : remove json.hpp from common.cpp (#12697)
9bacd6b37461608385360fd64326c13247ccf18e 267c1399f15a278ec8c3cdcf9c90dc94151fbc38 Chenguang Li 757486878@qq.com 2025-04-02T15:22:13+08:00 GitHub noreply@github.com 2025-04-02T15:22:13+08:00 [CANN] get_rows and dup optimization (#12671)
267c1399f15a278ec8c3cdcf9c90dc94151fbc38 f423981ac806bf031d83784bcb47d2721bc70f97 Xuan-Son Nguyen son@huggingface.co 2025-04-01T23:44:05+02:00 GitHub noreply@github.com 2025-04-01T23:44:05+02:00 common : refactor downloading system, handle mmproj with -hf option (#12694)
f423981ac806bf031d83784bcb47d2721bc70f97 e39e727e9a3daec7082b9c59540a429ad85914af Junil Kim logyourself@gmail.com 2025-04-02T01:54:34+09:00 GitHub noreply@github.com 2025-04-01T09:54:34-07:00 opencl : fix memory allocation size (#12649)
e39e727e9a3daec7082b9c59540a429ad85914af 5936a616e46cffad4579ccaff8f8fa315809da4c jklincn jklincn@outlook.com 2025-04-01T20:54:28+08:00 GitHub noreply@github.com 2025-04-01T14:54:28+02:00 llama : use LLM_KV_GENERAL_FILE_TYPE instead of gguf_find_key (#12672)
5936a616e46cffad4579ccaff8f8fa315809da4c 3fd072a54001a908c54e81fd2e82b682ecfdd475 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-04-01T14:37:13+02:00 GitHub noreply@github.com 2025-04-01T14:37:13+02:00 convert : BailingMoE : fix qkv split when head_dim is 0 (#12687)
3fd072a54001a908c54e81fd2e82b682ecfdd475 a6f32f0b3437ac79827ffb55521cb839343324ab Georgi Gerganov ggerganov@gmail.com 2025-04-01T14:57:19+03:00 GitHub noreply@github.com 2025-04-01T14:57:19+03:00 metal : use F32 prec in FA kernels (#12688)
a6f32f0b3437ac79827ffb55521cb839343324ab 2bb3597e426ce092c3e10ae0bdd876963765e6ed R0CKSTAR xiaodong.ye@mthreads.com 2025-04-01T19:12:53+08:00 GitHub noreply@github.com 2025-04-01T13:12:53+02:00 Fix clang warning in gguf_check_reserved_keys (#12686)
2bb3597e426ce092c3e10ae0bdd876963765e6ed 82939705421f4ef27e924879fdeed6d7b5f6d769 Wagner Bruna wbruna@users.noreply.github.com 2025-04-01T06:38:07-03:00 GitHub noreply@github.com 2025-04-01T11:38:07+02:00 vulkan: fix build when glslc doesn't support coopmat (#12683)
82939705421f4ef27e924879fdeed6d7b5f6d769 8bbf26083d93274240a20d16cda324441c57fcc6 Romain Biessy romain.biessy@codeplay.com 2025-04-01T10:24:29+02:00 GitHub noreply@github.com 2025-04-01T16:24:29+08:00 SYCL: Rename oneMKL to oneMath (#12192)
8bbf26083d93274240a20d16cda324441c57fcc6 35782aeedb4f65140c0d522070901843213ff157 Akarshan Biswas akarshan@menlo.ai 2025-04-01T13:41:39+05:30 GitHub noreply@github.com 2025-04-01T10:11:39+02:00 SYCL: switch to SYCL namespace (#12674)
35782aeedb4f65140c0d522070901843213ff157 c80a7759dab10657b9b6c3e87eef988a133b9b6a Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-31T23:09:48+02:00 GitHub noreply@github.com 2025-03-31T23:09:48+02:00 convert : BailingMoE : avoid setting rope_dim to 0 (#12678)
c80a7759dab10657b9b6c3e87eef988a133b9b6a 250d7953e829ff0e16074510fef0e7cec696461b Daniel Bevenius daniel.bevenius@gmail.com 2025-03-31T18:40:56+02:00 GitHub noreply@github.com 2025-03-31T18:40:56+02:00 vocab : add special infill tokens for CodeLlama (#11850)
250d7953e829ff0e16074510fef0e7cec696461b 403fbacbbcd3650abcc53e84f5e6bf88e51dd6f5 a3sh 38979186+A3shTnT@users.noreply.github.com 2025-04-01T00:05:13+08:00 GitHub noreply@github.com 2025-03-31T18:05:13+02:00 ggml : faster ssm scan (#10558)
403fbacbbcd3650abcc53e84f5e6bf88e51dd6f5 a8a1f3356786cbf8bcc3422e3c8737fc33b453e7 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-31T16:36:25+02:00 GitHub noreply@github.com 2025-03-31T16:36:25+02:00 convert : Qwerky : use lora_rank_tokenshift and lora_rank_decay if present (#12667)
a8a1f3356786cbf8bcc3422e3c8737fc33b453e7 1790e7315726ceda256ea91eaa66fc36b63f6067 0cc4m picard12@live.de 2025-03-31T14:37:01+02:00 GitHub noreply@github.com 2025-03-31T14:37:01+02:00 Vulkan: Add DP4A MMQ and Q8_1 quantization shader (#12135)
1790e7315726ceda256ea91eaa66fc36b63f6067 0114a32da0adc21dc04f05278cd8e3f67ba5f8c7 Georgi Gerganov ggerganov@gmail.com 2025-03-31T15:05:30+03:00 Georgi Gerganov ggerganov@gmail.com 2025-03-31T15:07:32+03:00 cmake : fix whitespace (#0)
0114a32da0adc21dc04f05278cd8e3f67ba5f8c7 a7724480fd7eb5451981ba7458c1b3829f4897d1 Georgi Gerganov ggerganov@gmail.com 2025-03-31T14:59:21+03:00 Georgi Gerganov ggerganov@gmail.com 2025-03-31T15:07:32+03:00 sync : ggml
a7724480fd7eb5451981ba7458c1b3829f4897d1 1a859490670c271e8e992c83cb11459ea0c3580c Sandro Hanea 40202887+sandrohanea@users.noreply.github.com 2025-03-31T12:44:36+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-31T15:07:32+03:00 cmake: improve Vulkan cooperative matrix support checks (whisper/2966)
1a859490670c271e8e992c83cb11459ea0c3580c 6c02a032fa21d69e881ef9a5c94ba28ebaf1d749 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-31T11:28:30+02:00 GitHub noreply@github.com 2025-03-31T11:28:30+02:00 llava : proper description fix (#12668)
6c02a032fa21d69e881ef9a5c94ba28ebaf1d749 f52d59d771dc231fc2ac39adacf157ddefc97730 Akarshan Biswas akarshan@menlo.ai 2025-03-31T14:55:24+05:30 GitHub noreply@github.com 2025-03-31T11:25:24+02:00 SYCL: Remove misleading ggml_sycl_op_flatten function (#12387)
f52d59d771dc231fc2ac39adacf157ddefc97730 52de2e594979be52f0da92601747aa44a13e50e4 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-31T11:07:07+02:00 GitHub noreply@github.com 2025-03-31T11:07:07+02:00 llava : fix clip loading GGUFs with missing description (#12660)
52de2e594979be52f0da92601747aa44a13e50e4 2c3f8b850a4a6cff0f5dda2135c03fc81d33ed8b marcoStocchi marcostocchi77@gmail.com 2025-03-31T10:20:30+02:00 GitHub noreply@github.com 2025-03-31T11:20:30+03:00 tts : remove printfs (#12640)
2c3f8b850a4a6cff0f5dda2135c03fc81d33ed8b 4663bd353c61c1136cd8a97b9908755e4ab30cec Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-30T22:21:03+02:00 GitHub noreply@github.com 2025-03-30T22:21:03+02:00 llama : support BailingMoE (Ling) (#12634)
4663bd353c61c1136cd8a97b9908755e4ab30cec b3de7cac732e7dc0e10e1bb07502a500b0ee9022 Georgi Gerganov ggerganov@gmail.com 2025-03-30T22:04:04+03:00 GitHub noreply@github.com 2025-03-30T22:04:04+03:00 metal : use constexpr in FA kernels + fix typedef (#12659)
b3de7cac732e7dc0e10e1bb07502a500b0ee9022 7242dd96756472f90ba41db4e5ebb3969dfc7e7c Juyoung Suk juyoung.suk@trillionlabs.co 2025-03-31T03:38:33+09:00 GitHub noreply@github.com 2025-03-30T20:38:33+02:00 llama : add Trillion 7B model support (#12556)
7242dd96756472f90ba41db4e5ebb3969dfc7e7c 492d7f1ff77e116e44962b832cc3db24cfc46d24 Sergei Vorobyov sergei.vorobyov01@gmail.com 2025-03-30T21:12:03+03:00 GitHub noreply@github.com 2025-03-30T20:12:03+02:00 llama-chat : Add Yandex instruct model template support (#12621)
492d7f1ff77e116e44962b832cc3db24cfc46d24 d3f1f0acfbf8aaafd2ce494309a10a7cc92042c8 R0CKSTAR xiaodong.ye@mthreads.com 2025-03-30T16:59:38+08:00 GitHub noreply@github.com 2025-03-30T10:59:38+02:00 musa: fix all warnings, re-enable `-DLLAMA_FATAL_WARNINGS=ON` in ci and update doc (#12611)
d3f1f0acfbf8aaafd2ce494309a10a7cc92042c8 360dc22c007c7b926d0721930e279d87317d8466 Georgi Gerganov ggerganov@gmail.com 2025-03-29T15:37:54+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-30T08:33:31+03:00 sync : ggml
360dc22c007c7b926d0721930e279d87317d8466 a62d7fa7a979d21bcd760f1d7c19a66e55611f44 Xuan-Son Nguyen son@huggingface.co 2025-03-29T11:59:56+01:00 Georgi Gerganov ggerganov@gmail.com 2025-03-30T08:33:31+03:00 cpu : rm unused variable (ggml/1166)
a62d7fa7a979d21bcd760f1d7c19a66e55611f44 e408d4351a4ad143656672479d8ae1479306ce31 cmdr2 secondary.cmdr2@gmail.com 2025-03-29T11:37:13+05:30 Georgi Gerganov ggerganov@gmail.com 2025-03-30T08:33:31+03:00 cpu: de-duplicate some of the operators and refactor (ggml/1144)
e408d4351a4ad143656672479d8ae1479306ce31 3891e183c61c1e25c2c61afe68d7b95019ea3f89 Daniel Bevenius daniel.bevenius@gmail.com 2025-03-24T09:53:38+01:00 Georgi Gerganov ggerganov@gmail.com 2025-03-30T08:33:31+03:00 ggml : add logging for native build options/vars (whisper/2935)
3891e183c61c1e25c2c61afe68d7b95019ea3f89 af6ae1efb27a9a7c3f7f7f84639d2243f7303ac1 Daniel Bevenius daniel.bevenius@gmail.com 2025-03-20T07:02:18+01:00 Georgi Gerganov ggerganov@gmail.com 2025-03-30T08:33:31+03:00 examples : command.wasm updates (whisper/2904)
af6ae1efb27a9a7c3f7f7f84639d2243f7303ac1 0bb2919335d00ff0bc79d5015da95c422de51f03 Xuan-Son Nguyen son@huggingface.co 2025-03-30T00:07:37+01:00 GitHub noreply@github.com 2025-03-30T00:07:37+01:00 llama : fix non-causal mask for gemma 3 (#12615)
0bb2919335d00ff0bc79d5015da95c422de51f03 a69f8463510a70fe0c85701b8c7ede283a0d1a7d Djip007 3705339+Djip007@users.noreply.github.com 2025-03-29T14:07:37+01:00 GitHub noreply@github.com 2025-03-29T14:07:37+01:00 llama : change cpu_buft_list order: ACCEL -> GPU host -> CPU extra -> CPU (#12632)
a69f8463510a70fe0c85701b8c7ede283a0d1a7d d07a0d7a79bafb01fb3b6c8992121e75c57c0c2f Jay BusyJay@users.noreply.github.com 2025-03-29T18:04:58+08:00 GitHub noreply@github.com 2025-03-29T11:04:58+01:00 cmake : fix ccache conflict (#12522)
d07a0d7a79bafb01fb3b6c8992121e75c57c0c2f 3714c3ee1a62ed64ac328ec7d699410ad1219150 hipudding huafengchun@gmail.com 2025-03-29T18:03:28+08:00 GitHub noreply@github.com 2025-03-29T11:03:28+01:00 CANN : remove clang-format in ggml-cann (#12607)
3714c3ee1a62ed64ac328ec7d699410ad1219150 b4ae50810e4304d052e630784c14bde7e79e4132 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-28T22:13:02+01:00 GitHub noreply@github.com 2025-03-28T22:13:02+01:00 llama : fix incorrect Qwen2Moe ffn_moe_out graph callback (#12631)
b4ae50810e4304d052e630784c14bde7e79e4132 b86f6007234da4bff51a3ebef2bdb952b52059c6 Georgi Gerganov ggerganov@gmail.com 2025-03-28T20:21:59+02:00 GitHub noreply@github.com 2025-03-28T20:21:59+02:00 metal : improve FA + improve MoE (#12612)
b86f6007234da4bff51a3ebef2bdb952b52059c6 dd373dd3bf81eced3e711fb7cb49123a6105933e Icenowy Zheng uwu@icenowy.me 2025-03-29T01:51:06+08:00 GitHub noreply@github.com 2025-03-28T14:51:06-03:00 vulkan: fix coopmat shader generation when cross-compiling (#12272)
dd373dd3bf81eced3e711fb7cb49123a6105933e 5d01670266859444366e4f333ade5e0e5e2ae63d Johannes Gäßler johannesg@5d6.de 2025-03-28T18:08:52+01:00 GitHub noreply@github.com 2025-03-28T18:08:52+01:00 llama: fix error on bad grammar (#12628)
5d01670266859444366e4f333ade5e0e5e2ae63d ef03229ff423dd1991f4f44ef1352f03334d86eb Benson Wong mostlygeek@gmail.com 2025-03-28T01:05:44-07:00 GitHub noreply@github.com 2025-03-28T10:05:44+02:00 server : include speculative decoding stats when timings_per_token is enabled (#12603)
ef03229ff423dd1991f4f44ef1352f03334d86eb 13731766db91ec927c1b61bf502ac7a9be2b11b9 Radoslav Gerganov rgerganov@gmail.com 2025-03-28T09:44:13+02:00 GitHub noreply@github.com 2025-03-28T09:44:13+02:00 rpc : update README for cache usage (#12620)
13731766db91ec927c1b61bf502ac7a9be2b11b9 ab6ab8f809bd514d88e02a63869b4d619f13fa86 amritahs-ibm amritahs@linux.vnet.ibm.com 2025-03-28T13:13:22+05:30 GitHub noreply@github.com 2025-03-28T09:43:22+02:00 llamafile : ppc64le GEMV forwarding for FP32. (#12594)
ab6ab8f809bd514d88e02a63869b4d619f13fa86 2099a9d5dbdcd2841d02dd396a71d0de70bf4490 Radoslav Gerganov rgerganov@gmail.com 2025-03-28T08:18:04+02:00 GitHub noreply@github.com 2025-03-28T08:18:04+02:00 rpc : send hash when tensor data is above some fixed threshold (#12496)
2099a9d5dbdcd2841d02dd396a71d0de70bf4490 296901983700f3c37449bcb555d85d27150a679d Piotr piotr.stankiewicz@docker.com 2025-03-27T23:41:04+01:00 GitHub noreply@github.com 2025-03-27T23:41:04+01:00 server : Support listening on a unix socket (#12613)
296901983700f3c37449bcb555d85d27150a679d 5dec47dcd411fdf815a3708fd6194e2b13d19006 Georgi Gerganov ggerganov@gmail.com 2025-03-27T23:09:05+02:00 GitHub noreply@github.com 2025-03-27T23:09:05+02:00 media : add SVG logo [no ci] (#12616)
5dec47dcd411fdf815a3708fd6194e2b13d19006 f125b8dccff34439a26bf750c9edef358c48c1f8 lhez quic_lih@quicinc.com 2025-03-27T08:08:08-07:00 GitHub noreply@github.com 2025-03-27T08:08:08-07:00 opencl: add multi and vision rope, `gelu_quick` and `im2col` (#12600)
f125b8dccff34439a26bf750c9edef358c48c1f8 953c2a62cf487e618140f3ea18d94e3b0257af93 Si1w 139008732+Si1w@users.noreply.github.com 2025-03-27T10:49:15Z GitHub noreply@github.com 2025-03-27T12:49:15+02:00 llama : add PLM GGUF Conversion & Inference Support (#12457)
953c2a62cf487e618140f3ea18d94e3b0257af93 d5c6309d91cb22ebc947920f92eb686d92f84eae HighDoping highdoping@gmail.com 2025-03-27T18:43:33+08:00 GitHub noreply@github.com 2025-03-27T11:43:33+01:00 model : restore support for T5Encoder (#12590)
d5c6309d91cb22ebc947920f92eb686d92f84eae 029c693fdcdc80e8508553df61a4337bb5fe49a9 Csaba Kecskemeti csaba.kecskemeti@gmail.com 2025-03-27T03:11:23-07:00 GitHub noreply@github.com 2025-03-27T11:11:23+01:00 convert : Support Qwen2_5_VLForConditionalGeneration (#12595)
029c693fdcdc80e8508553df61a4337bb5fe49a9 771d84371c0785c2554aef9a47cdd551b8c7ceaf Georgi Gerganov ggerganov@gmail.com 2025-03-27T09:36:13+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-27T10:09:29+02:00 sync : ggml
771d84371c0785c2554aef9a47cdd551b8c7ceaf df0665a483b08da1c9b55534b624ae4e1fe89767 Georgi Gerganov ggerganov@gmail.com 2025-03-27T09:22:30+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-27T10:09:29+02:00 scripts : update sync + fix cmake merge
df0665a483b08da1c9b55534b624ae4e1fe89767 0306aad1ca89a92fdb33450b35547b90b92f5dbe Georgi Gerganov ggerganov@gmail.com 2025-03-27T09:01:21+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-27T09:04:38+02:00 sync : ggml
0306aad1ca89a92fdb33450b35547b90b92f5dbe c7b43ab60855f752ae79937fb93d561bc30b69a4 Georgi Gerganov ggerganov@gmail.com 2025-03-27T09:00:57+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-27T09:04:38+02:00 cmake : sync/merge PowerPC build commands (#0)
c7b43ab60855f752ae79937fb93d561bc30b69a4 24feaec05792b972d5ff3e2b12d9237ebd50d1ac amritahs-ibm amritahs@linux.vnet.ibm.com 2025-03-27T12:21:47+05:30 GitHub noreply@github.com 2025-03-27T08:51:47+02:00 llamafile : ppc64le MMA implementation for Q4_0. (#12489)
24feaec05792b972d5ff3e2b12d9237ebd50d1ac f28bc4c286c453cd8385388eea057a404fdb6402 xctan xc-tan@outlook.com 2025-03-27T14:38:34+08:00 GitHub noreply@github.com 2025-03-27T08:38:34+02:00 ggml : riscv: add 128-bit RVV support (#12530)
f28bc4c286c453cd8385388eea057a404fdb6402 f17a3bb4e8b0aa24c0f86636d234aca7dc2cfa01 Georgi Gerganov ggerganov@gmail.com 2025-03-27T08:24:10+02:00 GitHub noreply@github.com 2025-03-27T08:24:10+02:00 llama : make loras compatible with repacking (#12593)
f17a3bb4e8b0aa24c0f86636d234aca7dc2cfa01 bd40678df768ab189b26b8b03e3de4062e3b71a3 Akarshan Biswas akarshan@menlo.ai 2025-03-27T07:16:00+05:30 GitHub noreply@github.com 2025-03-27T09:46:00+08:00 SYCL: implement memset ggml backend buffer interface (#12580)
bd40678df768ab189b26b8b03e3de4062e3b71a3 b3298fa47a2d56ae892127ea038942ab1cada190 Slobodan Josic 127323561+slojosic-amd@users.noreply.github.com 2025-03-26T23:46:30+01:00 GitHub noreply@github.com 2025-03-26T23:46:30+01:00 HIP: Add support for RDNA4 targets (#12372)
b3298fa47a2d56ae892127ea038942ab1cada190 2447ad8a981253a2b8e9f4b31cc8e7fdff83423e Georgi Gerganov ggerganov@gmail.com 2025-03-26T21:38:38+02:00 GitHub noreply@github.com 2025-03-26T21:38:38+02:00 metal : refactor mat-vec code (#12569)
2447ad8a981253a2b8e9f4b31cc8e7fdff83423e 02082f1519565fc7b49de211b28bc5404a69209b Michał Moskal michal@moskal.me 2025-03-26T11:06:09-07:00 GitHub noreply@github.com 2025-03-26T11:06:09-07:00 upgrade to llguidance 0.7.10 (#12576)
02082f1519565fc7b49de211b28bc5404a69209b df4d20cd53d5bb6fc21c1dc65f026d53b566d097 Ivy233 952254420@qq.com 2025-03-26T22:06:04+08:00 GitHub noreply@github.com 2025-03-26T15:06:04+01:00 clip: Fix llama-llava-clip-quantize-cli quantization error under CUDA backend (#12566)
df4d20cd53d5bb6fc21c1dc65f026d53b566d097 5ed38b6852bd509d56acfdae54bceec8ab3cc396 Georgi Gerganov ggerganov@gmail.com 2025-03-26T14:21:05+02:00 GitHub noreply@github.com 2025-03-26T08:21:05-04:00 convert : fix squeeze for ssm_conv tensors (#12573)
5ed38b6852bd509d56acfdae54bceec8ab3cc396 fd7855f8f522ab26681b25ae506ff99c654e2dd5 Georgi Gerganov ggerganov@gmail.com 2025-03-26T13:02:00+02:00 GitHub noreply@github.com 2025-03-26T13:02:00+02:00 ggml : fix MUL_MAT_ID repack with Q8_K (#12544)
fd7855f8f522ab26681b25ae506ff99c654e2dd5 53af4dba425768fd507ce6b45873cfbb3df2295f R0CKSTAR xiaodong.ye@mthreads.com 2025-03-26T15:09:48+08:00 GitHub noreply@github.com 2025-03-26T09:09:48+02:00 doc: [MUSA] minor changes (#12583)
53af4dba425768fd507ce6b45873cfbb3df2295f ef19c71769681a0b3dde6bc90911728376e5d236 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-25T23:03:10+01:00 GitHub noreply@github.com 2025-03-25T23:03:10+01:00 convert: fix Mistral3/Gemma3 model hparams init (#12571)
ef19c71769681a0b3dde6bc90911728376e5d236 053b3f9aae63151732eccf6b7408c6418ba8746e Eric Curtin ecurtin@redhat.com 2025-03-25T17:46:11Z GitHub noreply@github.com 2025-03-25T18:46:11+01:00 run: de-duplicate fmt and format functions and optimize (#11596)
053b3f9aae63151732eccf6b7408c6418ba8746e e2f560175a195f63c3276972a3d1caec0bd13e05 Dan Johansson dan.johansson@arm.com 2025-03-25T12:10:18+01:00 GitHub noreply@github.com 2025-03-25T13:10:18+02:00 ggml-cpu : update KleidiAI to v1.5.0 (#12568)
e2f560175a195f63c3276972a3d1caec0bd13e05 36ee06dd2dcf8d3d1157ebe36366d7670770e75f Akarshan Biswas akarshan@menlo.ai 2025-03-25T16:10:18+05:30 GitHub noreply@github.com 2025-03-25T18:40:18+08:00 SYCL: disable Q4_0 reorder optimization (#12560)
36ee06dd2dcf8d3d1157ebe36366d7670770e75f 3cd3a395323fa9cdf6ecfa1fea290bf228d4e856 Dan Johansson dan.johansson@arm.com 2025-03-25T10:35:20+01:00 GitHub noreply@github.com 2025-03-25T11:35:20+02:00 docs : add build instructions for KleidiAI (#12563)
3cd3a395323fa9cdf6ecfa1fea290bf228d4e856 2d77d88e70d017cd82c3f1a4517e3102e2028ac4 R0CKSTAR xiaodong.ye@mthreads.com 2025-03-25T15:45:08+08:00 GitHub noreply@github.com 2025-03-25T09:45:08+02:00 ci: [MUSA] add CI and update doc (#12562)
2d77d88e70d017cd82c3f1a4517e3102e2028ac4 c95fa362b3587d1822558f7e28414521075f254f Georgi Gerganov ggerganov@gmail.com 2025-03-25T09:19:23+02:00 GitHub noreply@github.com 2025-03-25T09:19:23+02:00 context : fix worst-case reserve outputs (#12545)
c95fa362b3587d1822558f7e28414521075f254f 2b65ae30299b9c67e25c51ee567e9a2ef22279ab Akarshan Biswas akarshan@menlo.ai 2025-03-24T23:05:38+05:30 GitHub noreply@github.com 2025-03-24T19:35:38+02:00 ci: [SYCL] ggml-ci Use main GPU and enable sysman (#12547)
2b65ae30299b9c67e25c51ee567e9a2ef22279ab 48d7021c61ceda6fcf1a7294d2115b8e1a53ae95 lhez quic_lih@quicinc.com 2025-03-24T09:20:47-07:00 GitHub noreply@github.com 2025-03-24T09:20:47-07:00 opencl: simplify kernel embedding logic in cmakefile (#12503)
48d7021c61ceda6fcf1a7294d2115b8e1a53ae95 3361e2deba0f24942877559b35c430dec68528c3 Akarshan Biswas akarshan@menlo.ai 2025-03-24T18:28:32+05:30 GitHub noreply@github.com 2025-03-24T14:58:32+02:00 CI: fix SYCL build (#12546)
3361e2deba0f24942877559b35c430dec68528c3 00d53800e00bb22a26bf710fa6bd1150e412cc1d Tei Home taiteitonghome@proton.me 2025-03-24T19:02:26+08:00 GitHub noreply@github.com 2025-03-24T11:02:26Z docs: update: improve the Fedoa CUDA guide (#12536)
00d53800e00bb22a26bf710fa6bd1150e412cc1d 7ea75035b67f44c22ed7039967f718011fd35ce5 compilade git@compilade.net 2025-03-24T06:47:24-04:00 GitHub noreply@github.com 2025-03-24T11:47:24+01:00 llama-vocab : add SuperBPE pre-tokenizer (#12532)
7ea75035b67f44c22ed7039967f718011fd35ce5 c54f6b7988b63714b87b0390e01a1e69aee79a12 R0CKSTAR xiaodong.ye@mthreads.com 2025-03-24T18:28:34+08:00 GitHub noreply@github.com 2025-03-24T11:28:34+01:00 CUDA: Fix clang warnings (#12540)
c54f6b7988b63714b87b0390e01a1e69aee79a12 9b169a4d4e01af7bc07a6981b53b27c18c9470d8 Prajwal B Mehendarkar prajwal.b.mehendarkar@ibm.com 2025-03-24T15:47:10+05:30 GitHub noreply@github.com 2025-03-24T12:17:10+02:00 mmap : skip resource limit checks on AIX (#12541)
9b169a4d4e01af7bc07a6981b53b27c18c9470d8 77f9c6bbe55fccd9ea567794024cb80943947901 Jeff Bolz jbolz@nvidia.com 2025-03-24T01:56:17-05:00 GitHub noreply@github.com 2025-03-24T07:56:17+01:00 vulkan: fix mul_mat_vec failure in backend tests (#12529)
77f9c6bbe55fccd9ea567794024cb80943947901 18b663d8e4ef352a9a15ff15d695fc3258801d60 Marius Gerdes 141485318+mglambda@users.noreply.github.com 2025-03-23T19:30:26+01:00 GitHub noreply@github.com 2025-03-23T19:30:26+01:00 server : Add verbose output to OAI compatible chat endpoint. (#12246)
18b663d8e4ef352a9a15ff15d695fc3258801d60 fbdfefe74e736f1a3687283c25ac21b11ba07b2e Lars Sonchocky-Helldorf lars.sonchocky-helldorf@hamburg.de 2025-03-23T09:21:48+01:00 GitHub noreply@github.com 2025-03-23T10:21:48+02:00 install : add macports (#12518)
fbdfefe74e736f1a3687283c25ac21b11ba07b2e ba932dfb50cc694645b1a148c72f8c06ee080b17 Xuan-Son Nguyen son@huggingface.co 2025-03-22T23:28:19+01:00 GitHub noreply@github.com 2025-03-22T23:28:19+01:00 llama : gemma3 : use output tensor if it exists in model weight (#12506)
ba932dfb50cc694645b1a148c72f8c06ee080b17 fac63a3d786b2a0f97876c30add02cb525a9648e Georgi Gerganov ggerganov@gmail.com 2025-03-22T16:23:26+02:00 GitHub noreply@github.com 2025-03-22T16:23:26+02:00 ggml : fix quantized cpy op (#12310)
fac63a3d786b2a0f97876c30add02cb525a9648e eddfb438502bd5d1014d63a812e9b6d03d326f8c R0CKSTAR xiaodong.ye@mthreads.com 2025-03-22T17:11:37+08:00 GitHub noreply@github.com 2025-03-22T10:11:37+01:00 musa: refine compute capability (#12493)
eddfb438502bd5d1014d63a812e9b6d03d326f8c 4375415b4abf94fb36a5fd15f233ac0ee23c0bd1 Jeff Bolz jbolz@nvidia.com 2025-03-22T03:40:11-05:00 GitHub noreply@github.com 2025-03-22T09:40:11+01:00 vulkan: Optimize mul_mat_vec p021 and nc shaders (#12505)
4375415b4abf94fb36a5fd15f233ac0ee23c0bd1 30c42ef5cbb2db756eff9aadc6b6c528ba68388d stduhpf stephduh@live.fr 2025-03-21T20:34:50+01:00 GitHub noreply@github.com 2025-03-21T20:34:50+01:00 Vulkan: RTE rounding for cpy to quant (#12480)
30c42ef5cbb2db756eff9aadc6b6c528ba68388d af04481e6b3b8dcdf5c1cc7d84bc7ece5658e9ab Eve 139727413+netrunnereve@users.noreply.github.com 2025-03-21T19:27:47Z GitHub noreply@github.com 2025-03-21T20:27:47+01:00 vulkan: workaround for AMD Windows driver 16 bit unpack8 bug (#12472)
af04481e6b3b8dcdf5c1cc7d84bc7ece5658e9ab 960e72607761eb2dd170b33f02a5a2840ec412fe Georgi Gerganov ggerganov@gmail.com 2025-03-21T16:14:29+02:00 GitHub noreply@github.com 2025-03-21T16:14:29+02:00 model : do not repack if a GPU device is present (#12498)
960e72607761eb2dd170b33f02a5a2840ec412fe ea1518e839abe668c20f7e0074c0721f803da898 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-21T10:21:36+01:00 GitHub noreply@github.com 2025-03-21T10:21:36+01:00 chore : cleanup llama_model_loader::TENSOR_ usage (#12492)
ea1518e839abe668c20f7e0074c0721f803da898 1aa87ee53d05505247c54612e40f6a38c680b433 marcoStocchi marcostocchi77@gmail.com 2025-03-21T10:12:45+01:00 GitHub noreply@github.com 2025-03-21T11:12:45+02:00 llama-tts : avoid crashes related to bad model file paths (#12482)
1aa87ee53d05505247c54612e40f6a38c680b433 9ffcc9e374080f73b16b7a351e6d76e7a8a19ce3 蕭澧邦 45505768+shou692199@users.noreply.github.com 2025-03-21T14:58:47+08:00 GitHub noreply@github.com 2025-03-21T14:58:47+08:00 [SYCL] Fix build on Windows when ccache enabled (#9954) (#9976)
9ffcc9e374080f73b16b7a351e6d76e7a8a19ce3 e04643063b3d240b8c0fdba98677dff6ba346784 Svetlozar Georgiev 55534064+sgeor255@users.noreply.github.com 2025-03-21T02:15:56Z GitHub noreply@github.com 2025-03-21T10:15:56+08:00 sycl: cleanup oneDNN related code (#12097)
e04643063b3d240b8c0fdba98677dff6ba346784 dbb3a4739e53226346c772dd72666e68b78dc583 Woof Dog 197125663+woof-dog@users.noreply.github.com 2025-03-20T14:57:43Z GitHub noreply@github.com 2025-03-20T15:57:43+01:00 webui : Prevent rerendering on textarea input (#12299)
dbb3a4739e53226346c772dd72666e68b78dc583 3d82dbcbce2c677fc35fbf99574ccd107d95a1f8 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-20T12:49:59+01:00 GitHub noreply@github.com 2025-03-20T12:49:59+01:00 llama : make Qwen2MoE QKV bias optional (#12477)
3d82dbcbce2c677fc35fbf99574ccd107d95a1f8 732b5fbf5e7f9cf069942f0c5850ee959ef321ba Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2025-03-20T17:05:34+05:30 GitHub noreply@github.com 2025-03-20T13:35:34+02:00 ggml : block interleaving support for Q4_K quantization for x86 AVX2 architecture (#12332)
732b5fbf5e7f9cf069942f0c5850ee959ef321ba 568013d0cd3d5add37c376b3d5e959809b711fc7 Bartowski ckealty1182@gmail.com 2025-03-20T02:36:37-04:00 GitHub noreply@github.com 2025-03-20T08:36:37+02:00 convert : avoid calls to tokenizer.added_tokens_decoder (#12473)
568013d0cd3d5add37c376b3d5e959809b711fc7 517b5ddbf002b91fd6d6daf5d8db8c88a0173039 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2025-03-19T21:01:57+01:00 GitHub noreply@github.com 2025-03-19T21:01:57+01:00 context : clear sets containing encoder output sequence ids before storing new values (#12470)
517b5ddbf002b91fd6d6daf5d8db8c88a0173039 a9b59288e222f39fc0311dc66944ed5a86c815fa Gaurav Garg 52341457+gaugarg-nv@users.noreply.github.com 2025-03-20T01:22:06+05:30 GitHub noreply@github.com 2025-03-19T20:52:06+01:00 CUDA: Improve flash decoding kernel GPU occupancy for BS=1 case (#12183)
a9b59288e222f39fc0311dc66944ed5a86c815fa 0fd8487b142b2b92565bc95b39ddc440955a237c Jeff Bolz jbolz@nvidia.com 2025-03-19T13:56:23-05:00 GitHub noreply@github.com 2025-03-19T19:56:23+01:00 vulkan: optimize iq1 coopmat2 dequant functions (#12427)
0fd8487b142b2b92565bc95b39ddc440955a237c 108e53c2f1b57b79c8bcf9773fba5a82e2f1eeb5 Guus Waals _@guusw.nl 2025-03-19T10:15:23Z GitHub noreply@github.com 2025-03-19T11:15:23+01:00 Fix visionOS build and add CI (#12415)
108e53c2f1b57b79c8bcf9773fba5a82e2f1eeb5 a686171ea71ed8cb8a324850d146cb65a001e141 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-19T09:08:49+01:00 GitHub noreply@github.com 2025-03-19T09:08:49+01:00 llama : add support for GPT2, Bloom and CodeShell tied word embeddings (#12456)
a686171ea71ed8cb8a324850d146cb65a001e141 c446b2edd2a9fe2772a1a18923c3e54a6749c364 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-19T08:58:13+01:00 GitHub noreply@github.com 2025-03-19T08:58:13+01:00 convert : Support chat_template.json (#12460)
c446b2edd2a9fe2772a1a18923c3e54a6749c364 d84635b1b085d54d6a21924e6171688d6e3dfb46 Jeff Bolz jbolz@nvidia.com 2025-03-19T02:26:26-05:00 GitHub noreply@github.com 2025-03-19T08:26:26+01:00 vulkan: Submit once enough matmul work has been recorded (#12406)
d84635b1b085d54d6a21924e6171688d6e3dfb46 75422e8bc42646005be0754f7aa438b97a5e777e lhez quic_lih@quicinc.com 2025-03-18T12:54:55-07:00 GitHub noreply@github.com 2025-03-18T12:54:55-07:00 opencl: improve profiling (#12442)
75422e8bc42646005be0754f7aa438b97a5e777e bb115d2bf7ed2cdd7dccd7ae74cc9cfe4b0adb71 Georgi Gerganov ggerganov@gmail.com 2025-03-18T21:35:19+02:00 GitHub noreply@github.com 2025-03-18T21:35:19+02:00 graph : normalize Q, K, V shapes + sync cross attention (#12449)
bb115d2bf7ed2cdd7dccd7ae74cc9cfe4b0adb71 29fff308c704c1c752cdb5153361e545e2bac09d R0CKSTAR xiaodong.ye@mthreads.com 2025-03-19T02:28:26+08:00 GitHub noreply@github.com 2025-03-18T19:28:26+01:00 musa: override warp_size of musa device to 32 (#12445)
29fff308c704c1c752cdb5153361e545e2bac09d c6af2161b200284d55633cf184a07406ca89908e Xuan-Son Nguyen son@huggingface.co 2025-03-18T19:16:19+01:00 GitHub noreply@github.com 2025-03-18T19:16:19+01:00 llama : support converting Mistral Small text-only (#12450)
c6af2161b200284d55633cf184a07406ca89908e 99aa304fb900654ec338749f64e62895b9a88afd Georgi Gerganov ggerganov@gmail.com 2025-03-18T19:35:11+02:00 GitHub noreply@github.com 2025-03-18T19:35:11+02:00 speculative : fix seg fault in certain cases (#12454)
99aa304fb900654ec338749f64e62895b9a88afd 8551c44d840a7db50adb958ccaf464dc3ded82e7 Xuan-Son Nguyen son@huggingface.co 2025-03-18T17:24:33+01:00 GitHub noreply@github.com 2025-03-18T17:24:33+01:00 llama : add support for EXAONE tied word embeddings (#12451)
8551c44d840a7db50adb958ccaf464dc3ded82e7 35cae5ba05a5292dc3108636a71ec59fa2f80ab7 Georgi Gerganov ggerganov@gmail.com 2025-03-18T13:05:49+02:00 GitHub noreply@github.com 2025-03-18T13:05:49+02:00 context : always use non-causal attention for encoder graphs (#12447)
35cae5ba05a5292dc3108636a71ec59fa2f80ab7 810e0af3f50379682dd46b7967c4aadf3f8286f6 Łukasz Ślusarczyk 112692748+lslusarczyk@users.noreply.github.com 2025-03-18T11:16:31+01:00 GitHub noreply@github.com 2025-03-18T11:16:31+01:00 SYCL: using graphs is configurable by environment variable and compile option (#12371)
810e0af3f50379682dd46b7967c4aadf3f8286f6 eba92d64c3f6d86de2e6b4dd3a540d2805a22b0c Georgi Gerganov ggerganov@gmail.com 2025-03-18T12:05:42+02:00 GitHub noreply@github.com 2025-03-18T12:05:42+02:00 server : fix warmup draft cache type (#12446)
eba92d64c3f6d86de2e6b4dd3a540d2805a22b0c d9a14523bb9074eef42d1b43ae4a1e149f81b7e2 Prajwal B Mehendarkar prajwal.b.mehendarkar@ibm.com 2025-03-18T15:07:33+05:30 GitHub noreply@github.com 2025-03-18T11:37:33+02:00 cmake : fix PowerPC build (#12241)
d9a14523bb9074eef42d1b43ae4a1e149f81b7e2 fd123cfead49eb32e386e26b8ef7a6d41554dda5 fj-y-saito 85871716+fj-y-saito@users.noreply.github.com 2025-03-18T17:14:39+09:00 GitHub noreply@github.com 2025-03-18T10:14:39+02:00 ggml : add SVE support for q6_K_q8_K (#12361)
fd123cfead49eb32e386e26b8ef7a6d41554dda5 a53f7f7b8859f3e634415ab03e1e295b9861d7e6 0cc4m picard12@live.de 2025-03-18T07:21:40+01:00 GitHub noreply@github.com 2025-03-18T07:21:40+01:00 Vulkan: Default to 1GB allocations instead of 4GB to avoid fragmentation and driver issues (#12434)
a53f7f7b8859f3e634415ab03e1e295b9861d7e6 7dfad387e3f6ac98d383ded2d175eb59736a3993 Łukasz Ślusarczyk 112692748+lslusarczyk@users.noreply.github.com 2025-03-18T01:51:25+01:00 GitHub noreply@github.com 2025-03-18T08:51:25+08:00 fixed compilation warnings in ggml-sycl (#12424)
7dfad387e3f6ac98d383ded2d175eb59736a3993 60c902926c928f9c2cd6390ce411876f92feeaf3 Molly Sophia mollysophia379@gmail.com 2025-03-18T07:27:50+08:00 GitHub noreply@github.com 2025-03-18T07:27:50+08:00 llama: Add support for RWKV v7 architecture (#12412)
60c902926c928f9c2cd6390ce411876f92feeaf3 b1b132efcba216c873715c483809730bb253f4a1 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-17T21:14:32+01:00 GitHub noreply@github.com 2025-03-17T21:14:32+01:00 docs : bring llama-cli conversation/template docs up-to-date (#12426)
b1b132efcba216c873715c483809730bb253f4a1 01e8f2138b2e40902afe2983ecbf503a08d74b1d Gaurav Garg 52341457+gaugarg-nv@users.noreply.github.com 2025-03-17T23:55:13+05:30 GitHub noreply@github.com 2025-03-17T20:25:13+02:00 cuda : enable CUDA Graph on CUDA Toolkit < 12.x (#12394)
01e8f2138b2e40902afe2983ecbf503a08d74b1d 484a8ab513bbd740cc49f30280c1acf52cb4e7e9 Guus Waals _@guusw.nl 2025-03-18T00:35:43+08:00 GitHub noreply@github.com 2025-03-17T13:35:43-03:00 ggml-vulkan: remove unused find_program(glslc) (#12416)
484a8ab513bbd740cc49f30280c1acf52cb4e7e9 cf2270e4d3685ac46f4a166d8718997ba7cbc45a Jeff Bolz jbolz@nvidia.com 2025-03-17T09:26:18-05:00 GitHub noreply@github.com 2025-03-17T09:26:18-05:00 vulkan: Add N/2 and N/4 optimized paths in coopmat2 shader (#12312)
cf2270e4d3685ac46f4a166d8718997ba7cbc45a f07690c930f74d82d4f108e567c7092544847f77 Daniele daniele.dilotorres@gmail.com 2025-03-17T12:42:33+01:00 GitHub noreply@github.com 2025-03-17T12:42:33+01:00 vulkan: subgroup size tuning (#12087)
f07690c930f74d82d4f108e567c7092544847f77 891c63956dbfbdf7ed2ecd0b5882cff49dbfe90f Jeff Bolz jbolz@nvidia.com 2025-03-17T04:43:35-05:00 GitHub noreply@github.com 2025-03-17T10:43:35+01:00 vulkan: use fp32 in coopmat2 q4_k dequant function (#12309)
891c63956dbfbdf7ed2ecd0b5882cff49dbfe90f 2f21123c1deb3ce1be3c0578c5f6980fe19ed077 Jeff Bolz jbolz@nvidia.com 2025-03-17T04:41:59-05:00 GitHub noreply@github.com 2025-03-17T10:41:59+01:00 vulkan: Pad N dimension of B matrix for coopmat2 perf, to avoid bounds checking (#12273)
2f21123c1deb3ce1be3c0578c5f6980fe19ed077 374101fd742bb35cb9bf46c86836e54d51191ffd Jeff Bolz jbolz@nvidia.com 2025-03-17T04:35:00-05:00 GitHub noreply@github.com 2025-03-17T10:35:00+01:00 vulkan: Adjust coopmat2 tile sizes and selection heuristic (#12258)
374101fd742bb35cb9bf46c86836e54d51191ffd b3c9a65673a63a6c9a75da24ee00d13499494e0c Christian Kastner ckk@kvr.at 2025-03-17T10:05:23+01:00 GitHub noreply@github.com 2025-03-17T11:05:23+02:00 cmake : enable building llama.cpp using system libggml (#12321)
b3c9a65673a63a6c9a75da24ee00d13499494e0c 8ba95dca2065c0073698afdfcda4c8a8f08bf0d9 Akarshan Biswas akarshan@menlo.ai 2025-03-17T07:15:12+05:30 GitHub noreply@github.com 2025-03-17T09:45:12+08:00 SYCL: set extras only on GGML_TYPE_Q4_0 (#12366)
8ba95dca2065c0073698afdfcda4c8a8f08bf0d9 dc079cfdffa1141a6caf5d41a33d73a1ef03da55 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-16T18:46:36+01:00 GitHub noreply@github.com 2025-03-16T19:46:36+02:00 llama : fix OLMo-2-0325-32B-Instruct K-norm size (#12400)
dc079cfdffa1141a6caf5d41a33d73a1ef03da55 7b61bcc87cfdeab88350e82df1c4b7be64331ea6 Georgi Gerganov ggerganov@gmail.com 2025-03-16T19:29:36+02:00 GitHub noreply@github.com 2025-03-16T19:29:36+02:00 context : fix init of n_outputs (#12397)
7b61bcc87cfdeab88350e82df1c4b7be64331ea6 f4c3dd5daa3a79f713813cf1aabdc5886071061d Daniel Bevenius daniel.bevenius@gmail.com 2025-03-16T18:22:05+01:00 GitHub noreply@github.com 2025-03-16T18:22:05+01:00 ci : add --symlinks to xcframework zip command (#12409)
f4c3dd5daa3a79f713813cf1aabdc5886071061d 3d35d87b4113648e224b837bb88e6b2c4c7f29e5 marcoStocchi marcostocchi77@gmail.com 2025-03-15T17:23:11+01:00 GitHub noreply@github.com 2025-03-15T17:23:11+01:00 llama-tts : add '-o' option (#12398)
3d35d87b4113648e224b837bb88e6b2c4c7f29e5 b19bd064c09822cb81efe4a38abafab3e979c9ce aubreyli aubreylee@gmail.com 2025-03-15T22:49:03+08:00 GitHub noreply@github.com 2025-03-15T15:49:03+01:00 SYCL: Delete redundant plus sign and space (#12391)
b19bd064c09822cb81efe4a38abafab3e979c9ce 92a391327e9201b9b5b32fdd3afb452026c22d4c fairydreaming 166155368+fairydreaming@users.noreply.github.com 2025-03-15T15:19:30+01:00 GitHub noreply@github.com 2025-03-15T22:19:30+08:00 SYCL : support non-contiguous tensors in binary ops (add, sub, etc) (#12399)
92a391327e9201b9b5b32fdd3afb452026c22d4c 9f2250ba722738ec0e6ab684636268a79160c854 Chenguang Li 757486878@qq.com 2025-03-15T09:31:08+08:00 GitHub noreply@github.com 2025-03-15T09:31:08+08:00 [CANN]MUL_MAT optimization (#12382)
9f2250ba722738ec0e6ab684636268a79160c854 774973b8f3d5e375b0b74d58638eeb1817e950a8 Eric Curtin ecurtin@redhat.com 2025-03-14T16:41:20Z GitHub noreply@github.com 2025-03-14T16:41:20Z Add CLI arg to llama-run to adjust the number of threads used (#12370)
774973b8f3d5e375b0b74d58638eeb1817e950a8 8fcb563613e20a04dd9791f0a9b8a41086428c09 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-14T16:57:05+01:00 GitHub noreply@github.com 2025-03-14T16:57:05+01:00 main : add -sysf / --system-prompt-file (#12249) (#12250)
8fcb563613e20a04dd9791f0a9b8a41086428c09 add2a3aa5a1571211aa5c7303b8e80c8d1824b91 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2025-03-14T13:47:05+01:00 GitHub noreply@github.com 2025-03-14T13:47:05+01:00 Load all MoE experts during warmup (#11571)
add2a3aa5a1571211aa5c7303b8e80c8d1824b91 c522ce4143a2b5c277f1e5f65cd570dbd0626466 Victor 194116445+dodekapod@users.noreply.github.com 2025-03-14T11:21:17+01:00 GitHub noreply@github.com 2025-03-14T11:21:17+01:00 server: fix "--grammar-file" parameter (#12285)
c522ce4143a2b5c277f1e5f65cd570dbd0626466 081bee8c643b1f6302e9edfe789ce2d5f0be6c77 Georgi Gerganov ggerganov@gmail.com 2025-03-14T10:47:44+02:00 GitHub noreply@github.com 2025-03-14T10:47:44+02:00 graph : simplify attn input build for unified KV cache (#12381)
081bee8c643b1f6302e9edfe789ce2d5f0be6c77 84d547554123a62e9ac77107cb20e4f6cc503af4 Georgi Gerganov ggerganov@gmail.com 2025-03-14T09:03:24+02:00 GitHub noreply@github.com 2025-03-14T09:03:24+02:00 hparams : add SWA rope parameters (#12374)
84d547554123a62e9ac77107cb20e4f6cc503af4 be7c3034108473beda214fd1d7c98fd6a7a3bdf5 Georgi Gerganov ggerganov@gmail.com 2025-03-13T19:08:07+02:00 GitHub noreply@github.com 2025-03-13T19:08:07+02:00 llama : fix Gemma3 SWA KV cache shift (#12373)
be7c3034108473beda214fd1d7c98fd6a7a3bdf5 e0dbec0bc6cd4b6230cda7a6ed1e9dac08d1600b Xuan-Son Nguyen son@huggingface.co 2025-03-13T12:34:54+01:00 GitHub noreply@github.com 2025-03-13T12:34:54+01:00 arg : no n_predict = -2 for examples except for main and infill (#12364)
e0dbec0bc6cd4b6230cda7a6ed1e9dac08d1600b 2048b5913d51beab82dfe29955f9008130b936c0 Georgi Gerganov ggerganov@gmail.com 2025-03-13T12:35:44+02:00 GitHub noreply@github.com 2025-03-13T12:35:44+02:00 llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)
2048b5913d51beab82dfe29955f9008130b936c0 f08f4b3187b691bb08a8884ed39ebaa94e956707 Ishaan Gandhi Ishaangandhi@gmail.com 2025-03-13T06:10:05-04:00 GitHub noreply@github.com 2025-03-13T11:10:05+01:00 server : fix crash when using verbose output with input tokens that are not in printable range (#12178) (#12338)
f08f4b3187b691bb08a8884ed39ebaa94e956707 80a02aa8588ef167d616f76f1781b104c245ace0 Oscar Barenys rtfss1@gmail.com 2025-03-12T20:06:58+01:00 GitHub noreply@github.com 2025-03-12T20:06:58+01:00 Update build.yml for Windows Vulkan builder to use Vulkan 1.4.304 SDK for VK_NV_cooperative_matrix2 support (#12301)
80a02aa8588ef167d616f76f1781b104c245ace0 363f8c5d67dcf80e00c39580dfa86dc2774d74c2 Daniel Bevenius daniel.bevenius@gmail.com 2025-03-12T13:45:32+01:00 GitHub noreply@github.com 2025-03-12T13:45:32+01:00 llama.swiftui : fix xcframework dir in README [no ci] (#12353)
363f8c5d67dcf80e00c39580dfa86dc2774d74c2 34c961b181836a4f06ab4c56d5ce61ce03fc478b Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2025-03-12T09:57:32Z GitHub noreply@github.com 2025-03-12T09:57:32Z sycl : variable sg_size support for mmvq kernels (#12336)
34c961b181836a4f06ab4c56d5ce61ce03fc478b 7841fc723e059d1fd9640e5c0ef19050fcc7c698 uvos philipp@uvos.xyz 2025-03-12T10:14:11+01:00 GitHub noreply@github.com 2025-03-12T10:14:11+01:00 CUDA/HIP: Fix fattn-vec-* when device warp size is not 32 (#12315)
7841fc723e059d1fd9640e5c0ef19050fcc7c698 bf69cfe62f9ccc01112c0232a55820b95a8c1fda Xuan-Son Nguyen son@huggingface.co 2025-03-12T09:30:24+01:00 GitHub noreply@github.com 2025-03-12T09:30:24+01:00 llama : Add Gemma 3 support (+ experimental vision capability) (#12343)
bf69cfe62f9ccc01112c0232a55820b95a8c1fda 10f2e81809bbb69ecfe64fc8b4686285f84b0c07 Jeff Bolz jbolz@nvidia.com 2025-03-12T00:59:19-05:00 GitHub noreply@github.com 2025-03-12T06:59:19+01:00 vulkan: fix bug in coopmat1 mul_mat_id (#12316)
10f2e81809bbb69ecfe64fc8b4686285f84b0c07 ba7654380a3c7c1b5ae154bea19134a3a9417a1e uvos philipp@uvos.xyz 2025-03-11T20:16:03+01:00 GitHub noreply@github.com 2025-03-11T20:16:03+01:00 CUDA/HIP: refractor mmqv to unify the calculation of nwarps and rows per block between host and device code. (#12177)
ba7654380a3c7c1b5ae154bea19134a3a9417a1e 6ab2e4765a673abcd162258a2671560a76106d69 jklincn 985765408@qq.com 2025-03-11T21:25:17+08:00 GitHub noreply@github.com 2025-03-11T14:25:17+01:00 ggml-backend : fix backend search path (#12330)
6ab2e4765a673abcd162258a2671560a76106d69 96e1280839561aaabb73851f94972a2cd37b2d96 BB-fat 45072480+BB-fat@users.noreply.github.com 2025-03-11T19:45:02+08:00 GitHub noreply@github.com 2025-03-11T13:45:02+02:00 metal : Cache the Metal library at the device context level (#12265)
96e1280839561aaabb73851f94972a2cd37b2d96 2c9f833d17bb5b8ea89dec663b072b5420fc5438 Xuan-Son Nguyen thichthat@gmail.com 2025-03-11T09:20:16+01:00 GitHub noreply@github.com 2025-03-11T09:20:16+01:00 clip : bring back GPU support (#12322)
2c9f833d17bb5b8ea89dec663b072b5420fc5438 251364549fe4a78d4e2e66f1adfaf2bd53041d2c Eve 139727413+netrunnereve@users.noreply.github.com 2025-03-10T19:28:11Z GitHub noreply@github.com 2025-03-10T19:28:11Z mat vec double buffer (#12188)
251364549fe4a78d4e2e66f1adfaf2bd53041d2c 8acdacb3ea00697477eb019efbb6fc183371055c R0CKSTAR xiaodong.ye@mthreads.com 2025-03-11T01:18:25+08:00 GitHub noreply@github.com 2025-03-10T18:18:25+01:00 musa: support new arch mp_31 and update doc (#12296)
8acdacb3ea00697477eb019efbb6fc183371055c 89b2b56e8658800375a8314200870b1ad4208a0b Henry Linjamäki henry.mikael.linjamaki@intel.com 2025-03-10T18:57:00+02:00 GitHub noreply@github.com 2025-03-10T09:57:00-07:00 opencl: use OpenCL C standard supported by the device (#12221)
89b2b56e8658800375a8314200870b1ad4208a0b e128a1bf5b65741b485dd094a4201264d0580d68 John Bean 113509988+johnbean393@users.noreply.github.com 2025-03-10T22:13:09+08:00 GitHub noreply@github.com 2025-03-10T16:13:09+02:00 readme: added Sidekick to available UIs (#12311)
e128a1bf5b65741b485dd094a4201264d0580d68 6ef79a67caf1159b0150b44ee80888c7ec98b83f Georgi Gerganov ggerganov@gmail.com 2025-03-10T14:07:15+02:00 GitHub noreply@github.com 2025-03-10T14:07:15+02:00 tests : fix test-quantize-fns to init the CPU backend (#12306)
6ef79a67caf1159b0150b44ee80888c7ec98b83f 4e39a3c332f84b890e91353ec448502cb8373a6f marcoStocchi marcostocchi77@gmail.com 2025-03-10T12:34:13+01:00 GitHub noreply@github.com 2025-03-10T13:34:13+02:00 common : refactor '-o' option (#12278)
4e39a3c332f84b890e91353ec448502cb8373a6f be421fc429795d135786f5a0e489709220a9c43a Olivier Chafik ochafik@users.noreply.github.com 2025-03-10T10:59:03Z GitHub noreply@github.com 2025-03-10T10:59:03Z `server`: extract <think> tags from qwq outputs (#12297)
be421fc429795d135786f5a0e489709220a9c43a 87c2630546cd8ccc836ae4c7d87d03fa597d2267 Olivier Chafik ochafik@users.noreply.github.com 2025-03-10T09:45:29Z GitHub noreply@github.com 2025-03-10T09:45:29Z `tool-call`: ensure there's always a non-empty tool call id (#12292)
87c2630546cd8ccc836ae4c7d87d03fa597d2267 2b3a25c212f8c4d8a49cec23e03343a7719d51c9 Olivier Chafik ochafik@users.noreply.github.com 2025-03-10T09:45:07Z GitHub noreply@github.com 2025-03-10T09:45:07Z allow missing content in message if tool_calls provided (#12293)
2b3a25c212f8c4d8a49cec23e03343a7719d51c9 8352cdc87b207a735d34c431a36c425728cb4586 Olivier Chafik ochafik@users.noreply.github.com 2025-03-10T09:44:42Z GitHub noreply@github.com 2025-03-10T09:44:42Z `sampler`: fixes trigger tokens + lazy grammars (fix typo cast from token to string) (#12291)
8352cdc87b207a735d34c431a36c425728cb4586 1e2f78a00450593e2dfa458796fcdd9987300dfc tc-mb 157115220+tc-mb@users.noreply.github.com 2025-03-10T16:33:24+08:00 GitHub noreply@github.com 2025-03-10T10:33:24+02:00 llava : fix bug in minicpm-v code (#11513)
1e2f78a00450593e2dfa458796fcdd9987300dfc 0fd7ca7a210bd4abc995cd728491043491dbdef7 Georgi Gerganov ggerganov@gmail.com 2025-03-09T19:08:20+02:00 GitHub noreply@github.com 2025-03-09T19:08:20+02:00 server : add speculative decoding presets for FIM (#12287)
0fd7ca7a210bd4abc995cd728491043491dbdef7 6fefc05a7a4e676780ae10b0a4d0728e5281f367 Georgi Gerganov ggerganov@gmail.com 2025-03-08T18:26:00+02:00 GitHub noreply@github.com 2025-03-08T18:26:00+02:00 authors : update (#12271)
6fefc05a7a4e676780ae10b0a4d0728e5281f367 7ab364390f92b0b8d83f69821a536b424838f3f8 Jason C.H ctrysbita@outlook.com 2025-03-09T00:02:39+08:00 GitHub noreply@github.com 2025-03-08T17:02:39+01:00 ggml-backend : make path_str compatible with C++20 (#12269)
7ab364390f92b0b8d83f69821a536b424838f3f8 7c7f3b7f435f41f2508e0e3010f0013cd8335156 Georgi Gerganov ggerganov@gmail.com 2025-03-07T20:54:30+02:00 GitHub noreply@github.com 2025-03-07T20:54:30+02:00 server : infill gen ends on new line (#12254)
7c7f3b7f435f41f2508e0e3010f0013cd8335156 102ac1891db32c346a7b6b96145a2a23c1e4c352 Daniel Bevenius daniel.bevenius@gmail.com 2025-03-07T14:15:27+01:00 GitHub noreply@github.com 2025-03-07T14:15:27+01:00 ggml : skip intermediate .air file when compiling .metallib (#12247)
102ac1891db32c346a7b6b96145a2a23c1e4c352 d6ae2fa06139e496880cbf65197c84341e9d98e7 Georgi Gerganov ggerganov@gmail.com 2025-03-07T14:00:27+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-07T14:49:44+02:00 sync : ggml
d6ae2fa06139e496880cbf65197c84341e9d98e7 68d0027f3d19eb579c1863814c91e37ffa699014 vmobilis 75476228+vmobilis@users.noreply.github.com 2025-03-07T11:11:40+03:00 Georgi Gerganov ggerganov@gmail.com 2025-03-07T14:49:44+02:00 ggml : ggml_compute_forward_concat() for arbitrary tensor type (ggml/1118)
68d0027f3d19eb579c1863814c91e37ffa699014 ea002810a209246d034d1b6ddac387f778751588 Rémy O remyoudompheng@gmail.com 2025-03-07T12:54:22+01:00 GitHub noreply@github.com 2025-03-07T13:54:22+02:00 ggml-cpu: faster AVX2 variant for IQ1_M (#12216)
ea002810a209246d034d1b6ddac387f778751588 8fad3c7a7c54a25a1ca38dfb08244df55288e675 Georgi Gerganov ggerganov@gmail.com 2025-03-07T12:19:31+02:00 GitHub noreply@github.com 2025-03-07T12:19:31+02:00 ci : fix save-load test invocations (#12245)
8fad3c7a7c54a25a1ca38dfb08244df55288e675 7cf64f6beecf54c6ac71503181f154667fd4228a Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-07T11:15:33+01:00 GitHub noreply@github.com 2025-03-07T11:15:33+01:00 server : Log original chat template parsing error (#12233)
7cf64f6beecf54c6ac71503181f154667fd4228a 5e2d57b2b2e43eadbe6d66ba3e873a824b95e725 Olivier Chafik ochafik@users.noreply.github.com 2025-03-07T09:33:37Z GitHub noreply@github.com 2025-03-07T09:33:37Z sync: minja - support QwQ-32B (#12235)
5e2d57b2b2e43eadbe6d66ba3e873a824b95e725 f1648e91cf6c52e9593810aa70857e412d474c09 BB-fat 45072480+BB-fat@users.noreply.github.com 2025-03-07T15:35:57+08:00 GitHub noreply@github.com 2025-03-07T08:35:57+01:00 metal : simplify kernel arguments using a struct (#3229) (#12194)
f1648e91cf6c52e9593810aa70857e412d474c09 d6c95b0740510231b3797b80d6d3440d8fe188b6 David Huang 1969802+hjc4869@users.noreply.github.com 2025-03-07T15:06:08+08:00 GitHub noreply@github.com 2025-03-07T08:06:08+01:00 HIP: fix rocWMMA build flags under Windows (#12230)
d6c95b0740510231b3797b80d6d3440d8fe188b6 d76a86d967ef491d530400b08bc8ef8a14807936 Daniel Bevenius daniel.bevenius@gmail.com 2025-03-07T06:23:16+01:00 GitHub noreply@github.com 2025-03-07T06:23:16+01:00 metal : fix default.metallib build (#12224)
d76a86d967ef491d530400b08bc8ef8a14807936 776f9e59cc8a85e840d1d4af8540d199c77190ac lhez quic_lih@quicinc.com 2025-03-06T16:20:35-08:00 GitHub noreply@github.com 2025-03-07T00:20:35Z opencl: Noncontiguous `norm`, `rms_norm`, disable `fp16` for some ops (#12217)
776f9e59cc8a85e840d1d4af8540d199c77190ac 3d652bfddfba09022525067e672c3c145c074649 xiaofei hbuxiaofei@gmail.com 2025-03-07T06:58:25+08:00 GitHub noreply@github.com 2025-03-06T22:58:25Z cmake : fix undefined reference errors for std::filesystem in ggml (#12092) (#12094)
3d652bfddfba09022525067e672c3c145c074649 5220a16d18563d3ffc509002f0514415fdda4036 Lucas Moura Belo lucas.belo@live.com 2025-03-06T16:15:13-03:00 GitHub noreply@github.com 2025-03-06T21:15:13+02:00 readme : update bindings (#12229)
5220a16d18563d3ffc509002f0514415fdda4036 3ffbbd5ce130859be91909e9b77d4c1962a6be2c Johannes Gäßler johannesg@5d6.de 2025-03-06T18:45:09+01:00 GitHub noreply@github.com 2025-03-06T18:45:09+01:00 CUDA: fix FA logic for PTX 7.0 and CC >= 7.5 (#12222)
3ffbbd5ce130859be91909e9b77d4c1962a6be2c 42994048a34b0bddd72b57c26f8ae2c7d417946d David Huang 1969802+hjc4869@users.noreply.github.com 2025-03-06T21:14:11+08:00 GitHub noreply@github.com 2025-03-06T14:14:11+01:00 HIP: rocWMMA documentation and enabling in workflow builds (#12179)
42994048a34b0bddd72b57c26f8ae2c7d417946d e9b2f84f145fbd458dcb98f227bd09370918be6e Olivier Chafik ochafik@users.noreply.github.com 2025-03-06T09:03:31Z GitHub noreply@github.com 2025-03-06T09:03:31Z update function-calling.md w/ template override for functionary-small-v3.2 (#12214)
e9b2f84f145fbd458dcb98f227bd09370918be6e e721c05c9336a72fbb59d5c75967360bc67036c6 Aaron Teo 57927438+taronaeo@users.noreply.github.com 2025-03-06T16:33:21+08:00 GitHub noreply@github.com 2025-03-06T09:33:21+01:00 llava: add big-endian conversion for image encoder (#12218)
e721c05c9336a72fbb59d5c75967360bc67036c6 57b6abf85acb1f697913a44e5e105e333d894b78 uvos philipp@uvos.xyz 2025-03-06T08:20:52+01:00 GitHub noreply@github.com 2025-03-06T08:20:52+01:00 HIP/CUDA: set the paramerter value in maintain_cuda_graph instead of replaceing it. (#12209)
57b6abf85acb1f697913a44e5e105e333d894b78 94bb63e4f0f6135579b88e5700ed40cefa374e09 Han Yin han.yin@arm.com 2025-03-05T22:22:49-08:00 GitHub noreply@github.com 2025-03-06T08:22:49+02:00 android : fix KV cache log message condition (#12212)
94bb63e4f0f6135579b88e5700ed40cefa374e09 f79243992cd31e4e4844d5158d2f3325b1d2d811 Henry Linjamäki henry.linjamaki@gmail.com 2025-03-06T03:33:40+02:00 GitHub noreply@github.com 2025-03-06T02:33:40+01:00 opencl : fix buffer alignment (#12197)
f79243992cd31e4e4844d5158d2f3325b1d2d811 ed4ce0dda24986c80d58e2ce112049af00f632f4 Henry Linjamäki henry.linjamaki@gmail.com 2025-03-06T03:31:14+02:00 GitHub noreply@github.com 2025-03-06T02:31:14+01:00 opencl : fix `ulong` kernel args were set from `int` variables (#12174)
ed4ce0dda24986c80d58e2ce112049af00f632f4 07d15723470a0a5b15d8ccad1aff5b20354ffbe1 simon886212 37953122+simon886212@users.noreply.github.com 2025-03-06T09:30:05+08:00 GitHub noreply@github.com 2025-03-06T02:30:05+01:00 opencl : fix profile-related errors (#12095)
07d15723470a0a5b15d8ccad1aff5b20354ffbe1 5e43f104cca1a14874e980326a506b44fde022b8 Rémy O remyoudompheng@gmail.com 2025-03-06T02:26:10+01:00 GitHub noreply@github.com 2025-03-06T02:26:10+01:00 ggml-cpu: Faster IQ1 mul_mat_vec on AVX2 using BMI2 instructions (#12154)
5e43f104cca1a14874e980326a506b44fde022b8 16e4b22c5e58ee200ede913fd7b7b8ba92132ce8 Akarshan Biswas akarshan@menlo.ai 2025-03-05T21:28:23+05:30 GitHub noreply@github.com 2025-03-05T16:58:23+01:00 SYCL: Disable f16 Unary OPs as not supported by the kernels (#12201)
16e4b22c5e58ee200ede913fd7b7b8ba92132ce8 074c4fd39df3af974e10fbee6cc6db5d9304655b Plamen Minev pacominev@gmail.com 2025-03-05T17:16:01+02:00 GitHub noreply@github.com 2025-03-05T17:16:01+02:00 ggml : fix GGMLMetalClass ODR (#12200)
074c4fd39df3af974e10fbee6cc6db5d9304655b 669912d9a5bf927312c553332ff997f0a99da8fb Daniel Bevenius daniel.bevenius@gmail.com 2025-03-05T14:16:40+01:00 GitHub noreply@github.com 2025-03-05T14:16:40+01:00 ci : add fetch-depth to xcframework upload (#12195)
669912d9a5bf927312c553332ff997f0a99da8fb fa31c438e0e709242ab7334d26fc3be3dcda07a0 Olivier Chafik ochafik@users.noreply.github.com 2025-03-05T13:05:13Z GitHub noreply@github.com 2025-03-05T13:05:13Z `tool-call`: fix Qwen 2.5 Coder support, add micro benchmarks, support trigger patterns for lazy grammars (#12034)
fa31c438e0e709242ab7334d26fc3be3dcda07a0 3ccbfe5a71c74ac574b00607067d0aa0a49df04c Daniel Bevenius daniel.bevenius@gmail.com 2025-03-05T10:22:29+01:00 GitHub noreply@github.com 2025-03-05T10:22:29+01:00 ci : fix xcframework artifact tag (#12191)
3ccbfe5a71c74ac574b00607067d0aa0a49df04c 06a92a193a07afe445929607be9d5e4d033956fb Daniel Bevenius daniel.bevenius@gmail.com 2025-03-05T08:34:02+01:00 GitHub noreply@github.com 2025-03-05T08:34:02+01:00 ci : remove xframework upload (#12190)
06a92a193a07afe445929607be9d5e4d033956fb a057897ad4e48e3df0354256e0cc80dadcb57595 Clauszy zhangyub@uniontech.com 2025-03-05T15:25:45+08:00 GitHub noreply@github.com 2025-03-05T09:25:45+02:00 server : fix cache reuse logic (#12161)
a057897ad4e48e3df0354256e0cc80dadcb57595 5bbe6a9fe9a8796a9389c85accec89dbc4d91e39 Daniel Bevenius daniel.bevenius@gmail.com 2025-03-05T06:30:31+01:00 GitHub noreply@github.com 2025-03-05T06:30:31+01:00 llama : add xcframework build script (#11996)
5bbe6a9fe9a8796a9389c85accec89dbc4d91e39 20a9b8f5e1380243ed03aeb50ae1bf94b8d68501 mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2025-03-04T17:53:26+01:00 GitHub noreply@github.com 2025-03-04T18:53:26+02:00 ggml : portability fixes for VS 2017 (#12150)
20a9b8f5e1380243ed03aeb50ae1bf94b8d68501 56d7a9f81274717fe035db192f315a049261c7fa Georgi Gerganov ggerganov@gmail.com 2025-03-04T18:42:44+02:00 GitHub noreply@github.com 2025-03-04T18:42:44+02:00 readme : fix roadmap link (#12185)
56d7a9f81274717fe035db192f315a049261c7fa 1a24c4621f0280306b0d53a4fa474fc65d3f1b2e Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-04T17:19:39+01:00 GitHub noreply@github.com 2025-03-04T12:19:39-04:00 main: allow preloading conversation with -p and add -st / --single-turn (#12145)
1a24c4621f0280306b0d53a4fa474fc65d3f1b2e becade5de77674696539163dfbaf5c041a1a8e97 Olivier Chafik ochafik@users.noreply.github.com 2025-03-04T06:24:07Z GitHub noreply@github.com 2025-03-04T08:24:07+02:00 `server`: fix deadly typo in response_format.json_schema.schema handling (#12168)
becade5de77674696539163dfbaf5c041a1a8e97 dfd6b2c0be191b3abe2fd9c1b25deff01c6249d8 David Huang 1969802+hjc4869@users.noreply.github.com 2025-03-04T05:10:54+08:00 GitHub noreply@github.com 2025-03-03T22:10:54+01:00 HIP: implement FlashAttention via rocWMMA for CDNA and RDNA3+ (#12032)
dfd6b2c0be191b3abe2fd9c1b25deff01c6249d8 b64d7cc2720a90c03480d9408fae40d936c110e2 Georgi Gerganov ggerganov@gmail.com 2025-03-03T17:57:38+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 sync : ggml
b64d7cc2720a90c03480d9408fae40d936c110e2 3d1cf3cf3394f2288dca3e1fb1fbb467d366e4bb cmdr2 secondary.cmdr2@gmail.com 2025-03-03T20:51:31+05:30 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 cuda: unary ops as float + de-duplicate (ggml/1130)
3d1cf3cf3394f2288dca3e1fb1fbb467d366e4bb 0cbee131ad332a8dab5bc72315f085a544682c26 Georgi Gerganov ggerganov@gmail.com 2025-02-28T12:37:35+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 sync : ggml
0cbee131ad332a8dab5bc72315f085a544682c26 8371d445958195f9ecdff39f67c0b357f1347e3a cmdr2 secondary.cmdr2@gmail.com 2025-02-28T12:36:46+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 cuda/vulkan: specify fp32-only support for some operations in supports_op (ggml/1129)
8371d445958195f9ecdff39f67c0b357f1347e3a 87abb7e903635a2660e89f9336122f374d683e0a Georgi Gerganov ggerganov@gmail.com 2025-02-28T09:09:58+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 sync : ggml
87abb7e903635a2660e89f9336122f374d683e0a 6d4c23b81b03c7b089a5f7a21ca73d1385d37191 cmdr2 secondary.cmdr2@gmail.com 2025-02-28T12:34:39+05:30 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 cuda/cpu: Increase support for fp16 unary operations (ggml/1125)
6d4c23b81b03c7b089a5f7a21ca73d1385d37191 6512a9003741bd3fe2e11bf3bb3608ec497d368d Diego Devesa slarengh@gmail.com 2025-02-27T13:35:07+01:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 whisper : support GGML_BACKEND_DL (whisper/2843)
6512a9003741bd3fe2e11bf3bb3608ec497d368d 4512055792cff7ea107f2d2231f79aa1af073c62 midnight midnightmagic@users.noreply.github.com 2025-02-05T04:41:10-08:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 cmake : fix compile assumptions for power9/etc (whisper/2777)
4512055792cff7ea107f2d2231f79aa1af073c62 f54a4ba11ed8ab59da778d5e280f7dc73c775a7a petterreinholdtsen pere-github@hungry.com 2025-02-26T21:44:00+01:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 Told cmake to install ggml-cpp.h as a public header file. (ggml/1126)
f54a4ba11ed8ab59da778d5e280f7dc73c775a7a aede2074f608d7bf6614cdd047bce687a2b3d908 cmdr2 shashank.shekhar.global@gmail.com 2025-02-25T18:06:34+05:30 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 Support pure float16 add/sub/mul/div operations in the CUDA (and CPU) backend (ggml/1121)
aede2074f608d7bf6614cdd047bce687a2b3d908 2679c3b55d1c9c3fed56dea00fea713622e42594 Georgi Gerganov ggerganov@gmail.com 2025-02-28T09:09:38+02:00 Georgi Gerganov ggerganov@gmail.com 2025-03-03T18:18:11+02:00 scripts : sync-ggml-am.sh fix
2679c3b55d1c9c3fed56dea00fea713622e42594 c43af9276b119dae7436b7878d59671d0f6b1a97 Daniel Bevenius daniel.bevenius@gmail.com 2025-03-03T16:17:36+01:00 GitHub noreply@github.com 2025-03-03T16:17:36+01:00 ci : set GITHUB_ACTION env var for server tests (#12162)
c43af9276b119dae7436b7878d59671d0f6b1a97 d5c63cd7f932663a20a860ef11a238e627abac53 dm4 dm4@secondstate.io 2025-03-03T21:09:29+08:00 GitHub noreply@github.com 2025-03-03T15:09:29+02:00 tts: add speaker file support (#12048)
d5c63cd7f932663a20a860ef11a238e627abac53 9660ffef582ca275092b621c87085a6eea136a90 Diego Devesa slarengh@gmail.com 2025-03-03T14:00:46+01:00 GitHub noreply@github.com 2025-03-03T14:00:46+01:00 test-backend-ops : add option -p to filter by op params (#12155)
9660ffef582ca275092b621c87085a6eea136a90 c950a1f69269bff416d74349a82d78181ce6f0f8 ag2s20150909 19373730+ag2s20150909@users.noreply.github.com 2025-03-03T20:54:08+08:00 GitHub noreply@github.com 2025-03-03T13:54:08+01:00 ggml : fix kleidiai build (#12159)
c950a1f69269bff416d74349a82d78181ce6f0f8 7b69003af74924ac04d97313c2e57c45ba56b616 Eric Curtin ecurtin@redhat.com 2025-03-03T12:44:56Z GitHub noreply@github.com 2025-03-03T12:44:56Z Adding UTF-8 support to llama.cpp (#12111)
7b69003af74924ac04d97313c2e57c45ba56b616 ece9745bb8079b9f4af45df29b67ad0c6e50584d Xuan-Son Nguyen thichthat@gmail.com 2025-03-03T11:42:45+01:00 GitHub noreply@github.com 2025-03-03T11:42:45+01:00 webui : add ?m=... and ?q=... params (#12148)
ece9745bb8079b9f4af45df29b67ad0c6e50584d cc473cac7cea1484c1f870231073b0bf0352c6f9 Akarshan Biswas akarshan@menlo.ai 2025-03-03T15:37:22+05:30 GitHub noreply@github.com 2025-03-03T11:07:22+01:00 SYCL: Move CPY kernels to a separate file and add few missing kernels (#12133)
cc473cac7cea1484c1f870231073b0bf0352c6f9 14dec0c2f29ae56917907dbf2eed6b19438d0a0e Diego Devesa slarengh@gmail.com 2025-03-02T22:11:00+01:00 GitHub noreply@github.com 2025-03-02T22:11:00+01:00 ggml-backend : keep paths in native string type when possible (#12144)
14dec0c2f29ae56917907dbf2eed6b19438d0a0e 1782cdfed60952f9ff333fc2ab5245f2be702453 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-02T14:53:48+01:00 GitHub noreply@github.com 2025-03-02T14:53:48+01:00 main: use jinja chat template system prompt by default (#12118)
1782cdfed60952f9ff333fc2ab5245f2be702453 45a8e7674548fca9b4ff6d8de97b73bf60f83d72 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-01T15:22:27+01:00 GitHub noreply@github.com 2025-03-01T15:22:27+01:00 main: update outdated system prompt message (followup to #12131) (#12132)
45a8e7674548fca9b4ff6d8de97b73bf60f83d72 80c41ddd8f11b8094018296d9820c2b6c119ac12 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-03-01T13:56:45+01:00 GitHub noreply@github.com 2025-03-01T13:56:45+01:00 common : add --system-prompt parameter, replace behavior of -p in conversation mode (#12131)
80c41ddd8f11b8094018296d9820c2b6c119ac12 2cc4a5e44a3a9ab94426816770d611b33fae8f77 Erik Scholz Green-Sky@users.noreply.github.com 2025-03-01T12:57:22+01:00 GitHub noreply@github.com 2025-03-01T12:57:22+01:00 CUDA: compress mode option and default to size (#12029)
2cc4a5e44a3a9ab94426816770d611b33fae8f77 06c2b1561d8b882bc018554591f8c35eb04ad30e Vivian vynride@gmail.com 2025-03-01T15:45:09+05:30 GitHub noreply@github.com 2025-03-01T11:15:09+01:00 webui : minor typo fixes (#12116)
06c2b1561d8b882bc018554591f8c35eb04ad30e 70680c48e5f77d2d3138712a6582bd8c1e548922 Xuan-Son Nguyen thichthat@gmail.com 2025-02-28T17:44:46+01:00 GitHub noreply@github.com 2025-02-28T17:44:46+01:00 convert : fix Norway problem when parsing YAML (#12114)
70680c48e5f77d2d3138712a6582bd8c1e548922 c43a3e7996e585e2addde1e44057a4f3cdbadef8 William Tambellini wtambellini@sdl.com 2025-02-28T05:41:47-08:00 GitHub noreply@github.com 2025-02-28T14:41:47+01:00 ggml : upgrade init_tensor API to return a ggml_status (#11854)
c43a3e7996e585e2addde1e44057a4f3cdbadef8 84d5f4bc195b9540fcb902d869015fba7ef6baa4 Xuan-Son Nguyen thichthat@gmail.com 2025-02-28T12:44:11+01:00 GitHub noreply@github.com 2025-02-28T12:44:11+01:00 llama : add Phi-4-mini support (supersede #12099) (#12108)
84d5f4bc195b9540fcb902d869015fba7ef6baa4 438a83926afcff3643ffef5543db67545ceffe39 Alex Brooks alex.brooks@ibm.com 2025-02-28T04:31:47-07:00 GitHub noreply@github.com 2025-02-28T11:31:47Z Update granite vision docs for 3.2 model (#12105)
438a83926afcff3643ffef5543db67545ceffe39 9c42b1718ca8299f9afeabdc122badeab64c9690 Rémy O remyoudompheng@gmail.com 2025-02-28T09:42:52+01:00 GitHub noreply@github.com 2025-02-28T09:42:52+01:00 vulkan: add specific MMV kernels for IQ2 and IQ3 quants + optimizations (#11595)
9c42b1718ca8299f9afeabdc122badeab64c9690 05e6f5aad0a4bb48fb2775f2a78505540fdbc47d Johannes Gäßler johannesg@5d6.de 2025-02-28T09:26:43+01:00 GitHub noreply@github.com 2025-02-28T09:26:43+01:00 CUDA: fix logic for V100 + GGML_CUDA_FORCE_MMQ (#12098)
05e6f5aad0a4bb48fb2775f2a78505540fdbc47d 673cfef9aa8daad09966208909f346eb996628a4 Prashant Vithule 119530321+Vithulep@users.noreply.github.com 2025-02-28T13:06:12+05:30 GitHub noreply@github.com 2025-02-28T09:36:12+02:00 ggml: aarch64: implement SVE kernels for q2_k_q8_k vector dot (#12064)
673cfef9aa8daad09966208909f346eb996628a4 fbeda9002d4b8b79a4f9288a7ff0d34ef4fb23d5 hipudding huafengchun@gmail.com 2025-02-28T15:23:47+08:00 GitHub noreply@github.com 2025-02-28T15:23:47+08:00 CANN: Fix build error with GCC 13 (#11990)
fbeda9002d4b8b79a4f9288a7ff0d34ef4fb23d5 581650b7cacec2872982fde381bd3bcda0f78699 Eve 139727413+netrunnereve@users.noreply.github.com 2025-02-28T07:20:08Z GitHub noreply@github.com 2025-02-28T08:20:08+01:00 vulkan: matmul dequantization improvements (#12015)
581650b7cacec2872982fde381bd3bcda0f78699 b95c8af37ccf169b0a3216b7ed691af0534e5091 Daniele 57776841+daniandtheweb@users.noreply.github.com 2025-02-28T06:52:51Z GitHub noreply@github.com 2025-02-28T07:52:51+01:00 vulkan: improve im2col (#11826)
b95c8af37ccf169b0a3216b7ed691af0534e5091 a800ae46da2ed7dac236aa6bf2b595da6b6294b5 Vladimir Vuksanovic 109677816+vvuksanovic@users.noreply.github.com 2025-02-27T08:42:48+01:00 GitHub noreply@github.com 2025-02-27T09:42:48+02:00 cmake: Fix ggml backend dependencies and installation (#11818)
a800ae46da2ed7dac236aa6bf2b595da6b6294b5 69050a11be0ae3e01329f11371ecb6850bdaded5 Ting Lou ting.lou@gmail.com 2025-02-26T22:26:52+08:00 GitHub noreply@github.com 2025-02-26T15:26:52+01:00 llava : add struct for FFI bindgen (#12079)
69050a11be0ae3e01329f11371ecb6850bdaded5 3567ee3a94d57ba72b6d023ca507d11e75767edb Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2025-02-26T14:04:48+01:00 GitHub noreply@github.com 2025-02-26T08:04:48-05:00 Refactor gguf scripts to improve metadata handling (#11909)
3567ee3a94d57ba72b6d023ca507d11e75767edb 53e4db10126e277486eccb94c6728f4146c75741 Aleksei Nikiforov 103434461+AlekseiNikiforovIBM@users.noreply.github.com 2025-02-26T12:39:27+01:00 GitHub noreply@github.com 2025-02-26T11:39:27Z gguf-py: enable reading non-native endian files (#12081)
53e4db10126e277486eccb94c6728f4146c75741 d7cfe1ffe0f435d0048a6058d529daf76e072d9c Kante Yin kerthcet@gmail.com 2025-02-26T15:49:36+08:00 GitHub noreply@github.com 2025-02-26T09:49:36+02:00 readme : update infra list (#9096)
d7cfe1ffe0f435d0048a6058d529daf76e072d9c a82c9e7c23ef6db48cebfa194dc9cebbc4ac3552 Olivier Chafik ochafik@users.noreply.github.com 2025-02-25T18:52:56Z GitHub noreply@github.com 2025-02-25T18:52:56Z docs: add docs/function-calling.md to lighten server/README.md's plight (#12069)
a82c9e7c23ef6db48cebfa194dc9cebbc4ac3552 401af80b546005a06854827b732e3b46979ae028 Jeff Bolz jbolz@nvidia.com 2025-02-25T09:30:21-06:00 GitHub noreply@github.com 2025-02-25T16:30:21+01:00 vulkan: fix assertion when qy_needs_dequant (#12068)
401af80b546005a06854827b732e3b46979ae028 c132239bfbc1aae3a96dfee3350afcb491f64ade rhjdvsgsgks 26178113+rhjdvsgsgks@users.noreply.github.com 2025-02-25T11:52:52Z GitHub noreply@github.com 2025-02-25T12:52:52+01:00 server: handle echo=false on /v1/completions (#12060)
c132239bfbc1aae3a96dfee3350afcb491f64ade 393fca629e6ec391b76edf778cb3f7a8a3bc56f9 Judd foldl@users.noreply.github.com 2025-02-25T19:32:20+08:00 GitHub noreply@github.com 2025-02-25T12:32:20+01:00 add OP sigmoid (#12056)
393fca629e6ec391b76edf778cb3f7a8a3bc56f9 61d4f39dfeaf3bbcf4e1535ac03953a5d766680b Molly Sophia mollysophia379@gmail.com 2025-02-25T19:28:22+08:00 GitHub noreply@github.com 2025-02-25T19:28:22+08:00 ggml-cpu: Fix build with sve (#12059)
61d4f39dfeaf3bbcf4e1535ac03953a5d766680b 0b52745649062c04b546aab662cfdb220f4f0621 Rémy O remyoudompheng@gmail.com 2025-02-25T12:04:45+01:00 GitHub noreply@github.com 2025-02-25T12:04:45+01:00 vulkan: implement more backpropagation operators (#11914)
0b52745649062c04b546aab662cfdb220f4f0621 4d1051a40ffc2fdff80bc532eea5b9568b85c156 Olivier Chafik ochafik@users.noreply.github.com 2025-02-25T10:40:22Z GitHub noreply@github.com 2025-02-25T10:40:22Z server: support add_generation_prompt query param (#12062)
4d1051a40ffc2fdff80bc532eea5b9568b85c156 3e9a2860e996657fc10db8393cf65adc40703082 Alex Brooks alex.brooks@ibm.com 2025-02-25T02:46:05-07:00 GitHub noreply@github.com 2025-02-25T10:46:05+01:00 Add Doc for Converting Granite Vision -> GGUF (#12006)
3e9a2860e996657fc10db8393cf65adc40703082 58d07a8043a1395177cf77b3e4f388e34182ae64 Vitali Lovich vlovich+github@gmail.com 2025-02-25T01:29:33-08:00 GitHub noreply@github.com 2025-02-25T11:29:33+02:00 llama : expose llama_model_n_head_kv in the API (#11997)
58d07a8043a1395177cf77b3e4f388e34182ae64 34a846b5847a18d133b360074f1fb485b2632b2d Gian-Carlo Pascutto gcp@sjeng.org 2025-02-25T10:27:58+01:00 GitHub noreply@github.com 2025-02-25T11:27:58+02:00 metal : copy kernels for quant to F32/F16 conversions (#12017)
34a846b5847a18d133b360074f1fb485b2632b2d 7a2c913e66353362d7f28d612fd3c9d51a831eda lhez quic_lih@quicinc.com 2025-02-24T13:47:07-08:00 GitHub noreply@github.com 2025-02-24T14:47:07-07:00 opencl: fix for small models (#11950)
7a2c913e66353362d7f28d612fd3c9d51a831eda 08d5986290cc42d2c52739e046642b8252f97e4b Alex Brooks alex.brooks@ibm.com 2025-02-24T09:09:51-07:00 GitHub noreply@github.com 2025-02-24T17:09:51+01:00 llava : Add Granite Vision Support (#11794)
08d5986290cc42d2c52739e046642b8252f97e4b 651adf4b6675339465179b81b194d98cc14704d6 Neo Zhang Jianyu jianyu.zhang@intel.com 2025-02-24T22:33:23+08:00 GitHub noreply@github.com 2025-02-24T22:33:23+08:00 [SYCL] Optimize mul_mat for Q4_0 on Intel GPU (#12035)
651adf4b6675339465179b81b194d98cc14704d6 8303e8b0fb2c1e26a8edd58071f9120a5bd6930a Aleksei Nikiforov 103434461+AlekseiNikiforovIBM@users.noreply.github.com 2025-02-24T12:27:01+01:00 GitHub noreply@github.com 2025-02-24T11:27:01Z gguf_convert_endian.py: implement byteswapping for q4_k and q6_k (#11349)
8303e8b0fb2c1e26a8edd58071f9120a5bd6930a 7ad0779f5de84a68143b2c00ab5dc94a948925d3 Akarshan Biswas akarshan.biswas@gmail.com 2025-02-24T15:48:25+05:30 GitHub noreply@github.com 2025-02-24T10:18:25Z SYCL: Fix GGML_SYCL_DEBUG macro (#11995)
7ad0779f5de84a68143b2c00ab5dc94a948925d3 f777a73e18c218848bca0748581c043987348a5d Florent BENOIT fbenoit@redhat.com 2025-02-23T18:15:51+01:00 GitHub noreply@github.com 2025-02-23T17:15:51Z run: allow to customize prompt by env var LLAMA_PROMPT_PREFIX (#12041)
f777a73e18c218848bca0748581c043987348a5d af7747c95ae71a5db4184947f837179e82c70b77 Eric Curtin ecurtin@redhat.com 2025-02-23T13:14:32Z GitHub noreply@github.com 2025-02-23T13:14:32Z Some llama-run cleanups (#11973)
af7747c95ae71a5db4184947f837179e82c70b77 a28e0d5eb18c18e6a4598286158f427269b1444e Aaron Teo 57927438+taronaeo@users.noreply.github.com 2025-02-23T05:39:24+08:00 GitHub noreply@github.com 2025-02-22T21:39:24Z ggml-cpu: Support s390x SIMD Instruction Set (#12019)
a28e0d5eb18c18e6a4598286158f427269b1444e 36c258ee921dbb5c96bdc57c0872e4a9a129bef6 Johannes Gäßler johannesg@5d6.de 2025-02-22T20:44:34+01:00 GitHub noreply@github.com 2025-02-22T20:44:34+01:00 CUDA: app option to compile without FlashAttention (#12025)
36c258ee921dbb5c96bdc57c0872e4a9a129bef6 f3e64859edb0d55d4223ead78672597cd1a218df Ting Lou ting.lou@gmail.com 2025-02-22T22:28:28+08:00 GitHub noreply@github.com 2025-02-22T15:28:28+01:00 llava: build clip image from pixels (#11999)
f3e64859edb0d55d4223ead78672597cd1a218df 5fa07c2f93c73161bf09ef0b23b5d2686f9a073e Georgi Gerganov ggerganov@gmail.com 2025-02-22T15:03:00+02:00 GitHub noreply@github.com 2025-02-22T15:03:00+02:00 ci : fix arm upload artifacts (#12024)
5fa07c2f93c73161bf09ef0b23b5d2686f9a073e 335eb04a91f481f37c0c9b302ee31b449b04c3e9 Johannes Gäßler johannesg@5d6.de 2025-02-22T12:20:17+01:00 GitHub noreply@github.com 2025-02-22T12:20:17+01:00 CUDA: optimize FA for GQA + large batches (#12014)
335eb04a91f481f37c0c9b302ee31b449b04c3e9 cf756d6e0a314e0fe25ff944341d79ea8c94cc96 Rohanjames1997 rohan.james4@gmail.com 2025-02-22T04:48:57-06:00 GitHub noreply@github.com 2025-02-22T11:48:57+01:00 ci : Build on Github-hosted arm64 runners (#12009)
cf756d6e0a314e0fe25ff944341d79ea8c94cc96 d70908421ff22b013e8209f2d12e5c750663c620 Georgi Gerganov ggerganov@gmail.com 2025-02-22T12:46:31+02:00 GitHub noreply@github.com 2025-02-22T11:46:31+01:00 server : disable Nagle's algorithm (#12020)
d70908421ff22b013e8209f2d12e5c750663c620 de8b5a3624499bdb9fa6e99840259998638f093f Gian-Carlo Pascutto gcp@sjeng.org 2025-02-22T09:43:24+01:00 GitHub noreply@github.com 2025-02-22T09:43:24+01:00 cuda: Add Q5_1, Q5_0, Q4_1 and Q4_0 to F32 conversion support. (#12000)
de8b5a3624499bdb9fa6e99840259998638f093f 51f311e057723b7454d0ebe20f545a1a2c4db6b2 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-22T06:33:29+01:00 GitHub noreply@github.com 2025-02-22T06:33:29+01:00 llama.swiftui : add "Done" dismiss button to help view (#11998)
51f311e057723b7454d0ebe20f545a1a2c4db6b2 586d5fe6ebb8f92e9dd53420e04cec2697046073 Georgi Gerganov ggerganov@gmail.com 2025-02-21T18:33:18+02:00 GitHub noreply@github.com 2025-02-21T18:33:18+02:00 llama : skip loading unused tensors (#12004)
586d5fe6ebb8f92e9dd53420e04cec2697046073 ecc8e3aeff269037aa786dfb393e8b531f96832f Johannes Gäßler johannesg@5d6.de 2025-02-21T12:51:25+01:00 GitHub noreply@github.com 2025-02-21T12:51:25+01:00 doc: update contributing guidelines [no ci] (#11969)
ecc8e3aeff269037aa786dfb393e8b531f96832f 0b3863ff9576872855b0265da2cab2ce7e25c4d9 PureJourney edward.pong@qq.com 2025-02-21T19:21:05+08:00 GitHub noreply@github.com 2025-02-21T12:21:05+01:00 CUDA: correct the lowest Maxwell supported by CUDA 12 (#11984)
0b3863ff9576872855b0265da2cab2ce7e25c4d9 ee02ad02c56ff36a5edd22d8617ab3f9546ce7fe Bodhi 3882561+BodhiHu@users.noreply.github.com 2025-02-21T15:46:23+08:00 GitHub noreply@github.com 2025-02-21T09:46:23+02:00 MUSA: support ARM64 and enable dp4a .etc (#11843)
ee02ad02c56ff36a5edd22d8617ab3f9546ce7fe c392e5094deaf2d1a7c18683214f007fad3fe42b Alex Brooks alex.brooks@ibm.com 2025-02-20T23:11:03-07:00 GitHub noreply@github.com 2025-02-21T08:11:03+02:00 clip : fix visual encoders with no CLS (#11982)
c392e5094deaf2d1a7c18683214f007fad3fe42b c5d91a74006c49376590ef2b67420bc7ce206397 momonga 146910567+mmngays@users.noreply.github.com 2025-02-21T03:43:22+09:00 GitHub noreply@github.com 2025-02-20T19:43:22+01:00 server (webui): Fix Premature Submission During IME Conversion (#11971)
c5d91a74006c49376590ef2b67420bc7ce206397 4806498bf15ef767de3776b00856e56c373dd1b1 Charles Xu charles.xu@arm.com 2025-02-20T14:06:51+01:00 GitHub noreply@github.com 2025-02-20T15:06:51+02:00 ggml-cpu: Add CPU backend support for KleidiAI library (#11390)
4806498bf15ef767de3776b00856e56c373dd1b1 0d559580a0a74c842c3a876035ba96a338aabfb2 Prashant Vithule 119530321+Vithulep@users.noreply.github.com 2025-02-20T15:38:32+05:30 GitHub noreply@github.com 2025-02-20T12:08:32+02:00 ggml: aarch64: implement SVE kernels for q3_K_q8_K vector dot (#11917)
0d559580a0a74c842c3a876035ba96a338aabfb2 d04e7163c85a847bc61d58c22f2c503596db7aa8 Michael Engel mengel@redhat.com 2025-02-20T09:35:11+01:00 GitHub noreply@github.com 2025-02-20T10:35:11+02:00 run : add --chat-template-file (#11961)
d04e7163c85a847bc61d58c22f2c503596db7aa8 d07c621393fab6cf32f3add2aa65e4d5331d4a67 Johannes Gäßler johannesg@5d6.de 2025-02-19T20:45:17+01:00 GitHub noreply@github.com 2025-02-19T20:45:17+01:00 doc: add links to ggml examples [no ci] (#11958)
d07c621393fab6cf32f3add2aa65e4d5331d4a67 abd4d0bc4f1a9a0e429bc8ee0d5ece2a394a0a39 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-19T12:29:52+01:00 GitHub noreply@github.com 2025-02-19T12:29:52+01:00 common : add llama.vim preset for Qwen2.5 Coder (#11945)
abd4d0bc4f1a9a0e429bc8ee0d5ece2a394a0a39 9626d9351a6dfb665400d9fccbda876a0a96ef67 Georgi Gerganov ggerganov@gmail.com 2025-02-19T13:29:42+02:00 GitHub noreply@github.com 2025-02-19T13:29:42+02:00 speculative : update default params (#11954)
9626d9351a6dfb665400d9fccbda876a0a96ef67 b58934c1836b5ea51dbacbe899eee125775e77c9 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-19T06:16:23+01:00 GitHub noreply@github.com 2025-02-19T06:16:23+01:00 llama : fix indentation in llama-grammar [no ci] (#11943)
b58934c1836b5ea51dbacbe899eee125775e77c9 63e489c025d61c7ca5ec06c5d10f36e2b76aaa1d igardev 49397134+igardev@users.noreply.github.com 2025-02-19T00:01:44+02:00 GitHub noreply@github.com 2025-02-18T23:01:44+01:00 server : (webui) Enable communication with parent html (if webui is in iframe) (#11940)
63e489c025d61c7ca5ec06c5d10f36e2b76aaa1d 63ac12856303108ee46635e6c9e751f81415ee64 Olivier Chafik ochafik@users.noreply.github.com 2025-02-18T18:03:23Z GitHub noreply@github.com 2025-02-18T18:03:23Z tool-call: refactor common chat / tool-call api (+ tests / fixes) (#11900)
63ac12856303108ee46635e6c9e751f81415ee64 5137da7b8c3eaa090476a632888ca178ba109f8a Xuan-Son Nguyen thichthat@gmail.com 2025-02-18T14:21:41+01:00 GitHub noreply@github.com 2025-02-18T14:21:41+01:00 server : add TEI API format for /rerank endpoint (#11942)
5137da7b8c3eaa090476a632888ca178ba109f8a 09aaf4f1f5b69b5173b7fcd24eab96729f6b242a MoonRide303 130458190+MoonRide303@users.noreply.github.com 2025-02-18T10:30:16+01:00 GitHub noreply@github.com 2025-02-18T10:30:16+01:00 scripts: corrected encoding when getting chat template (#11866) (#11907)
09aaf4f1f5b69b5173b7fcd24eab96729f6b242a 73e2ed3ce3492d3ed70193dd09ae8aa44779651d xiaobing318 71554036+xiaobing318@users.noreply.github.com 2025-02-18T17:12:49+08:00 GitHub noreply@github.com 2025-02-18T10:12:49+01:00 docs : Fix duplicated file extension in test command (#11935)
73e2ed3ce3492d3ed70193dd09ae8aa44779651d f7b1116af102bcac450c1a522e9c59db241c6767 Johannes Gäßler johannesg@5d6.de 2025-02-17T14:03:24+01:00 GitHub noreply@github.com 2025-02-17T14:03:24+01:00 CUDA: use async data loading for FlashAttention (#11894)
f7b1116af102bcac450c1a522e9c59db241c6767 c4d29baf3236b011730b6ccaae6852e781fb1b91 Eve 139727413+netrunnereve@users.noreply.github.com 2025-02-17T11:20:23Z GitHub noreply@github.com 2025-02-17T12:20:23+01:00 update release requirements (#11897)
c4d29baf3236b011730b6ccaae6852e781fb1b91 2eea03d86a2d132c8245468c26290ce07a27a8e8 Antoine Viallon antoine@lesviallon.fr 2025-02-17T11:25:12+01:00 GitHub noreply@github.com 2025-02-17T11:25:12+01:00 server : fix divide-by-zero in metrics reporting (#11915)
2eea03d86a2d132c8245468c26290ce07a27a8e8 0f2bbe656473177538956d22b6842bcaa0449fab Rémy O remyoudompheng@gmail.com 2025-02-17T07:55:57+01:00 GitHub noreply@github.com 2025-02-17T07:55:57+01:00 vulkan: implement several ops relevant for ggml_opt (#11769)
0f2bbe656473177538956d22b6842bcaa0449fab fe163d5bf3fb8e99c96ff6d21775fa3d1daee3cf Xuan-Son Nguyen thichthat@gmail.com 2025-02-16T18:11:22+01:00 GitHub noreply@github.com 2025-02-16T17:11:22Z server : bump httplib to 0.19.0 (#11908)
fe163d5bf3fb8e99c96ff6d21775fa3d1daee3cf 818a340ea8be55b3706e1772527cb8738e90a8c7 standby24x7 standby24x7@gmail.com 2025-02-16T18:51:13+09:00 GitHub noreply@github.com 2025-02-16T10:51:13+01:00 common : Fix a typo in help (#11899)
818a340ea8be55b3706e1772527cb8738e90a8c7 bf42a23d0a515a508aecf10fde1d52c5ed5104c6 Xuan-Son Nguyen thichthat@gmail.com 2025-02-16T10:36:39+01:00 GitHub noreply@github.com 2025-02-16T10:36:39+01:00 ci : fix (again) arm64 build fails (#11895)
bf42a23d0a515a508aecf10fde1d52c5ed5104c6 c2ea16f260445d1f2bd9fee78fb8acb2b3aaaad4 Jeff Bolz jbolz@nvidia.com 2025-02-16T01:52:23-06:00 GitHub noreply@github.com 2025-02-16T08:52:23+01:00 vulkan: support multi/vision rope, and noncontiguous rope (#11902)
c2ea16f260445d1f2bd9fee78fb8acb2b3aaaad4 6dde1782483d6b0a1d59f5a5fbcb3119b9d34c27 Hale Chan halechan@qq.com 2025-02-16T14:50:26+08:00 GitHub noreply@github.com 2025-02-16T08:50:26+02:00 metal : fix the crash caused by the lack of residency set support on Intel Macs. (#11904)
6dde1782483d6b0a1d59f5a5fbcb3119b9d34c27 fc10c38ded84670955d4c44770f8acfb64617e15 Johannes Gäßler johannesg@5d6.de 2025-02-15T20:23:22+01:00 GitHub noreply@github.com 2025-02-15T20:23:22+01:00 scripts: fix compare-llama-bench commit hash logic (#11891)
fc10c38ded84670955d4c44770f8acfb64617e15 22885105a6b034abaf1c1471ad90fb2ae96146bb 708-145 40387547+708-145@users.noreply.github.com 2025-02-15T20:03:30+01:00 GitHub noreply@github.com 2025-02-15T21:03:30+02:00 examples: fix typo in imatrix/README.md (#11884)
22885105a6b034abaf1c1471ad90fb2ae96146bb c2cd24fbfdfeacc2fc6ad03878379de104264114 Adrian Kretz me@akretz.com 2025-02-15T19:39:20+01:00 GitHub noreply@github.com 2025-02-15T20:39:20+02:00 metal : optimize dequant q6_K kernel (#11892)
c2cd24fbfdfeacc2fc6ad03878379de104264114 68ff663a04ed92044a9937bcae353e9d9733f9cd Georgi Gerganov ggerganov@gmail.com 2025-02-15T20:29:56+02:00 GitHub noreply@github.com 2025-02-15T20:29:56+02:00 readme : add notice about new package registry (#11890)
68ff663a04ed92044a9937bcae353e9d9733f9cd f3552296924e704c11ca936907b9cad7873db8e2 Georgi Gerganov ggerganov@gmail.com 2025-02-15T16:40:57+02:00 GitHub noreply@github.com 2025-02-15T16:40:57+02:00 repo : update links to new url (#11886)
f3552296924e704c11ca936907b9cad7873db8e2 fc1b0d0936e4dfc52a81f38e7420c7d23f6caa88 Olivier Chafik ochafik@users.noreply.github.com 2025-02-15T10:11:36Z GitHub noreply@github.com 2025-02-15T10:11:36Z server: fix type promotion typo causing crashes w/ --jinja w/o tools (#11880)
fc1b0d0936e4dfc52a81f38e7420c7d23f6caa88 89daa2564f6eab33be53c6a1b39273af536d6bb3 Rémy O remyoudompheng@gmail.com 2025-02-15T09:01:40+01:00 GitHub noreply@github.com 2025-02-15T09:01:40+01:00 vulkan: initial support for IQ1_S and IQ1_M quantizations (#11528)
89daa2564f6eab33be53c6a1b39273af536d6bb3 300907b2110cc17b4337334dc397e05de2d8f5e0 Michał Moskal michal@moskal.me 2025-02-14T12:46:08-08:00 GitHub noreply@github.com 2025-02-14T12:46:08-08:00 llguidance build fixes for Windows (#11664)
300907b2110cc17b4337334dc397e05de2d8f5e0 94b87f87b502d2ab6c8b28d2b5935cf008ca1505 lhez quic_lih@quicinc.com 2025-02-14T11:12:23-08:00 GitHub noreply@github.com 2025-02-14T12:12:23-07:00 opencl: Fix rope and softmax (#11833)
94b87f87b502d2ab6c8b28d2b5935cf008ca1505 dbc2ec59b5603fbd25f3833b2407b4f3612b9f02 Diego Devesa slarengh@gmail.com 2025-02-14T15:33:52+01:00 GitHub noreply@github.com 2025-02-14T15:33:52+01:00 cuda : add ampere to the list of default architectures (#11870)
dbc2ec59b5603fbd25f3833b2407b4f3612b9f02 3d68f034dad53f0f27ad626b2732ef48fbcea4ee Georgi Gerganov ggerganov@gmail.com 2025-02-14T14:48:40+02:00 GitHub noreply@github.com 2025-02-14T14:48:40+02:00 docker : drop to CUDA 12.4 (#11869)
3d68f034dad53f0f27ad626b2732ef48fbcea4ee 38e32eb6a0cec32130b699ce9fefad6c74571953 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-14T11:16:56+01:00 GitHub noreply@github.com 2025-02-14T11:16:56+01:00 llama : add completion for --chat-template-file (#11860)
38e32eb6a0cec32130b699ce9fefad6c74571953 a4f011e8d02179f032627130f961eb77ee30401c Jinyang He hejinyang@loongson.cn 2025-02-14T16:54:27+08:00 GitHub noreply@github.com 2025-02-14T10:54:27+02:00 ggml: optimize some vec dot functions for LoongArch ASX (#11842)
a4f011e8d02179f032627130f961eb77ee30401c a7b8ce226071b2b0faaad0d36cc5ebd7fb074730 Eve 139727413+netrunnereve@users.noreply.github.com 2025-02-14T02:59:40Z GitHub noreply@github.com 2025-02-14T02:59:40Z vulkan: linux builds + small subgroup size fixes (#11767)
a7b8ce226071b2b0faaad0d36cc5ebd7fb074730 04045bb84288dc7e9e424d4e2bf7f40d259b4c6a theraininsky 76763719+theraininsky@users.noreply.github.com 2025-02-14T09:13:43+08:00 GitHub noreply@github.com 2025-02-14T02:13:43+01:00 llama-bench : fix unexpected global variable initialize sequence issue (#11832)
04045bb84288dc7e9e424d4e2bf7f40d259b4c6a 8a8c4ceb6050bd9392609114ca56ae6d26f5b8f5 Georgi Gerganov ggerganov@gmail.com 2025-02-14T00:16:56+02:00 GitHub noreply@github.com 2025-02-14T00:16:56+02:00 readme : minor
8a8c4ceb6050bd9392609114ca56ae6d26f5b8f5 c1f958c0381e797135b7d42a677542133264193c Jeffrey Morgan jmorganca@gmail.com 2025-02-13T09:05:04-08:00 GitHub noreply@github.com 2025-02-13T18:05:04+01:00 llamafile: use member variable instead of constant for iq4nlt (#11780)
c1f958c0381e797135b7d42a677542133264193c c48f630d1c1942fce08aa7cb18a53ace443cd611 Reza Rahemtola 49811529+RezaRahemtola@users.noreply.github.com 2025-02-13T17:22:44+01:00 GitHub noreply@github.com 2025-02-13T17:22:44+01:00 server : (docs) Update wrong tool calling example (#11809)
c48f630d1c1942fce08aa7cb18a53ace443cd611 bd6e55bfd3f3af3e5705cf87a10e5178cef7c3c1 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-13T14:46:59+01:00 GitHub noreply@github.com 2025-02-13T14:46:59+01:00 llama : add --completion-bash option (#11846)
bd6e55bfd3f3af3e5705cf87a10e5178cef7c3c1 c7f460ab882065ec5696e3d51e24dbf67b539287 R0CKSTAR xiaodong.ye@mthreads.com 2025-02-13T20:28:18+08:00 GitHub noreply@github.com 2025-02-13T13:28:18+01:00 musa: bump MUSA SDK version to rc3.1.1 (#11822)
c7f460ab882065ec5696e3d51e24dbf67b539287 27e8a23300e30cd6ff6107ce262acf832ca60597 Olivier Chafik ochafik@users.noreply.github.com 2025-02-13T10:05:16Z GitHub noreply@github.com 2025-02-13T10:05:16Z `server`: fix tool-call of DeepSeek R1 Qwen, return reasoning_content (Command 7RB & DeepSeek R1) unless `--reasoning-format none` (#11607)
27e8a23300e30cd6ff6107ce262acf832ca60597 e4376270d971cff7992bdb6c5412a739195b1459 Vinesh Janarthanan 36610342+VJHack@users.noreply.github.com 2025-02-13T00:45:57-06:00 GitHub noreply@github.com 2025-02-13T08:45:57+02:00 sampling: add Top-nσ sampler (#11223)
e4376270d971cff7992bdb6c5412a739195b1459 3e693197724c31d53a9b69018c2f1bd0b93ebab2 Oleksandr Kuvshynov 661042+okuvshynov@users.noreply.github.com 2025-02-13T01:25:34-05:00 GitHub noreply@github.com 2025-02-13T08:25:34+02:00 llama.cpp: fix warning message (#11839)
3e693197724c31d53a9b69018c2f1bd0b93ebab2 a394039db004c6ee00098250d160b5aa018c2314 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-13T07:07:51+01:00 GitHub noreply@github.com 2025-02-13T08:07:51+02:00 llama : update llama_decode_internal ref [no ci] (#11840)
a394039db004c6ee00098250d160b5aa018c2314 be3bbd62153820ff6d358c817360927f429105c4 Diego Devesa slarengh@gmail.com 2025-02-13T01:02:38+01:00 GitHub noreply@github.com 2025-02-13T01:02:38+01:00 ggml-cpu : add chunking support to mul_mat_id (#11666)
be3bbd62153820ff6d358c817360927f429105c4 31afcbee0eebbc998ab311aa314e389d60aa2127 Xuan-Son Nguyen thichthat@gmail.com 2025-02-13T00:33:45+01:00 GitHub noreply@github.com 2025-02-13T00:33:45+01:00 ggml : x2 speed for WASM by optimizing SIMD (#11453)
31afcbee0eebbc998ab311aa314e389d60aa2127 5c4284d57bbc613121e90de5271f1cbc95d55872 Woof Dog 197125663+woof-dog@users.noreply.github.com 2025-02-12T22:47:11Z GitHub noreply@github.com 2025-02-12T23:47:11+01:00 server : (webui) Give copy button back to all message bubbles (#11814)
5c4284d57bbc613121e90de5271f1cbc95d55872 bfd11a2344ce6005bfbd5432a06fc7325bc0ecae uvos philipp@uvos.xyz 2025-02-12T22:25:28+01:00 GitHub noreply@github.com 2025-02-12T22:25:28+01:00 HIP: Remove GCN from list of devices that avoid MMQ (#11831)
bfd11a2344ce6005bfbd5432a06fc7325bc0ecae 0fb77f821f6e70ad8b8247a97d1022f0fef78991 JC 43374599+MrSMlT@users.noreply.github.com 2025-02-12T20:36:11Z GitHub noreply@github.com 2025-02-12T21:36:11+01:00 Fix: Compile failure due to Microsoft STL breaking change (#11836)
0fb77f821f6e70ad8b8247a97d1022f0fef78991 e598697d63d062b4af939f1b0383d6adc6292d1a Georgi Gerganov ggerganov@gmail.com 2025-02-12T21:46:02+02:00 Georgi Gerganov ggerganov@gmail.com 2025-02-12T21:46:02+02:00 sync : ggml
e598697d63d062b4af939f1b0383d6adc6292d1a fef0cbeadf5c8e221f832158cb4006ade39ef786 uvos philipp@uvos.xyz 2025-02-12T17:25:03+01:00 GitHub noreply@github.com 2025-02-12T17:25:03+01:00 HIP: Switch to std::vector in rocblas version check (#11820)
fef0cbeadf5c8e221f832158cb4006ade39ef786 748ee9fe9312acb8755ee4bf46fd9de2e6a45f29 bandoti 141645996+bandoti@users.noreply.github.com 2025-02-12T10:06:53-04:00 GitHub noreply@github.com 2025-02-12T10:06:53-04:00 cleanup: fix compile warnings associated with gnu_printf (#11811)
748ee9fe9312acb8755ee4bf46fd9de2e6a45f29 198b1ec611a2c551ea40e5b9c0b862f37555a4cc Richard r-burton@hotmail.co.uk 2025-02-12T13:57:33Z GitHub noreply@github.com 2025-02-12T15:57:33+02:00 ggml : fix multi-threaded clamp_f32 (#11824)
198b1ec611a2c551ea40e5b9c0b862f37555a4cc c3d6af7cd2d79dc89da086b2d08c777d0319d829 Weizhao Ouyang o451686892@gmail.com 2025-02-12T20:22:58+08:00 GitHub noreply@github.com 2025-02-12T13:22:58+01:00 ggml-cpu: Fix duplicate MATMUL_INT8 (#11817)
c3d6af7cd2d79dc89da086b2d08c777d0319d829 369be5598ac5f71f6aa6d6606e9aec769a23dafa Johannes Gäßler johannesg@5d6.de 2025-02-12T13:16:39+01:00 GitHub noreply@github.com 2025-02-12T13:16:39+01:00 CUDA: fix CUDART_VERSION checks (#11821)
369be5598ac5f71f6aa6d6606e9aec769a23dafa 4078c77f9891831f29ffc7c315c8ec6695ba5ce7 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-12T08:40:01+01:00 GitHub noreply@github.com 2025-02-12T09:40:01+02:00 llama : fix typo in llama-grammar.h [no ci] (#11816)
4078c77f9891831f29ffc7c315c8ec6695ba5ce7 90e4dba461b07e635fd1daf3b491c978c7dd0013 lhez quic_lih@quicinc.com 2025-02-11T14:04:13-08:00 GitHub noreply@github.com 2025-02-11T15:04:13-07:00 docs: add OpenCL (#11697)
90e4dba461b07e635fd1daf3b491c978c7dd0013 a18f481f99962638092d6f1c98b1d34d3e3256de Sheldon Robinson sheldon.robinson@live.com 2025-02-11T10:55:45-05:00 GitHub noreply@github.com 2025-02-11T16:55:45+01:00 Fix #11802: Compile bug - RegQueryValueExA changed to RegQueryValueEx (#11803)
a18f481f99962638092d6f1c98b1d34d3e3256de b9ab0a4d0b2ed19effec130921d05fb5c30b68c5 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-11T14:06:45+01:00 GitHub noreply@github.com 2025-02-11T14:06:45+01:00 server : use common_token_to_piece instead of common_detokenize (#11740)
b9ab0a4d0b2ed19effec130921d05fb5c30b68c5 7b891bdc863663b4dc1155aa9429b58c721524b2 Johannes Gäßler johannesg@5d6.de 2025-02-11T00:17:22+01:00 GitHub noreply@github.com 2025-02-11T00:17:22+01:00 CUDA: use arch list for compatibility check (#11775)
7b891bdc863663b4dc1155aa9429b58c721524b2 81732619fd2d570712dc78db5965cee9224b38bd Maxim Evtush 154841002+maximevtush@users.noreply.github.com 2025-02-10T23:21:31+01:00 GitHub noreply@github.com 2025-02-10T23:21:31+01:00 fix: typos in documentation files (#11791)
81732619fd2d570712dc78db5965cee9224b38bd 507f9174fe856772b6c7c17e81be442de7ee6d1e jason_w jason.wang@126.com 2025-02-11T06:17:48+08:00 GitHub noreply@github.com 2025-02-10T23:17:48+01:00 docs: utilize the forward slash (/) as the path separator for Unix-like systems (#11770)
507f9174fe856772b6c7c17e81be442de7ee6d1e 19b392d58dc08c366d0b29bd3b9c6991fa4e1662 Xuan-Son Nguyen thichthat@gmail.com 2025-02-10T21:23:17+01:00 GitHub noreply@github.com 2025-02-10T21:23:17+01:00 server : (webui) introduce conversation branching + idb storage (#11792)
19b392d58dc08c366d0b29bd3b9c6991fa4e1662 0893e0114e934bdd0eba0ff69d9ef8c59343cbc3 Wilken Gottwalt 12194808+wgottwalt@users.noreply.github.com 2025-02-10T19:58:18+01:00 GitHub noreply@github.com 2025-02-10T20:58:18+02:00 llama-mmap: fix missing include (#11796)
0893e0114e934bdd0eba0ff69d9ef8c59343cbc3 d7b31a9d84297b493a61c9a8ee3a458e7ccc64a7 Xuan-Son Nguyen thichthat@gmail.com 2025-02-10T18:03:28+01:00 GitHub noreply@github.com 2025-02-10T18:03:28+01:00 server : correct signal handler (#11795)
d7b31a9d84297b493a61c9a8ee3a458e7ccc64a7 9ac3457b39c78c5eb103280e08fb6163642484ff Olivier Chafik ochafik@users.noreply.github.com 2025-02-10T09:34:09Z GitHub noreply@github.com 2025-02-10T09:34:09Z sync: minja (https://github.com/google/minja/commit/a72057e5190de2c612d4598bb10b4bfd0f53011f) (#11774)
9ac3457b39c78c5eb103280e08fb6163642484ff c2a67efe38e6782c0217e1de3edc68de6951612b pascal-lc 49066376+pascal-lc@users.noreply.github.com 2025-02-10T16:05:57+08:00 GitHub noreply@github.com 2025-02-10T09:05:57+01:00 Update README.md [no ci] (#11781)
c2a67efe38e6782c0217e1de3edc68de6951612b b044a0fe3ca0cbef9dd041edce3ebda8c501fae4 Danny Milosavljevic dannym@friendly-machines.com 2025-02-10T07:17:21+01:00 GitHub noreply@github.com 2025-02-10T07:17:21+01:00 vulkan: Make Vulkan optional at runtime (#11493). (#11494)
b044a0fe3ca0cbef9dd041edce3ebda8c501fae4 19d3c8293b1f61acbe2dab1d49a17950fd788a4a Wagner Bruna wbruna@users.noreply.github.com 2025-02-10T03:08:22-03:00 GitHub noreply@github.com 2025-02-10T07:08:22+01:00 vulkan: add environment variable GGML_VK_PREFER_HOST_MEMORY to avoid VRAM allocation (#11592)
19d3c8293b1f61acbe2dab1d49a17950fd788a4a 98f6b0fd1ea88ce33f4fcd1775d9a463067156ac Eric Curtin ecurtin@redhat.com 2025-02-09T10:34:49Z GitHub noreply@github.com 2025-02-09T10:34:49Z There's a better way of clearing lines (#11756)
98f6b0fd1ea88ce33f4fcd1775d9a463067156ac 55ac8c7791ff44aeb82bd7fe3ca2041844fc77f1 Jeff Bolz jbolz@nvidia.com 2025-02-09T01:43:51-06:00 GitHub noreply@github.com 2025-02-09T08:43:51+01:00 vulkan: account for lookup tables when checking shared memory size (#11502)
55ac8c7791ff44aeb82bd7fe3ca2041844fc77f1 e6e658319952f7ad269dc11275b9edddc721fc6d Xuan-Son Nguyen thichthat@gmail.com 2025-02-08T21:54:50+01:00 GitHub noreply@github.com 2025-02-08T21:54:50+01:00 server : (webui) revamp Settings dialog, add Pyodide interpreter (#11759)
e6e658319952f7ad269dc11275b9edddc721fc6d aaa55053076f3d5d7da4d9a877cb77e112dabed4 Woof Dog 197125663+woof-dog@users.noreply.github.com 2025-02-08T19:09:55Z GitHub noreply@github.com 2025-02-08T20:09:55+01:00 server : (webui) increase edit textarea size (#11763)
aaa55053076f3d5d7da4d9a877cb77e112dabed4 bdcf8b6a56f4d49d3420ae5b21cdf9a116040551 Georgi Gerganov ggerganov@gmail.com 2025-02-08T18:08:43+02:00 GitHub noreply@github.com 2025-02-08T18:08:43+02:00 server : minor log updates (#11760)
bdcf8b6a56f4d49d3420ae5b21cdf9a116040551 4d3465c5aeca8be29cac77f1535c35f4fb274eca Georgi Gerganov ggerganov@gmail.com 2025-02-08T16:49:38+02:00 Georgi Gerganov ggerganov@gmail.com 2025-02-08T16:49:38+02:00 cont : fix mmap flag print (#11699)
4d3465c5aeca8be29cac77f1535c35f4fb274eca d80be897acdca45f8f7ea2e52c930b1d0e738a14 Karol Kontny 82021046+kkontny@users.noreply.github.com 2025-02-08T15:30:53+01:00 GitHub noreply@github.com 2025-02-08T15:30:53+01:00 ggml: Fix data race in ggml threadpool (#11736)
d80be897acdca45f8f7ea2e52c930b1d0e738a14 3ab410f55f26038de45d6cc569aaa5cb29acb918 Johannes Gäßler johannesg@5d6.de 2025-02-08T10:46:07+01:00 GitHub noreply@github.com 2025-02-08T10:46:07+01:00 CUDA: fix min. version for movmatrix (#11751)
3ab410f55f26038de45d6cc569aaa5cb29acb918 0cf867160ca9d492e06c0bc688b337cffd1eb187 Nikolaos Pothitos pothitos@di.uoa.gr 2025-02-08T11:43:04+02:00 GitHub noreply@github.com 2025-02-08T10:43:04+01:00 readme : update front-end framework (#11753)
0cf867160ca9d492e06c0bc688b337cffd1eb187 d2fe216fb2fb7ca8627618c9ea3a2e7886325780 Xuan-Son Nguyen thichthat@gmail.com 2025-02-08T10:42:34+01:00 GitHub noreply@github.com 2025-02-08T10:42:34+01:00 server : (webui) fix numeric settings being saved as string (#11739)
d2fe216fb2fb7ca8627618c9ea3a2e7886325780 ed926d8833df3c29797edbc98dafd9b575aa0729 Eric Curtin ecurtin@redhat.com 2025-02-07T14:42:46Z GitHub noreply@github.com 2025-02-07T14:42:46Z Make logging more verbose (#11714)
ed926d8833df3c29797edbc98dafd9b575aa0729 2d219b389e8c8c40bce547b08c8aa7add60fde1f Georgi Gerganov ggerganov@gmail.com 2025-02-07T16:05:34+02:00 GitHub noreply@github.com 2025-02-07T16:05:34+02:00 llama : fix defrag logic (#11707)
2d219b389e8c8c40bce547b08c8aa7add60fde1f 333820d7491cd31c707a340ff23b984a84e40154 Christian Fillion cfillion@users.noreply.github.com 2025-02-07T08:55:47-05:00 GitHub noreply@github.com 2025-02-07T15:55:47+02:00 vocab : ignore invalid UTF-8 input in the BPE tokenizer (#11729)
333820d7491cd31c707a340ff23b984a84e40154 c026ba3c23765a648ca27c7a15ecf179f8e27f26 magicse magicse@users.noreply.github.com 2025-02-07T15:48:47+02:00 GitHub noreply@github.com 2025-02-07T15:48:47+02:00 llama : fix progress dots (#11730)
c026ba3c23765a648ca27c7a15ecf179f8e27f26 7ee953a64a40c09438b2064539becdbc577cefd0 Jeff Bolz jbolz@nvidia.com 2025-02-07T04:26:03-06:00 GitHub noreply@github.com 2025-02-07T11:26:03+01:00 vulkan: print shared memory size (#11719)
7ee953a64a40c09438b2064539becdbc577cefd0 ec3bc8270bc67b58955748d40a3e558a05b2d8f2 Christian Fillion cfillion@users.noreply.github.com 2025-02-07T04:33:27-05:00 GitHub noreply@github.com 2025-02-07T11:33:27+02:00 llama : add llama_sampler_init for safe usage of llama_sampler_free (#11727)
ec3bc8270bc67b58955748d40a3e558a05b2d8f2 b7552cfcbc7defccd8bdefd0a7b9c47d145ed3d7 Akarshan Biswas akarshan.biswas@gmail.com 2025-02-07T14:57:53+05:30 GitHub noreply@github.com 2025-02-07T09:27:53Z SYCL: remove XMX info from print devices (#11712)
b7552cfcbc7defccd8bdefd0a7b9c47d145ed3d7 225bbbfa39930cda38a2e5d1f3e5b38226732009 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-07T09:15:22+01:00 GitHub noreply@github.com 2025-02-07T09:15:22+01:00 common : add default embeddings presets (#11677)
225bbbfa39930cda38a2e5d1f3e5b38226732009 855cd0734aca26c86cc23d94aefd34f934464ac9 Jinyang He hejinyang@loongson.cn 2025-02-07T15:38:31+08:00 GitHub noreply@github.com 2025-02-07T09:38:31+02:00 ggml : optimize and build warning fix for LoongArch (#11709)
855cd0734aca26c86cc23d94aefd34f934464ac9 8a59053f63fffc24e730cd3ea067760abfe4a919 tv1wnd 55383215+tv1wnd@users.noreply.github.com 2025-02-06T22:48:51+01:00 GitHub noreply@github.com 2025-02-06T22:48:51+01:00 llama : fix old glm4 models (#11670)
8a59053f63fffc24e730cd3ea067760abfe4a919 1d20e53c40c3cc848ba2b95f5bf7c075eeec8b19 Georgi Gerganov ggerganov@gmail.com 2025-02-06T21:23:03+02:00 Georgi Gerganov ggerganov@gmail.com 2025-02-06T21:23:03+02:00 sync : ggml
1d20e53c40c3cc848ba2b95f5bf7c075eeec8b19 2fb3c32a1634488a5265d1304ab37628eeb5480d Patrick Peng retr0@retr0.blog 2025-02-06T09:29:13-05:00 Georgi Gerganov ggerganov@gmail.com 2025-02-06T21:22:54+02:00 rpc: fix known RCE in rpc-server (ggml/1103)
2fb3c32a1634488a5265d1304ab37628eeb5480d 9ab42dc722ad19a12af80f38a06474d498f96da3 Xuan-Son Nguyen thichthat@gmail.com 2025-02-06T17:32:29+01:00 GitHub noreply@github.com 2025-02-06T17:32:29+01:00 server : (webui) migrate project to ReactJS with typescript (#11688)
9ab42dc722ad19a12af80f38a06474d498f96da3 194b2e69f8da3a22395c74fd9acd6d5835437b96 Tei Home taiteitonghome@proton.me 2025-02-06T20:16:15+08:00 GitHub noreply@github.com 2025-02-06T12:16:15Z docs: update fedora cuda guide for 12.8 release (#11393)
194b2e69f8da3a22395c74fd9acd6d5835437b96 9dd7a0390feffcc1f4b17eb7692a6e43030d85af Akarshan Biswas akarshan.biswas@gmail.com 2025-02-06T17:12:35+05:30 GitHub noreply@github.com 2025-02-06T11:42:35Z SYCL: Adjust support condition for norm operators (#11674)
9dd7a0390feffcc1f4b17eb7692a6e43030d85af c0d4843225eed38903ea71ef302a02fa0b27f048 Georgi Gerganov ggerganov@gmail.com 2025-02-06T13:41:37+02:00 GitHub noreply@github.com 2025-02-06T13:41:37+02:00 llama : add log about loading model tensors (#11699)
c0d4843225eed38903ea71ef302a02fa0b27f048 8d4d2be143a3919c3d194b9bf7a221e50abed893 Adrien Gallouët adrien@gallouet.fr 2025-02-06T12:08:13+01:00 GitHub noreply@github.com 2025-02-06T13:08:13+02:00 build : fix llama.pc (#11658)
8d4d2be143a3919c3d194b9bf7a221e50abed893 2c6c8df56d8a3edd657b9a295e95d469a37f0044 junchao-zhao 68935141+junchao-loongson@users.noreply.github.com 2025-02-06T17:20:00+08:00 GitHub noreply@github.com 2025-02-06T11:20:00+02:00 ggml : fix LoongArch compile error with 128-bit SIMD (#11701)
2c6c8df56d8a3edd657b9a295e95d469a37f0044 8a7e3bf17aa5a8412854787746c92a28623a8925 Jeff Bolz jbolz@nvidia.com 2025-02-06T00:15:30-06:00 GitHub noreply@github.com 2025-02-06T07:15:30+01:00 vulkan: optimize coopmat2 iq2/iq3 callbacks (#11521)
8a7e3bf17aa5a8412854787746c92a28623a8925 1b598b30581bad59e5af86c94362f9a30f261fac Rémy O remyoudompheng@gmail.com 2025-02-06T07:09:59+01:00 GitHub noreply@github.com 2025-02-06T07:09:59+01:00 vulkan: initial support for IQ4_XS quantization (#11501)
1b598b30581bad59e5af86c94362f9a30f261fac 902368a06b915b860236cfc97ff885b2aceae256 Jeff Bolz jbolz@nvidia.com 2025-02-06T00:02:18-06:00 GitHub noreply@github.com 2025-02-06T07:02:18+01:00 vulkan: use smaller combined allocations to avoid fragmentation (#11551)
902368a06b915b860236cfc97ff885b2aceae256 c3db0480bb820e120249014b380e1f4badf2a1c1 Charles Duffy charles@dyfis.net 2025-02-05T19:52:31-06:00 GitHub noreply@github.com 2025-02-06T09:52:31+08:00 metal : avoid breaking build when metal API predates TARGET_OS_VISION (#11690)
c3db0480bb820e120249014b380e1f4badf2a1c1 d774ab3acc4fee41fbed6dbfc192b57d5f79f34b Matvey Soloviev blackhole89@gmail.com 2025-02-06T01:55:25+01:00 GitHub noreply@github.com 2025-02-06T01:55:25+01:00 readme : add link to Autopen under UIs (#11684)
d774ab3acc4fee41fbed6dbfc192b57d5f79f34b fa62da9b2dc03539a30f1306f59c4c6ffbe4f50a Georgi Gerganov ggerganov@gmail.com 2025-02-05T10:57:42+02:00 GitHub noreply@github.com 2025-02-05T10:57:42+02:00 metal : adjust support conditions for norm operators (#11671)
fa62da9b2dc03539a30f1306f59c4c6ffbe4f50a 1ec208083cb896dd8772a2f962151fa3d4a74e36 Johannes Gäßler johannesg@5d6.de 2025-02-05T08:58:31+01:00 GitHub noreply@github.com 2025-02-05T08:58:31+01:00 CUDA: support for mat. mul. with ne03 != ne13 (#11656)
1ec208083cb896dd8772a2f962151fa3d4a74e36 9f4cc8f8d310b13ab3fc93a9be81b1d1376a7fa6 SAMI samuel.koesnadi@stud.uni-due.de 2025-02-05T14:45:40+07:00 GitHub noreply@github.com 2025-02-05T10:45:40+03:00 llava: add quantization for the visual projector LLAVA, Qwen2VL (#11644)
9f4cc8f8d310b13ab3fc93a9be81b1d1376a7fa6 fd08255d0dea6625596c0367ee0a11d195f36762 Olivier Chafik ochafik@users.noreply.github.com 2025-02-05T01:00:12Z GitHub noreply@github.com 2025-02-05T01:00:12Z `sync`: minja (#11641)
fd08255d0dea6625596c0367ee0a11d195f36762 3ec9fd4b77b6aca03a3c2bf678eae3f9517d6904 Johannes Gäßler johannesg@5d6.de 2025-02-04T22:21:42+01:00 GitHub noreply@github.com 2025-02-04T22:21:42+01:00 CUDA: non-contiguous (RMS) norm support (#11659)
3ec9fd4b77b6aca03a3c2bf678eae3f9517d6904 3962fc1a7956dd0afacfbce10fd1a3ffd3ad857e fxzjshm 11426482+fxzjshm@users.noreply.github.com 2025-02-05T02:18:38+08:00 GitHub noreply@github.com 2025-02-04T19:18:38+01:00 HIP: force max threads per block to be 1024 (#11621)
3962fc1a7956dd0afacfbce10fd1a3ffd3ad857e 1bef571f6a23c36a26dabacba631763f9a893b83 Xuan-Son Nguyen thichthat@gmail.com 2025-02-04T18:25:42+01:00 GitHub noreply@github.com 2025-02-04T18:25:42+01:00 server : add try..catch to places not covered by set_exception_handler (#11620)
1bef571f6a23c36a26dabacba631763f9a893b83 db288b60cb49eab69703f995379b8d75c18bf5b3 Radoslav Gerganov rgerganov@gmail.com 2025-02-04T18:16:20+02:00 GitHub noreply@github.com 2025-02-04T18:16:20+02:00 arg : list RPC devices first when using --list-devices (#11655)
db288b60cb49eab69703f995379b8d75c18bf5b3 106045e7bb8db481bb2ebbc60e3b53cb27ada117 Olivier Chafik ochafik@users.noreply.github.com 2025-02-04T15:48:53Z GitHub noreply@github.com 2025-02-04T15:48:53Z `tool-call`: command r7b fix for normal responses (#11608)
106045e7bb8db481bb2ebbc60e3b53cb27ada117 f117d84b48104992ba16961b35a96fa93efbb355 Shelby Jenkins 47464908+ShelbyJenkins@users.noreply.github.com 2025-02-04T05:20:55-06:00 GitHub noreply@github.com 2025-02-04T13:20:55+02:00 readme : add llm_client Rust crate to readme bindings (#11628)
f117d84b48104992ba16961b35a96fa93efbb355 534c46b53c23613628d72e93c63ea01ea4d1e2ac Jhen-Jie Hong iainst0409@gmail.com 2025-02-04T19:15:24+08:00 GitHub noreply@github.com 2025-02-04T13:15:24+02:00 swift : fix llama-vocab api usage (#11645)
534c46b53c23613628d72e93c63ea01ea4d1e2ac 387a1598ca094a4755303ec964c3b09b4c5c300e Jhen-Jie Hong iainst0409@gmail.com 2025-02-04T19:07:18+08:00 GitHub noreply@github.com 2025-02-04T13:07:18+02:00 metal : use residency set for other platforms (#11648)
387a1598ca094a4755303ec964c3b09b4c5c300e 7c9e0ca52039530b6ddf16015cf1263491e92d75 Georgi Gerganov ggerganov@gmail.com 2025-02-04T13:04:10+02:00 Georgi Gerganov ggerganov@gmail.com 2025-02-04T13:04:10+02:00 authors : update
7c9e0ca52039530b6ddf16015cf1263491e92d75 8f8290ada96194138bb3d4dc6958323bc728ef83 Georgi Gerganov ggerganov@gmail.com 2025-02-04T12:59:21+02:00 Georgi Gerganov ggerganov@gmail.com 2025-02-04T12:59:21+02:00 sync : ggml
8f8290ada96194138bb3d4dc6958323bc728ef83 b34aedd558dcf67f7e96b0260d6b554877bd3499 Christian Kastner ckk@kvr.at 2025-02-04T00:17:15+01:00 Georgi Gerganov ggerganov@gmail.com 2025-02-04T12:59:15+02:00 cmake: Add ability to pass in GGML_BUILD_NUMBER (ggml/1096)
b34aedd558dcf67f7e96b0260d6b554877bd3499 cde383323959544abe10a4d79e1d3e1ee479933c Georgi Gerganov ggerganov@gmail.com 2025-02-04T09:30:42+02:00 Georgi Gerganov ggerganov@gmail.com 2025-02-04T09:31:01+02:00 ci : do not stale-close roadmap issues
cde383323959544abe10a4d79e1d3e1ee479933c b3451785aca16fbf4214eeba7e4612676cdf8ccb Olivier Chafik ochafik@users.noreply.github.com 2025-02-03T23:49:27Z GitHub noreply@github.com 2025-02-03T23:49:27Z `tool-call`: allow `--chat-template chatml` w/ `--jinja`, default to chatml upon parsing issue, avoid double bos (#11616)
b3451785aca16fbf4214eeba7e4612676cdf8ccb 1d1e6a90bcf485ad2dee309c31cf19bd802465e5 Xuan-Son Nguyen thichthat@gmail.com 2025-02-04T00:10:52+01:00 GitHub noreply@github.com 2025-02-04T00:10:52+01:00 server : (webui) revert hacky solution from #11626 (#11634)
1d1e6a90bcf485ad2dee309c31cf19bd802465e5 5598f475be3e31430fbe17ebb85654ec90dc201e Woof Dog 197125663+woof-dog@users.noreply.github.com 2025-02-03T22:16:27Z GitHub noreply@github.com 2025-02-03T23:16:27+01:00 server : (webui) allow typing and submitting during llm response (#11626)
5598f475be3e31430fbe17ebb85654ec90dc201e 8ec05832fa8409c49b3bbd13f957c6ae8486e618 Daniel Bevenius daniel.bevenius@gmail.com 2025-02-03T16:45:38+01:00 GitHub noreply@github.com 2025-02-03T16:45:38+01:00 server : remove CPPHTTPLIB_NO_EXCEPTIONS define (#11622)
8ec05832fa8409c49b3bbd13f957c6ae8486e618 21c84b5d2dc04050714567501bf78762bfa17846 Georgi Gerganov ggerganov@gmail.com 2025-02-03T14:57:08+02:00 Georgi Gerganov ggerganov@gmail.com 2025-02-03T14:57:08+02:00 sync : ggml
21c84b5d2dc04050714567501bf78762bfa17846 d92cb67e37abc23b1c6f7b0ef27a9889da8537e3 Johannes Gäßler johannesg@5d6.de 2025-02-03T13:25:56+01:00 GitHub noreply@github.com 2025-02-03T14:25:56+02:00 CUDA: fix Volta FlashAttention logic (#11615)
d92cb67e37abc23b1c6f7b0ef27a9889da8537e3 6eecde3cc8fda44da7794042e3668de4af3c32c6 mashdragon 122402293+mashdragon@users.noreply.github.com 2025-02-03T09:42:55Z GitHub noreply@github.com 2025-02-03T10:42:55+01:00 server : (webui) Fix Shift+Enter handling (#11609)
6eecde3cc8fda44da7794042e3668de4af3c32c6 396856b40029dd6747d2fbdb179e828683418045 Johannes Gäßler johannesg@5d6.de 2025-02-02T23:48:29+01:00 GitHub noreply@github.com 2025-02-02T23:48:29+01:00 HIP: fix flash_attn_stream_k_fixup warning (#11604)
396856b40029dd6747d2fbdb179e828683418045 4d0598e1445a64c99cf2faac72f8d5d023f1e6a1 uvos philipp@uvos.xyz 2025-02-02T22:40:09+01:00 GitHub noreply@github.com 2025-02-02T22:40:09+01:00 CUDA/HIP: add support for selectable warp size to mmv (#11519)
4d0598e1445a64c99cf2faac72f8d5d023f1e6a1 90f9b88afb6447d3929843a2aa98c0f11074762d uvos philipp@uvos.xyz 2025-02-02T22:08:05+01:00 GitHub noreply@github.com 2025-02-02T22:08:05+01:00 HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (#11601)
90f9b88afb6447d3929843a2aa98c0f11074762d 864a0b67a6c8f648c43ce8271f9cb2e12dd5df6e Olivier Chafik ochafik@users.noreply.github.com 2025-02-02T19:58:34Z GitHub noreply@github.com 2025-02-02T19:58:34Z nit: more informative crash when grammar sampler fails (#11593)
864a0b67a6c8f648c43ce8271f9cb2e12dd5df6e 84ec8a58f7b6aad6887bbfbd1321f3ff417341a5 Johannes Gäßler johannesg@5d6.de 2025-02-02T19:31:09+01:00 GitHub noreply@github.com 2025-02-02T19:31:09+01:00 CUDA: use mma PTX instructions for FlashAttention (#11583)
84ec8a58f7b6aad6887bbfbd1321f3ff417341a5 bfcce4d693617ec843d0b2510f6ee16e6bc6720d Eric Curtin ecurtin@redhat.com 2025-02-02T16:14:48+01:00 GitHub noreply@github.com 2025-02-02T15:14:48Z Name colors (#11573)
bfcce4d693617ec843d0b2510f6ee16e6bc6720d 69804487e0b10f2c5c06316f0ac0eb6ada68433f Olivier Chafik ochafik@users.noreply.github.com 2025-02-02T09:25:38Z GitHub noreply@github.com 2025-02-02T09:25:38Z `tool-call`: support Command R7B (+ return tool_plan "thoughts" in API) (#11585)
69804487e0b10f2c5c06316f0ac0eb6ada68433f ff227703d6d6e1888bdc7af6138514092ffcdb96 Olivier Chafik ochafik@users.noreply.github.com 2025-02-02T09:10:15Z GitHub noreply@github.com 2025-02-02T09:10:15Z Fix exotic ci env that lacks ostringstream::str (#11581)
ff227703d6d6e1888bdc7af6138514092ffcdb96 0cec062a638700495673f5494d200b74340538be Michał Moskal michal@moskal.me 2025-02-01T23:55:32-08:00 GitHub noreply@github.com 2025-02-02T09:55:32+02:00 sampling : support for llguidance grammars (#10224)
0cec062a638700495673f5494d200b74340538be 53debe6f3c9cca87e9520a83ee8c14d88977afa4 piDack 104877312+piDack@users.noreply.github.com 2025-02-02T15:48:46+08:00 GitHub noreply@github.com 2025-02-02T09:48:46+02:00 llama : add support for GLM-Edge and GLM-Edge-V series models (#10573)
53debe6f3c9cca87e9520a83ee8c14d88977afa4 cfd74c86dbaa95ed30aa6b30e14d8801eb975d63 Olivier Chafik ochafik@users.noreply.github.com 2025-02-01T18:22:38Z GitHub noreply@github.com 2025-02-01T18:22:38Z ci: use sccache on windows HIP jobs (#11553)
cfd74c86dbaa95ed30aa6b30e14d8801eb975d63 ecef206ccb186a1cde8dd2523b1da3e12f593f9e Olivier Chafik ochafik@users.noreply.github.com 2025-02-01T12:24:51Z GitHub noreply@github.com 2025-02-01T12:24:51Z `sync`: minja (https://github.com/google/minja/commit/418a2364b56dc9be4ed9a1a2b0fb16fb53a7a22e) (#11574)
ecef206ccb186a1cde8dd2523b1da3e12f593f9e 5bbc7362cb93265f4c853fd89800a6255cc26985 Eric Curtin ecurtin@redhat.com 2025-02-01T11:30:54+01:00 GitHub noreply@github.com 2025-02-01T10:30:54Z Implement s3:// protocol (#11511)
5bbc7362cb93265f4c853fd89800a6255cc26985 aa6fb1321333fae8853d0cdc26bcb5d438e650a1 Olivier Chafik ochafik@users.noreply.github.com 2025-02-01T00:01:20Z GitHub noreply@github.com 2025-02-01T00:01:20Z ci: simplify cmake build commands (#11548)
aa6fb1321333fae8853d0cdc26bcb5d438e650a1 a83f528688324a21484a97af1d1be5e1bc8d4c8e Olivier Chafik ochafik@users.noreply.github.com 2025-01-31T17:12:40Z GitHub noreply@github.com 2025-01-31T17:12:40Z `ci`: use sccache on windows instead of ccache (#11545)
a83f528688324a21484a97af1d1be5e1bc8d4c8e b1bcd309fc8ac929cbd4a6207b3a19886bda031f Olivier Chafik ochafik@users.noreply.github.com 2025-01-31T14:15:25Z GitHub noreply@github.com 2025-01-31T14:15:25Z `tool-call`: fix llama 3.x and functionary 3.2, play nice w/ pydantic_ai package, update readme (#11539)
b1bcd309fc8ac929cbd4a6207b3a19886bda031f 5783575c9d99c4d9370495800663aa5397ceb0be Olivier Chafik ochafik@users.noreply.github.com 2025-01-31T13:48:31Z GitHub noreply@github.com 2025-01-31T13:48:31Z fix stop regression (#11543)
5783575c9d99c4d9370495800663aa5397ceb0be 4a2b196d03d52da31236390e9f5694a88d43d11d Olivier Chafik ochafik@users.noreply.github.com 2025-01-31T08:24:29Z GitHub noreply@github.com 2025-01-31T08:24:29Z Fix chatml fallback for unsupported builtin templates (when --jinja not enabled) (#11533)
4a2b196d03d52da31236390e9f5694a88d43d11d 1bd3047a939e561adfb3c7dd2e17c4cc7a4e4e6f Olivier Chafik ochafik@users.noreply.github.com 2025-01-31T08:12:40Z GitHub noreply@github.com 2025-01-31T10:12:40+02:00 server : fix --jinja when there's no tools or schema (typo was forcing JSON) (#11531)
1bd3047a939e561adfb3c7dd2e17c4cc7a4e4e6f a2df2787b32e0846205f7151dfad88ceab592beb Steve Grubb ausearch.1@gmail.com 2025-01-31T00:58:55-05:00 GitHub noreply@github.com 2025-01-31T07:58:55+02:00 common: Add missing va_end (#11529)
a2df2787b32e0846205f7151dfad88ceab592beb 553f1e46e9e864514bbd6bf4009146db66be0541 Daniel Bevenius daniel.bevenius@gmail.com 2025-01-31T06:04:53+01:00 GitHub noreply@github.com 2025-01-31T06:04:53+01:00 server : update help metrics processing/deferred (#11512)
553f1e46e9e864514bbd6bf4009146db66be0541 8b576b6c55bc4e6be898b47522f0ef402b93ef62 Olivier Chafik ochafik@users.noreply.github.com 2025-01-30T22:01:06Z GitHub noreply@github.com 2025-01-30T22:01:06Z `ci`: ccache for all github worfklows (#11516)
8b576b6c55bc4e6be898b47522f0ef402b93ef62 27d135c970c00f655d486f870edacded792bef5c Olivier Chafik ochafik@users.noreply.github.com 2025-01-30T19:13:58Z GitHub noreply@github.com 2025-01-30T19:13:58Z Tool call support (generic + native for Llama, Functionary, Hermes, Mistral, Firefunction, DeepSeek) w/ lazy grammars (#9639)
27d135c970c00f655d486f870edacded792bef5c 6af1ca48cbdf9a438438afd0a9a549a272bc95bf uvos devnull@uvos.xyz 2025-01-29T19:36:00+01:00 uvos devnull@uvos.xyz 2025-01-30T16:25:44+01:00 HIP: require at least HIP 5.5
6af1ca48cbdf9a438438afd0a9a549a272bc95bf c300e68ef490e6cf6c04ed96fd27a6a53ab8a422 uvos devnull@uvos.xyz 2025-01-29T19:12:42+01:00 uvos devnull@uvos.xyz 2025-01-30T16:25:44+01:00 HIP: Prepare reduction operators for wave 64
c300e68ef490e6cf6c04ed96fd27a6a53ab8a422 3d804dec7661fbb7de9b7f93267e2fc3ca0193c1 uvos devnull@uvos.xyz 2025-01-29T17:46:23+01:00 uvos devnull@uvos.xyz 2025-01-30T16:25:44+01:00 CUDA/HIP: add warp_size to cuda_device_info
3d804dec7661fbb7de9b7f93267e2fc3ca0193c1 ffd0821c57edc7e5d04338ab0c6b1461198df15f Olivier Chafik ochafik@users.noreply.github.com 2025-01-30T10:30:27Z GitHub noreply@github.com 2025-01-30T10:30:27Z sync: minja (#11499)
ffd0821c57edc7e5d04338ab0c6b1461198df15f 4314e56c4f8c5091f45732f39bd94c0c6c323798 mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2025-01-30T11:10:59+01:00 GitHub noreply@github.com 2025-01-30T12:10:59+02:00 vocab : correctly identify LF token for GPT-2 style BPE tokenizer (#11496)
4314e56c4f8c5091f45732f39bd94c0c6c323798 496e5bf46bab757d05e18227cb4e17055ae42f42 Daniel Bevenius daniel.bevenius@gmail.com 2025-01-30T11:05:00+01:00 GitHub noreply@github.com 2025-01-30T11:05:00+01:00 server : use lambda instead of std::bind (#11507)
496e5bf46bab757d05e18227cb4e17055ae42f42 7919256c57f05c09b0b50ec9abb37ff62dab7251 Isaac McFadyen isaac@imcf.me 2025-01-30T04:11:53-05:00 GitHub noreply@github.com 2025-01-30T10:11:53+01:00 server : (docs) added response format for /apply-template [no ci] (#11503)
7919256c57f05c09b0b50ec9abb37ff62dab7251 e0449763a4f335cca374254a72892141f41eaa59 Guspan Tanadi 36249910+guspan-tanadi@users.noreply.github.com 2025-01-30T12:58:02+07:00 GitHub noreply@github.com 2025-01-30T06:58:02+01:00 readme : reference examples relative links (#11505)
e0449763a4f335cca374254a72892141f41eaa59 eb7cf15a808d4d7a71eef89cc6a9b96fe82989dc Daniel Bevenius daniel.bevenius@gmail.com 2025-01-30T05:48:14+01:00 GitHub noreply@github.com 2025-01-30T05:48:14+01:00 server : update json snippets in README.md [no ci] (#11492)
eb7cf15a808d4d7a71eef89cc6a9b96fe82989dc 66ee4f297cff3c7ce98b31dbc0ce909d41b9e408 Nigel Bosch pnigelb@gmail.com 2025-01-29T12:45:44-06:00 GitHub noreply@github.com 2025-01-29T19:45:44+01:00 server : add /apply-template endpoint for additional use cases of Minja functionality (#11489)
66ee4f297cff3c7ce98b31dbc0ce909d41b9e408 e51c47b401f8cb5f21630a05171e2529cde4d186 Rémy Oudompheng oudomphe@phare.normalesup.org 2025-01-29T18:29:39+01:00 GitHub noreply@github.com 2025-01-29T18:29:39+01:00 vulkan: implement initial support for IQ2 and IQ3 quantizations (#11360)
e51c47b401f8cb5f21630a05171e2529cde4d186 2711d0215ff6a1ecb2202ac8c1f135bceed7057b Daniel Bevenius daniel.bevenius@gmail.com 2025-01-29T16:34:18+01:00 GitHub noreply@github.com 2025-01-29T16:34:18+01:00 server : update auto gen files comments [no ci] (#11484)
2711d0215ff6a1ecb2202ac8c1f135bceed7057b f0d4b29edfc633ec3ba6442482a6b43a5cd80a01 Jeff Bolz jbolz@nvidia.com 2025-01-29T09:26:50-06:00 GitHub noreply@github.com 2025-01-29T09:26:50-06:00 vulkan: Catch pipeline creation failure and print an error message (#11436)
f0d4b29edfc633ec3ba6442482a6b43a5cd80a01 815857791d3639a4d544d0a8cf25a49b0325c08c Eric Curtin ecurtin@redhat.com 2025-01-29T12:23:10+01:00 GitHub noreply@github.com 2025-01-29T11:23:10Z Parse https://ollama.com/library/ syntax (#11480)
815857791d3639a4d544d0a8cf25a49b0325c08c 1a0e87d29152cb9d4ce13b4ad64b0382c9ba1ab6 Georgi Gerganov ggerganov@gmail.com 2025-01-29T11:25:29+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-29T11:25:29+02:00 sync : ggml
1a0e87d29152cb9d4ce13b4ad64b0382c9ba1ab6 d2e518e9b4231077013ffb0e5c4cc24a0c4c1b7d William Tambellini wtambellini@sdl.com 2025-01-23T11:59:08-08:00 Georgi Gerganov ggerganov@gmail.com 2025-01-29T11:24:53+02:00 ggml : add option to not print stack on abort (ggml/1081)
d2e518e9b4231077013ffb0e5c4cc24a0c4c1b7d b636228c0ad0db95bf73008094c6145a05615cf6 issixx 46835150+issixx@users.noreply.github.com 2025-01-17T21:29:08+09:00 Georgi Gerganov ggerganov@gmail.com 2025-01-29T11:24:51+02:00 ggml-cpu : fix ggml_graph_compute_thread did not terminate on abort. (ggml/1065)
b636228c0ad0db95bf73008094c6145a05615cf6 325afb370a1a7b32b5fe46a749bc840c66db9765 Daniel Bevenius daniel.bevenius@gmail.com 2025-01-29T09:38:54+01:00 GitHub noreply@github.com 2025-01-29T10:38:54+02:00 embedding : enable --no-warmup option (#11475)
325afb370a1a7b32b5fe46a749bc840c66db9765 794fe23f29fb40104975c91fe19f23798f7c726e Molly Sophia mollysophia379@gmail.com 2025-01-29T12:07:21+08:00 GitHub noreply@github.com 2025-01-29T12:07:21+08:00 llama: fix missing k_cache store for rwkv6qwen2 (#11445)
794fe23f29fb40104975c91fe19f23798f7c726e cf8cc856d7d02165bd08593b4757e1256a62d501 Emreerdog 34742675+Emreerdog@users.noreply.github.com 2025-01-29T02:22:06+03:00 GitHub noreply@github.com 2025-01-28T19:22:06-04:00 cmake: add hints for locating ggml on Windows using Llama find-package (#11466)
cf8cc856d7d02165bd08593b4757e1256a62d501 d0c08040b6c8bebeade7b8d5764df6cf901678d5 peidaqi peidaqi@gmail.com 2025-01-28T16:03:42-07:00 GitHub noreply@github.com 2025-01-29T00:03:42+01:00 server : Fixed wrong function name in llamacpp server unit test (#11473)
d0c08040b6c8bebeade7b8d5764df6cf901678d5 be5ef7963fcf14a9c77c963fdd3f7b606eacb498 Xuan-Son Nguyen thichthat@gmail.com 2025-01-29T00:02:56+01:00 GitHub noreply@github.com 2025-01-29T00:02:56+01:00 ci : fix build CPU arm64 (#11472)
be5ef7963fcf14a9c77c963fdd3f7b606eacb498 cae9fb4361138b937464524eed907328731b81f6 uvos philipp@uvos.xyz 2025-01-28T23:06:32+01:00 GitHub noreply@github.com 2025-01-28T23:06:32+01:00 HIP: Supress transformation warning in softmax.cu
cae9fb4361138b937464524eed907328731b81f6 7fee2889e6565830631fbe76d47ef85cf8fd946a Nikita Sarychev 42014488+sARY77@users.noreply.github.com 2025-01-28T07:42:20-08:00 GitHub noreply@github.com 2025-01-28T16:42:20+01:00 HIP: Only call rocblas_initialize on rocblas versions with the multiple instantation bug (#11080)
7fee2889e6565830631fbe76d47ef85cf8fd946a d7d1eccacccaa698c9232014b96a82b359595d6e Eric Curtin ecurtin@redhat.com 2025-01-28T15:45:41+01:00 GitHub noreply@github.com 2025-01-28T14:45:41Z Add github protocol pulling and http:// (#11465)
d7d1eccacccaa698c9232014b96a82b359595d6e 4bf3119d61c1de5660025fd5a611effe503e3d2b Nuno rare-magma@posteo.eu 2025-01-28T15:17:25+01:00 GitHub noreply@github.com 2025-01-28T14:17:25Z docker: allow installing pip packages system-wide (#11437)
4bf3119d61c1de5660025fd5a611effe503e3d2b f643120bad8ab3a753daa64aaac8288ee5800e06 someone13574 81528246+someone13574@users.noreply.github.com 2025-01-28T09:15:34-05:00 GitHub noreply@github.com 2025-01-28T15:15:34+01:00 cmake : don't fail on `GGML_CPU=OFF` (#11457)
f643120bad8ab3a753daa64aaac8288ee5800e06 6e84b0ab8e10b8f6f99a32855f976ebcd35b0353 Nuno rare-magma@posteo.eu 2025-01-28T11:42:32+01:00 GitHub noreply@github.com 2025-01-28T10:42:32Z docker: add perplexity and bench commands to full image (#11438)
6e84b0ab8e10b8f6f99a32855f976ebcd35b0353 2b8525d5c89b124c4578a2621cbeb64354ff3d9c Akarshan Biswas akarshan.biswas@gmail.com 2025-01-28T15:26:58+05:30 GitHub noreply@github.com 2025-01-28T09:56:58Z SYCL : SOFTMAX F16 mask support and other fixes (#11261)
2b8525d5c89b124c4578a2621cbeb64354ff3d9c a4417ddda98fd0558fb4d802253e68a933704b59 Michael Engel mengel@redhat.com 2025-01-28T09:32:40+01:00 GitHub noreply@github.com 2025-01-28T08:32:40Z Handle missing model in CLI parameters for llama-run (#11399)
a4417ddda98fd0558fb4d802253e68a933704b59 d6d24cd9ed6d0b9558643dcc28f2124bef488c52 Eric Curtin ecurtin@redhat.com 2025-01-27T19:36:10+01:00 GitHub noreply@github.com 2025-01-27T19:36:10+01:00 Add new hf protocol for ollama (#11449)
d6d24cd9ed6d0b9558643dcc28f2124bef488c52 a5203b4465c5c87813936bde98170e25bb09024f Haus1 haus.xda@gmail.com 2025-01-27T08:58:17-05:00 GitHub noreply@github.com 2025-01-27T14:58:17+01:00 AMD: parse the architecture as supplied by gcnArchName (#11244)
a5203b4465c5c87813936bde98170e25bb09024f df984e014714cba4c99ef894b20b51cbcef31b16 lexasub lexakopp2212@gmail.com 2025-01-27T17:42:09+04:00 GitHub noreply@github.com 2025-01-27T14:42:09+01:00 llama : minor fixes for up llama load model speed (#11448)
df984e014714cba4c99ef894b20b51cbcef31b16 acd38efee316f3a5ed2e6afcbc5814807c347053 Johannes Gäßler johannesg@5d6.de 2025-01-27T12:07:12+01:00 GitHub noreply@github.com 2025-01-27T12:07:12+01:00 llama: refactor llama_decode_impl (#11381)
acd38efee316f3a5ed2e6afcbc5814807c347053 caf773f249aa267c78d3da5567b8ab156080ea59 Ihar Hrachyshka ihrachys@redhat.com 2025-01-27T02:41:59-05:00 GitHub noreply@github.com 2025-01-27T09:41:59+02:00 metal: Handle null returned from MTLCreateSystemDefaultDevice() (#11441)
caf773f249aa267c78d3da5567b8ab156080ea59 178a7eb952d211b8d4232d5e50ae1b64519172a9 Xuan Son Nguyen thichthat@gmail.com 2025-01-26T22:45:32+01:00 GitHub noreply@github.com 2025-01-26T22:45:32+01:00 docker : fix ARM build and Vulkan build (#11434)
178a7eb952d211b8d4232d5e50ae1b64519172a9 6f53d8a6b41e48c73b345fc6c712c3b00ea4fb93 Georgi Gerganov ggerganov@gmail.com 2025-01-26T20:06:16+02:00 GitHub noreply@github.com 2025-01-26T20:06:16+02:00 metal : use residency sets (#11427)
6f53d8a6b41e48c73b345fc6c712c3b00ea4fb93 19f65187cbf009801288861133267ee5573ceead Nuno rare-magma@posteo.eu 2025-01-26T18:22:43+01:00 GitHub noreply@github.com 2025-01-26T18:22:43+01:00 docker: add missing vulkan library to base layer and update to 24.04 (#11422)
19f65187cbf009801288861133267ee5573ceead 1d8ee06000ecdd274e7f0a0465d6bf26ad2b3491 bandoti 141645996+bandoti@users.noreply.github.com 2025-01-26T12:07:48-04:00 GitHub noreply@github.com 2025-01-26T12:07:48-04:00 cmake: add ggml find package (#11369)
1d8ee06000ecdd274e7f0a0465d6bf26ad2b3491 2cc9b8c32c78d09cd1b4df0aaa605ab2d0176243 Frank Mai thxcode0824@gmail.com 2025-01-26T23:20:34+08:00 GitHub noreply@github.com 2025-01-26T16:20:34+01:00 rpc: fix register position (#11424)
2cc9b8c32c78d09cd1b4df0aaa605ab2d0176243 f35726c2fb0a824246e004ab4bedcde37f3f0dd0 Georgi Gerganov ggerganov@gmail.com 2025-01-26T14:30:15+02:00 GitHub noreply@github.com 2025-01-26T14:30:15+02:00 readme : update hot topics
f35726c2fb0a824246e004ab4bedcde37f3f0dd0 4a75d19376f2f00dbae6c266eb9c4f3001872b52 Jeff Bolz jbolz@nvidia.com 2025-01-25T20:10:03-06:00 GitHub noreply@github.com 2025-01-26T03:10:03+01:00 build: apply MSVC /bigobj option to c/cpp files only (#11423)
4a75d19376f2f00dbae6c266eb9c4f3001872b52 26771a1491f3a4c3d5b99c4c267b81aca9a7dfa0 Jeff Bolz jbolz@nvidia.com 2025-01-25T15:29:57-06:00 GitHub noreply@github.com 2025-01-25T22:29:57+01:00 vulkan: compile shaders on-demand (#11406)
26771a1491f3a4c3d5b99c4c267b81aca9a7dfa0 ca6baf76c1a7adb9134b08d2bc4c65557297ff87 uvos devnull@uvos.xyz 2025-01-25T21:01:12+01:00 GitHub noreply@github.com 2025-01-25T21:01:12+01:00 Hip: disable VMM on hip as it seams that it dosent work in some configurations (#11420)
ca6baf76c1a7adb9134b08d2bc4c65557297ff87 6e264a905bec9e4c0111eb4c91379c88accef7c6 Jeff Bolz jbolz@nvidia.com 2025-01-25T11:26:37-06:00 GitHub noreply@github.com 2025-01-25T11:26:37-06:00 build: add /bigobj to MSVC build (#11407)
6e264a905bec9e4c0111eb4c91379c88accef7c6 49b0e3cec4b67dc9f4debe3a16acd4c819f751d6 Diego Devesa slarengh@gmail.com 2025-01-25T17:22:41+01:00 GitHub noreply@github.com 2025-01-25T17:22:41+01:00 docker : add GGML_CPU_ARM_ARCH arg to select ARM architecture to build for (#11419)
49b0e3cec4b67dc9f4debe3a16acd4c819f751d6 20a758155bc5f37290b20ea44d76ba99c4e7f2cb Xuan Son Nguyen thichthat@gmail.com 2025-01-25T16:36:44+01:00 GitHub noreply@github.com 2025-01-25T16:36:44+01:00 server : fix cleaning up stream task (#11418)
20a758155bc5f37290b20ea44d76ba99c4e7f2cb 00c24acb2ac49d9f8318e808b6ada2f5649f253f Diego Devesa slarengh@gmail.com 2025-01-25T15:22:29+01:00 GitHub noreply@github.com 2025-01-25T15:22:29+01:00 docker : fix CPU ARM build (#11403)
00c24acb2ac49d9f8318e808b6ada2f5649f253f 466ea66f338d63109540dae1df97ccfdbf4cd08f Georgi Gerganov ggerganov@gmail.com 2025-01-25T13:36:48+02:00 GitHub noreply@github.com 2025-01-25T13:36:48+02:00 ci : fix line breaks on windows builds (#11409)
466ea66f338d63109540dae1df97ccfdbf4cd08f 5f0db9522f347b095f84c3033d6c1c1895402e25 jiahao su damow890@gmail.com 2025-01-25T07:26:01+08:00 GitHub noreply@github.com 2025-01-25T00:26:01+01:00 CANN: Add Ascend CANN build ci (#10217)
5f0db9522f347b095f84c3033d6c1c1895402e25 c5d9effb49649db80a52caf5c0626de6f342f526 uvos devnull@uvos.xyz 2025-01-25T00:02:23+01:00 GitHub noreply@github.com 2025-01-25T00:02:23+01:00 hip : Add hipGraph and VMM support to ROCM (#11362)
c5d9effb49649db80a52caf5c0626de6f342f526 9fbadaef4f0903c64895ba9c70f02ac6e6a4b41c Johannes Gäßler johannesg@5d6.de 2025-01-24T21:02:43+01:00 GitHub noreply@github.com 2025-01-24T21:02:43+01:00 CUDA: fix FP16 cuBLAS GEMM (#11396)
9fbadaef4f0903c64895ba9c70f02ac6e6a4b41c 9755129c27da76d768bd7e47e206bac61b40cf18 uvos devnull@uvos.xyz 2025-01-24T17:50:49+01:00 GitHub noreply@github.com 2025-01-24T17:50:49+01:00 rocBLAS: Avoid fp32->fp16->fp32 conversion on cdna (#11356)
9755129c27da76d768bd7e47e206bac61b40cf18 a07c2c8a52464646ce13040e62c1ea04459f721e Georgi Gerganov ggerganov@gmail.com 2025-01-24T18:41:30+02:00 GitHub noreply@github.com 2025-01-24T18:41:30+02:00 release : pack /lib in the packages (#11392)
a07c2c8a52464646ce13040e62c1ea04459f721e 8137b4bb2b5fd4cb4a752bfe69ccfd915a313d58 Jafar Uruç jafar.uruc@gmail.com 2025-01-24T13:30:13Z GitHub noreply@github.com 2025-01-24T14:30:13+01:00 docs : Update readme to build targets for local docker build (#11368)
8137b4bb2b5fd4cb4a752bfe69ccfd915a313d58 1af6945eb0d0e97525dc0ec18167abf05c28f482 Johannes Gäßler johannesg@5d6.de 2025-01-24T12:38:31+01:00 GitHub noreply@github.com 2025-01-24T12:38:31+01:00 CPU/CUDA: fix (GQA) mul mat back, add CUDA support (#11380)
1af6945eb0d0e97525dc0ec18167abf05c28f482 01f37edf1a6fae76fd9e2e02109aae6995a914f0 Bernhard M. Wiedemann githubbmwprimary@lsmod.de 2025-01-24T12:21:35+01:00 GitHub noreply@github.com 2025-01-24T13:21:35+02:00 cmake : avoid -march=native when reproducible build is wanted (#11366)
01f37edf1a6fae76fd9e2e02109aae6995a914f0 c07e87f38bd0c22ec6dbc852ae50aaa1c64632d4 Eric Curtin ecurtin@redhat.com 2025-01-24T09:39:24Z GitHub noreply@github.com 2025-01-24T09:39:24Z Update llama-run README.md (#11386)
c07e87f38bd0c22ec6dbc852ae50aaa1c64632d4 564804b79b78df1469ec8646869972de5e885ec4 stduhpf stephduh@live.fr 2025-01-24T09:02:38+01:00 GitHub noreply@github.com 2025-01-24T09:02:38+01:00 server : (webui) put DeepSeek R1 CoT in a collapsible <details> element (#11364)
564804b79b78df1469ec8646869972de5e885ec4 05f63cc9ee859de07f585f7b12939345f39ada8b Jeff Bolz jbolz@nvidia.com 2025-01-23T14:51:24-06:00 GitHub noreply@github.com 2025-01-23T14:51:24-06:00 tests: fix some mul_mat test gaps (#11375)
05f63cc9ee859de07f585f7b12939345f39ada8b f7fb43cd0b84280c261f440dc8e85eafad4a0ca6 Eric Curtin ecurtin@redhat.com 2025-01-23T20:04:31Z GitHub noreply@github.com 2025-01-23T20:04:31Z Update documentation (#11373)
f7fb43cd0b84280c261f440dc8e85eafad4a0ca6 58456616408c828a1ce6d2481940fd1c97bce3b5 Eric Curtin ecurtin@redhat.com 2025-01-23T16:16:18Z GitHub noreply@github.com 2025-01-23T16:16:18Z Add -ngl (#11372)
58456616408c828a1ce6d2481940fd1c97bce3b5 f211d1dc10332b7e89a4abd1041903fa63bfed27 Xuan Son Nguyen thichthat@gmail.com 2025-01-23T13:56:05+01:00 GitHub noreply@github.com 2025-01-23T13:56:05+01:00 server : add more clean up when cancel_tasks is called (#11340)
f211d1dc10332b7e89a4abd1041903fa63bfed27 955a6c2d91480954e90469517c4b91c4052c6a59 Eric Curtin ecurtin@redhat.com 2025-01-23T10:38:20Z GitHub noreply@github.com 2025-01-23T10:38:20Z Treat hf.co/ prefix the same as hf:// (#11350)
955a6c2d91480954e90469517c4b91c4052c6a59 1971adf55e3ebbfab83771d6174d37b77c352c71 amd-dwang dong.wang@amd.com 2025-01-23T15:14:28+08:00 GitHub noreply@github.com 2025-01-23T08:14:28+01:00 Vulkan-run-test: fix mmq_wg_denoms (#11343)
1971adf55e3ebbfab83771d6174d37b77c352c71 5245729e3317064959faefc5e5cbc63f4e9cfbea Jeff Bolz jbolz@nvidia.com 2025-01-23T01:07:50-06:00 GitHub noreply@github.com 2025-01-23T08:07:50+01:00 vulkan: sort shaders for more deterministic binary (#11315)
5245729e3317064959faefc5e5cbc63f4e9cfbea 6152129d05870cb38162c422c6ba80434e021e9f Jeff Bolz jbolz@nvidia.com 2025-01-23T01:01:17-06:00 GitHub noreply@github.com 2025-01-23T08:01:17+01:00 vulkan: fix diag_mask_inf (#11323)
6152129d05870cb38162c422c6ba80434e021e9f 16d3df7ab0bb9def78047eab4d9b15393997f495 Diego Devesa slarengh@gmail.com 2025-01-22T19:22:20+01:00 GitHub noreply@github.com 2025-01-22T19:22:20+01:00 main : update README documentation for batch size (#11353)
16d3df7ab0bb9def78047eab4d9b15393997f495 12c2bdf2de34f747d13b270fc9d3b52490bf194f Georgi Gerganov ggerganov@gmail.com 2025-01-22T19:44:26+02:00 GitHub noreply@github.com 2025-01-22T19:44:26+02:00 readme : add plugin links (#11355)
12c2bdf2de34f747d13b270fc9d3b52490bf194f c64d2becb13f2a7c0145b516a05f028d949a046b Diego Devesa slarengh@gmail.com 2025-01-22T17:44:40+01:00 GitHub noreply@github.com 2025-01-22T17:44:40+01:00 server : fix draft context not being released (#11354)
c64d2becb13f2a7c0145b516a05f028d949a046b 96f405393461062450692430e4916809bf71c3c4 Olivier Chafik ochafik@users.noreply.github.com 2025-01-22T16:16:27Z GitHub noreply@github.com 2025-01-22T16:16:27Z `minja`: sync at https://github.com/google/minja/commit/0f5f7f2b3770eb682fbc11763266d45204173686 (#11352)
96f405393461062450692430e4916809bf71c3c4 a94f3b2727e97eb6c904006eb786960c069282bc Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2025-01-22T12:51:32+01:00 GitHub noreply@github.com 2025-01-22T12:51:32+01:00 Adding logprobs to /v1/completions (#11344)
a94f3b2727e97eb6c904006eb786960c069282bc 3e3357fd77bcf5bd8cfcc53ca53d4a5532e67e1b Olivier Chafik ochafik@users.noreply.github.com 2025-01-22T09:51:44Z GitHub noreply@github.com 2025-01-22T09:51:44Z `common`: utils to split / join / repeat strings (from json converter) (#11342)
3e3357fd77bcf5bd8cfcc53ca53d4a5532e67e1b 6171c9d25820ccf676b243c172868819d882848f tc-mb 157115220+tc-mb@users.noreply.github.com 2025-01-22T15:35:48+08:00 GitHub noreply@github.com 2025-01-22T09:35:48+02:00 llava : support Minicpm-omni (#11289)
6171c9d25820ccf676b243c172868819d882848f e28245f35f2faaf249dd352998b3693a8cc28c51 Olivier Chafik ochafik@users.noreply.github.com 2025-01-21T13:18:51Z GitHub noreply@github.com 2025-01-21T13:18:51Z Add Jinja template support (#11016)
e28245f35f2faaf249dd352998b3693a8cc28c51 6da5bec81c34f3b8a8f1b367cf23ad016e83d332 Xuan Son Nguyen thichthat@gmail.com 2025-01-21T14:07:12+01:00 GitHub noreply@github.com 2025-01-21T14:07:12+01:00 export-lora : fix tok_embd tensor (#11330)
6da5bec81c34f3b8a8f1b367cf23ad016e83d332 2e2f8f093cd4fb6bbb87ba84f6b9684fa082f3fa Radoslav Gerganov rgerganov@gmail.com 2025-01-21T15:06:41+02:00 GitHub noreply@github.com 2025-01-21T15:06:41+02:00 rpc : better caching of the base buffer pointer (#11331)
2e2f8f093cd4fb6bbb87ba84f6b9684fa082f3fa 2139667ec419cdd953987f4df2ace9da87bbda28 Eric Curtin ecurtin@redhat.com 2025-01-21T09:32:35Z GitHub noreply@github.com 2025-01-21T09:32:35Z linenoise.cpp refactoring (#11301)
2139667ec419cdd953987f4df2ace9da87bbda28 80d0d6b4b7abca342fc990399e78af8d213eabaf Georgi Gerganov ggerganov@gmail.com 2025-01-21T08:48:13+02:00 GitHub noreply@github.com 2025-01-21T08:48:13+02:00 metal : fix out-of-bounds write (#11314)
80d0d6b4b7abca342fc990399e78af8d213eabaf aea8ddd5165d525a449e2fc3839db77a71f4a318 Georgi Gerganov ggerganov@gmail.com 2025-01-20T22:29:43+02:00 GitHub noreply@github.com 2025-01-20T22:29:43+02:00 common : add -hfd option for the draft model (#11318)
aea8ddd5165d525a449e2fc3839db77a71f4a318 9f7add1cde670ff744b791f5ed6649e86a0c586c Jeff Bolz jbolz@nvidia.com 2025-01-20T10:38:32-06:00 GitHub noreply@github.com 2025-01-20T10:38:32-06:00 vulkan: fix coopmat2 validation failures (#11284)
9f7add1cde670ff744b791f5ed6649e86a0c586c 90d987b105db6016cb395f673b4d21e02129721e Georgi Gerganov ggerganov@gmail.com 2025-01-20T16:36:08+02:00 GitHub noreply@github.com 2025-01-20T16:36:08+02:00 examples : fix add_special conditions (#11311)
90d987b105db6016cb395f673b4d21e02129721e a4251edd6fc16d168f517a7e4b0936212b296603 Christopher Nielsen 62156882+mascguy@users.noreply.github.com 2025-01-20T09:02:43-05:00 GitHub noreply@github.com 2025-01-20T16:02:43+02:00 mmap: add include for cerrno (#11296)
a4251edd6fc16d168f517a7e4b0936212b296603 ec7f3ac9ab33e46b136eb5ab6a76c4d81f57c7f1 Michael Podvitskiy podvitskiymichael@gmail.com 2025-01-20T15:02:15+01:00 GitHub noreply@github.com 2025-01-20T16:02:15+02:00 cmake: fix shell command quoting in build-info script (#11309)
ec7f3ac9ab33e46b136eb5ab6a76c4d81f57c7f1 ef6dada60ca710f4edfbb6fd9e1258685d8ea49d Xuan Son Nguyen thichthat@gmail.com 2025-01-20T14:35:07+01:00 GitHub noreply@github.com 2025-01-20T14:35:07+01:00 llama : add support for Deepseek-R1-Qwen distill model (#11310)
ef6dada60ca710f4edfbb6fd9e1258685d8ea49d ae3c1db2f9c4beec0737c6a82d1f3791fd6fcdb2 Georgi Gerganov ggerganov@gmail.com 2025-01-20T09:29:32+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-20T09:29:32+02:00 cont : fix whitespaces (#11305)
ae3c1db2f9c4beec0737c6a82d1f3791fd6fcdb2 92bc493917d43b83e592349e138b54c90b1c3ea7 Kyle Bruene KyleBruene@users.noreply.github.com 2025-01-20T01:21:01-06:00 GitHub noreply@github.com 2025-01-20T09:21:01+02:00 llama : re-add LLM_ARCH_PHIMOE (#11305)
92bc493917d43b83e592349e138b54c90b1c3ea7 b9daaffe02d6a77d85f0420bce5dfe0e00daeff6 Georgi Gerganov ggerganov@gmail.com 2025-01-19T20:22:30+02:00 GitHub noreply@github.com 2025-01-19T20:22:30+02:00 tests : increase timeout when sanitizers are enabled (#11300)
b9daaffe02d6a77d85f0420bce5dfe0e00daeff6 99487b57d47e14dc342b7b89d238ca11c0345241 Georgi Gerganov ggerganov@gmail.com 2025-01-19T18:12:09+02:00 GitHub noreply@github.com 2025-01-19T18:12:09+02:00 simple-chat : fix BOS being added to each message (#11278)
99487b57d47e14dc342b7b89d238ca11c0345241 a1649cc13f89946322358f92ea268ae1b7b5096c Nicolò Scipione nicolo.scipione@codeplay.com 2025-01-19T14:33:34+01:00 GitHub noreply@github.com 2025-01-19T21:33:34+08:00 SYCL: Introducing memory host pool (#11251)
a1649cc13f89946322358f92ea268ae1b7b5096c 4dd34ff83165a483ebff7bd43621b28490fa1fd6 Eric Curtin ecurtin@redhat.com 2025-01-18T14:42:31Z GitHub noreply@github.com 2025-01-18T14:42:31Z Adding linenoise.cpp to llama-run (#11252)
4dd34ff83165a483ebff7bd43621b28490fa1fd6 f30f099228f774209aa3010b78dfbe5d262e69aa Georgi Gerganov ggerganov@gmail.com 2025-01-18T16:18:15+02:00 GitHub noreply@github.com 2025-01-18T16:18:15+02:00 cmake : add sanitizer flags for llama.cpp (#11279)
f30f099228f774209aa3010b78dfbe5d262e69aa f26c87417999209e7f4576b4f3ecf7a5b9c66a29 Xuan Son Nguyen thichthat@gmail.com 2025-01-18T14:12:05+01:00 GitHub noreply@github.com 2025-01-18T14:12:05+01:00 server : implement cancellable request (#11285)
f26c87417999209e7f4576b4f3ecf7a5b9c66a29 6390a998bfc63241fde8509022b0768a71bf20bb Georgi Gerganov ggerganov@gmail.com 2025-01-18T13:18:32+02:00 GitHub noreply@github.com 2025-01-18T13:18:32+02:00 scripts : restore hf.sh (#11288)
6390a998bfc63241fde8509022b0768a71bf20bb 44e18ef93995f3040660750b527e5becf85899d0 LostRuins Concedo 39025047+LostRuins@users.noreply.github.com 2025-01-18T18:20:57+08:00 GitHub noreply@github.com 2025-01-18T12:20:57+02:00 tts : add guide tokens support (#11186)
44e18ef93995f3040660750b527e5becf85899d0 3edfa7d3753c29e44b964c0ff424d2ea8d5fdee6 Jeff Bolz jbolz@nvidia.com 2025-01-18T02:26:50-06:00 GitHub noreply@github.com 2025-01-18T09:26:50+01:00 vulkan: fix coopmat2 flash attention for non-contiguous inputs (#11281)
3edfa7d3753c29e44b964c0ff424d2ea8d5fdee6 667d72846c06b2cf4f7c8a4265e210991a49706b codezjx code.zjx@gmail.com 2025-01-17T20:57:56+08:00 GitHub noreply@github.com 2025-01-17T14:57:56+02:00 llama.android: add field formatChat to control whether to parse special tokens when send message (#11270)
667d72846c06b2cf4f7c8a4265e210991a49706b a133566d34a1dd3693c504786963bf1b7b7d8c0e Radoslav Gerganov rgerganov@gmail.com 2025-01-17T10:57:09+02:00 GitHub noreply@github.com 2025-01-17T10:57:09+02:00 rpc : early register backend devices (#11262)
a133566d34a1dd3693c504786963bf1b7b7d8c0e 960ec65273a9554ff2510ba285a970289256ebfb Georgi Gerganov ggerganov@gmail.com 2025-01-17T09:28:00+02:00 GitHub noreply@github.com 2025-01-17T09:28:00+02:00 vocab : fix double-eos check (#11273)
960ec65273a9554ff2510ba285a970289256ebfb 7a689c415e2aecea1a5ae438542afeaf69815d52 David Renshaw dwrenshaw@gmail.com 2025-01-17T02:12:01-05:00 GitHub noreply@github.com 2025-01-17T08:12:01+01:00 llama : fix deprecation message: vocabable -> vocab (#11269)
7a689c415e2aecea1a5ae438542afeaf69815d52 bd38ddea0181bc717de7cae66fd4323975c85656 musoles 135031143+musoles@users.noreply.github.com 2025-01-17T00:10:49Z GitHub noreply@github.com 2025-01-17T01:10:49+01:00 README : added kalavai to infrastructure list (#11216)
bd38ddea0181bc717de7cae66fd4323975c85656 466300fe1416de2802b710215817db28d4496f41 Jeff Bolz jbolz@nvidia.com 2025-01-16T15:47:10-06:00 GitHub noreply@github.com 2025-01-16T22:47:10+01:00 vulkan: support copy from f32 to q4_0/q4_1/q5_0/q5_1/q8_0/iq4_nl (#11166)
466300fe1416de2802b710215817db28d4496f41 206bc53422521a67de5e2caba661e21d590d2bae Jeff Bolz jbolz@nvidia.com 2025-01-16T15:23:49-06:00 GitHub noreply@github.com 2025-01-16T22:23:49+01:00 vulkan: optimize coopmat2 q4_k/q5_k dequant functions. (#11206)
206bc53422521a67de5e2caba661e21d590d2bae 4dbc8b9cb71876e005724f4e8f73a3544646bcf5 Jeff Bolz jbolz@nvidia.com 2025-01-16T15:16:39-06:00 GitHub noreply@github.com 2025-01-16T22:16:39+01:00 vulkan: optimize coopmat2 q2_k dequant function (#11130)
4dbc8b9cb71876e005724f4e8f73a3544646bcf5 9c8dcefe171ba70ed14f2a64d1433da12d0dd1c1 RunningLeon mnsheng@yeah.net 2025-01-17T02:10:38+08:00 GitHub noreply@github.com 2025-01-16T20:10:38+02:00 llama : add internlm3 support (#11233)
9c8dcefe171ba70ed14f2a64d1433da12d0dd1c1 681149ced2582f48c24792403df1307fed5eb951 Johannes Gäßler johannesg@5d6.de 2025-01-16T16:43:38+01:00 GitHub noreply@github.com 2025-01-16T16:43:38+01:00 CUDA: backwards pass for misc. ops, add tests (#11257)
681149ced2582f48c24792403df1307fed5eb951 c67cc9837d48ea7f612b5666b90d189e63dfd7d3 Xuan Son Nguyen thichthat@gmail.com 2025-01-16T13:54:08+01:00 GitHub noreply@github.com 2025-01-16T13:54:08+01:00 llama : add `llama_model_load_from_splits` (#11255)
c67cc9837d48ea7f612b5666b90d189e63dfd7d3 adc5dd92e8aea98f5e7ac84f6e1bc15de35130b5 fj-y-saito 85871716+fj-y-saito@users.noreply.github.com 2025-01-16T18:11:49+09:00 GitHub noreply@github.com 2025-01-16T11:11:49+02:00 ggml: aarch64: implement SVE kernels for q4_K_q8_K vector dot (#11227)
adc5dd92e8aea98f5e7ac84f6e1bc15de35130b5 f11cfdfd7fe29436fce512d934c2ff6b94bd89d2 Eve 139727413+netrunnereve@users.noreply.github.com 2025-01-15T19:50:13Z GitHub noreply@github.com 2025-01-15T19:50:13Z vulkan: scale caching for k quants + misc fixes (#11081)
f11cfdfd7fe29436fce512d934c2ff6b94bd89d2 1d8504338ea27c7331998c11afdbd47ce5a9daac Georgi Gerganov ggerganov@gmail.com 2025-01-15T18:28:35+02:00 GitHub noreply@github.com 2025-01-15T18:28:35+02:00 ci : use -no-cnv in gguf-split tests (#11254)
1d8504338ea27c7331998c11afdbd47ce5a9daac 432df2d5f901a8ac93d1befad916144a0478bd9e Junil Kim logyourself@gmail.com 2025-01-15T22:17:42+09:00 GitHub noreply@github.com 2025-01-15T14:17:42+01:00 fix: ggml: fix vulkan-shaders-gen build (#10448)
432df2d5f901a8ac93d1befad916144a0478bd9e 0ccd7f3eb2debe477ffe3c44d5353cc388c9418d Johannes Gäßler johannesg@5d6.de 2025-01-15T12:51:37+01:00 GitHub noreply@github.com 2025-01-15T12:51:37+01:00 RoPE: fix back, CUDA support for back + noncont. (#11240)
0ccd7f3eb2debe477ffe3c44d5353cc388c9418d f446c2cf6a56a750b67c967505e717a996d2f2fd Daniel Bevenius daniel.bevenius@gmail.com 2025-01-15T05:44:38+01:00 GitHub noreply@github.com 2025-01-15T05:44:38+01:00 examples : add embd_to_audio to tts-outetts.py [no ci] (#11235)
f446c2cf6a56a750b67c967505e717a996d2f2fd b4d92a59a20eea400d8dd30844a339b76210daa0 Akarshan Biswas akarshan.biswas@gmail.com 2025-01-15T08:50:17+05:30 GitHub noreply@github.com 2025-01-15T11:20:17+08:00 SYCL: Add gated linear attention kernel (#11175)
b4d92a59a20eea400d8dd30844a339b76210daa0 bbf3e55e352d309573bdafee01a014b0a2492155 Xuan Son Nguyen thichthat@gmail.com 2025-01-14T15:42:23+01:00 GitHub noreply@github.com 2025-01-14T16:42:23+02:00 ci : add -no-cnv for tests (#11238)
bbf3e55e352d309573bdafee01a014b0a2492155 c5bf0d1bd7eb8762edca47e0a95f64e6fbfaf5b1 Georgi Gerganov ggerganov@gmail.com 2025-01-14T12:54:58+02:00 GitHub noreply@github.com 2025-01-14T11:54:58+01:00 vocab : add dummy tokens for "no_vocab" type (#11231)
c5bf0d1bd7eb8762edca47e0a95f64e6fbfaf5b1 091592d758cb55af7bfadd6c397f61db387aa8f3 ebraminio ebrahim@gnu.org 2025-01-14T14:09:33+03:30 GitHub noreply@github.com 2025-01-14T11:39:33+01:00 server : Improve code snippets direction between RTL text (#11221)
091592d758cb55af7bfadd6c397f61db387aa8f3 44d1e796d08641e7083fcbf37b33c79842a2f01e Olivier Chafik ochafik@users.noreply.github.com 2025-01-14T10:16:41Z GitHub noreply@github.com 2025-01-14T10:16:41Z Refactor test-chat-template.cpp (#11224)
44d1e796d08641e7083fcbf37b33c79842a2f01e a4f3f5d8e64b10fcf59913d487c1782dd0bc23e0 Georgi Gerganov ggerganov@gmail.com 2025-01-14T10:39:42+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-14T10:39:42+02:00 sync : ggml
a4f3f5d8e64b10fcf59913d487c1782dd0bc23e0 48e1ae0e61b04c6a458ac207e84e86f5ce0abccd Georgi Gerganov ggerganov@gmail.com 2025-01-14T09:40:15+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-14T09:40:52+02:00 scripts : sync gguf (cont)
48e1ae0e61b04c6a458ac207e84e86f5ce0abccd d00a80e89dc6632dde7391e1e1ecbbbac088dd25 Georgi Gerganov ggerganov@gmail.com 2025-01-14T09:36:58+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-14T09:36:58+02:00 scripts : sync gguf
d00a80e89dc6632dde7391e1e1ecbbbac088dd25 504af20ee4eae72080a56d59d744f6774f7901ce Georgi Gerganov ggerganov@gmail.com 2025-01-14T09:19:58+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-14T09:19:58+02:00 scripts : sync opencl
504af20ee4eae72080a56d59d744f6774f7901ce 84a44815f704aaed8e8edec7a39e846a975c7ba9 ebraminio ebrahim@gnu.org 2025-01-13T22:53:31+03:30 GitHub noreply@github.com 2025-01-13T20:23:31+01:00 server : (UI) Improve messages bubble shape in RTL (#11220)
84a44815f704aaed8e8edec7a39e846a975c7ba9 39509fb082895d1eae2486f8ad2cbf0e905346c4 Xuan Son Nguyen thichthat@gmail.com 2025-01-13T20:18:12+01:00 GitHub noreply@github.com 2025-01-13T20:18:12+01:00 cli : auto activate conversation mode if chat template is available (#11214)
39509fb082895d1eae2486f8ad2cbf0e905346c4 a29f0870d4846f52eda14ae28cea612ab66d903c Andreas Kieslinger 47689530+aendk@users.noreply.github.com 2025-01-13T16:45:53+01:00 GitHub noreply@github.com 2025-01-13T16:45:53+01:00 cuda : CUDA Graph Compute Function Refactor (precursor for performance improvements) (#11042)
a29f0870d4846f52eda14ae28cea612ab66d903c 437e05f714cdd67757405221578d3f95f8228b63 Georgi Gerganov ggerganov@gmail.com 2025-01-13T15:59:26+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-13T15:59:26+02:00 contrib : add naming guidelines (cont) (#11177)
437e05f714cdd67757405221578d3f95f8228b63 ca001f6656c1c3d29ef479b3aa5d669453e63be5 ebraminio ebrahim@gnu.org 2025-01-13T17:16:39+03:30 GitHub noreply@github.com 2025-01-13T14:46:39+01:00 server : (UI) Support for RTL text as models input or output (#11208)
ca001f6656c1c3d29ef479b3aa5d669453e63be5 00b4c3da6202e855087a4986bf19bb41b959e333 Georgi Gerganov ggerganov@gmail.com 2025-01-13T15:08:44+02:00 GitHub noreply@github.com 2025-01-13T15:08:44+02:00 contrib : add naming guidelines (cont) (#11177)
00b4c3da6202e855087a4986bf19bb41b959e333 7426a26b2492fc546a4db6991e871ee605714093 Xuan Son Nguyen thichthat@gmail.com 2025-01-13T13:56:23+01:00 GitHub noreply@github.com 2025-01-13T13:56:23+01:00 common : support tag-based --hf-repo like on ollama (#11195)
7426a26b2492fc546a4db6991e871ee605714093 8f70fc3d1b1d3c17b61842330dd106d391cc1227 Georgi Gerganov ggerganov@gmail.com 2025-01-13T14:46:36+02:00 GitHub noreply@github.com 2025-01-13T14:46:36+02:00 contrib : add naming guidelines (#11177)
8f70fc3d1b1d3c17b61842330dd106d391cc1227 1244cdcf14900dd199907b13f25d9c91a507f578 Daniel Bevenius daniel.bevenius@gmail.com 2025-01-13T13:38:20+01:00 GitHub noreply@github.com 2025-01-13T13:38:20+01:00 llama : remove 'd' from bad special token log (#11212)
1244cdcf14900dd199907b13f25d9c91a507f578 924518e2e5726e81f3aeb2518fb85963a500e93a Radoslav Gerganov rgerganov@gmail.com 2025-01-13T13:31:41+02:00 GitHub noreply@github.com 2025-01-13T13:31:41+02:00 ggml : do not define GGML_USE_CUDA when building with GGML_BACKEND_DL (#11211)
924518e2e5726e81f3aeb2518fb85963a500e93a 9a483999a6fda350772aaf7bc541f1cb246f8a29 Eric Curtin ecurtin@redhat.com 2025-01-12T18:23:10Z GitHub noreply@github.com 2025-01-12T18:23:10Z Reset color before we exit (#11205)
9a483999a6fda350772aaf7bc541f1cb246f8a29 08f10f69c38288e9e8bb1f933af63a3fc9013d40 Xuan Son Nguyen thichthat@gmail.com 2025-01-12T13:45:14+01:00 GitHub noreply@github.com 2025-01-12T13:45:14+01:00 llama : fix chat template gguf key (#11201)
08f10f69c38288e9e8bb1f933af63a3fc9013d40 afa8a9ec9b520137bbd1ca6838cda93ee39baf20 Georgi Gerganov ggerganov@gmail.com 2025-01-12T12:15:53+02:00 GitHub noreply@github.com 2025-01-12T12:15:53+02:00 llama : remove notion of CLS token (#11064)
afa8a9ec9b520137bbd1ca6838cda93ee39baf20 c05e8c9934f94fde49bc1bc9dc51eed282605150 Georgi Gerganov ggerganov@gmail.com 2025-01-12T11:32:42+02:00 GitHub noreply@github.com 2025-01-12T11:32:42+02:00 llama : add `llama_vocab`, functions -> methods, naming (#11110)
c05e8c9934f94fde49bc1bc9dc51eed282605150 2739a71e4b88474833b64aa974ca4515574fd3c4 Vinesh Janarthanan 36610342+VJHack@users.noreply.github.com 2025-01-11T03:42:31-06:00 GitHub noreply@github.com 2025-01-11T11:42:31+02:00 gguf-py: fixed local detection of gguf package (#11180)
2739a71e4b88474833b64aa974ca4515574fd3c4 ba8a1f9c5b675459c55a83e3f97f10df3a66c788 Daniel Bevenius daniel.bevenius@gmail.com 2025-01-11T05:50:33+01:00 GitHub noreply@github.com 2025-01-11T05:50:33+01:00 convert : sort print supported models [no ci] (#11179)
ba8a1f9c5b675459c55a83e3f97f10df3a66c788 ff3fcabc727b2dd0c477d23a258217b27cc639fb Daniel Bevenius daniel.bevenius@gmail.com 2025-01-10T13:16:16+01:00 GitHub noreply@github.com 2025-01-10T13:16:16+01:00 examples : add README.md to tts example [no ci] (#11155)
ff3fcabc727b2dd0c477d23a258217b27cc639fb c3f9d25706ac84297067aeaa662c1f1af42ed443 Daniel Bevenius daniel.bevenius@gmail.com 2025-01-10T11:30:53+01:00 GitHub noreply@github.com 2025-01-10T11:30:53+01:00 convert : add --print-supported-models option (#11172)
c3f9d25706ac84297067aeaa662c1f1af42ed443 ee7136c6d1e0ba7633294dad137b1573048031ec 0cc4m picard12@live.de 2025-01-10T06:39:33+01:00 GitHub noreply@github.com 2025-01-10T06:39:33+01:00 Vulkan: Fix float16 use on devices without float16 support + fix subgroup_size_control validation error (#11161)
ee7136c6d1e0ba7633294dad137b1573048031ec c6860cc7346c90219475e4467bb8a288e0df975c Molly Sophia mollysophia379@gmail.com 2025-01-10T09:58:08+08:00 GitHub noreply@github.com 2025-01-10T09:58:08+08:00 llama: add support for QRWKV6 model architecture (#11001)
c6860cc7346c90219475e4467bb8a288e0df975c 1204f9727005974587d6fc1dcd4d4f0ead87c856 Akarshan Biswas akarshan.biswas@gmail.com 2025-01-10T05:43:03+05:30 GitHub noreply@github.com 2025-01-10T08:13:03+08:00 SYCL: Refactor ggml_sycl_compute_forward (#11121)
1204f9727005974587d6fc1dcd4d4f0ead87c856 8eceb888d7b7f5e93d20a4f85ca6511022b87040 Tei Home taiteitonghome@proton.me 2025-01-09T19:32:06+08:00 GitHub noreply@github.com 2025-01-09T11:32:06Z doc: add cuda guide for fedora (#11135)
8eceb888d7b7f5e93d20a4f85ca6511022b87040 f8feb4b01af374ad2fce302fd5790529c615710b Daniel Bevenius daniel.bevenius@gmail.com 2025-01-09T11:28:29+01:00 GitHub noreply@github.com 2025-01-09T11:28:29+01:00 server : add tooltips to settings and themes btn (#11154)
f8feb4b01af374ad2fce302fd5790529c615710b be0e950c91cde2d8488ae32162b549d7023482f0 Pierrick Hymbert pierrick.hymbert@gmail.com 2025-01-09T11:21:41+01:00 GitHub noreply@github.com 2025-01-09T11:21:41+01:00 model: Add support for PhiMoE arch (#11003)
be0e950c91cde2d8488ae32162b549d7023482f0 d9feae1c06321aac9662fd4b4249452dccaec553 Georgi Gerganov ggerganov@gmail.com 2025-01-09T11:15:15+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-09T11:15:15+02:00 media : remove old img [no ci]
d9feae1c06321aac9662fd4b4249452dccaec553 8d59d911711b8f1ba9ec57c4b192ccd2628af033 Xuan Son Nguyen thichthat@gmail.com 2025-01-09T10:07:33+01:00 GitHub noreply@github.com 2025-01-09T10:07:33+01:00 llama-chat : add phi 4 template (#11148)
8d59d911711b8f1ba9ec57c4b192ccd2628af033 8a1d9c25fafbaf4182dd0b785dd6303ee40d55bc hydai z54981220@gmail.com 2025-01-09T04:03:28+08:00 GitHub noreply@github.com 2025-01-08T20:03:28Z fix: add missing msg in static_assert (#11143)
8a1d9c25fafbaf4182dd0b785dd6303ee40d55bc 1bf839b1e8b9d043306c65eddd9021fe4337733e Vinesh Janarthanan 36610342+VJHack@users.noreply.github.com 2025-01-08T12:54:58-06:00 GitHub noreply@github.com 2025-01-08T20:54:58+02:00 gguf-py : move scripts directory (#11116)
1bf839b1e8b9d043306c65eddd9021fe4337733e f7cd13301c2a88f97073fd119072b4cc92c08df1 Eric Curtin ecurtin@redhat.com 2025-01-08T18:47:05Z GitHub noreply@github.com 2025-01-08T18:47:05Z Enhance user input handling for llama-run (#11138)
f7cd13301c2a88f97073fd119072b4cc92c08df1 4d2b3d88041705b20c30b3219838aa435e7ffbde Xuan Son Nguyen thichthat@gmail.com 2025-01-08T16:09:20+01:00 GitHub noreply@github.com 2025-01-08T16:09:20+01:00 ci : use actions from ggml-org (#11140)
4d2b3d88041705b20c30b3219838aa435e7ffbde c07d437bbd417f42b122e767ad42b3298767dca0 Xuan Son Nguyen thichthat@gmail.com 2025-01-08T15:59:53+01:00 GitHub noreply@github.com 2025-01-08T15:59:53+01:00 lora : improve compat with `mergekit-extract-lora` (#11131)
c07d437bbd417f42b122e767ad42b3298767dca0 99a3755a3c518119d0156766122f7b4b796ea576 Georgi Gerganov ggerganov@gmail.com 2025-01-08T16:19:36+02:00 GitHub noreply@github.com 2025-01-08T16:19:36+02:00 llama : avoid hardcoded QK_K (#11061)
99a3755a3c518119d0156766122f7b4b796ea576 c792dcf4880461c2b5f3960584db241ac71a893a Georgi Gerganov ggerganov@gmail.com 2025-01-08T13:40:30+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-08T13:40:30+02:00 sync : ggml
c792dcf4880461c2b5f3960584db241ac71a893a 80ccf5d725571035b454659e3c1b4b2b07b65e71 Radoslav Gerganov rgerganov@gmail.com 2025-01-05T09:50:37+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-08T13:40:18+02:00 ggml : allow loading backend with env variable (ggml/1059)
80ccf5d725571035b454659e3c1b4b2b07b65e71 a3c1232c3f475f0a77b9cc5225516ac31c567a06 Xuan Son Nguyen thichthat@gmail.com 2025-01-08T12:07:20+01:00 GitHub noreply@github.com 2025-01-08T12:07:20+01:00 ci : pin dependency to specific version (#11137)
a3c1232c3f475f0a77b9cc5225516ac31c567a06 8cef75c743ba13ebbd6d380c531200c768a8b8aa Georgi Gerganov ggerganov@gmail.com 2025-01-08T12:55:36+02:00 GitHub noreply@github.com 2025-01-08T12:55:36+02:00 arg : option to exclude arguments from specific examples (#11136)
8cef75c743ba13ebbd6d380c531200c768a8b8aa 0d52a69e4bf0d6181beec7853307bdcdeec9905b amritahs-ibm amritahs@linux.vnet.ibm.com 2025-01-08T16:24:19+05:30 GitHub noreply@github.com 2025-01-08T12:54:19+02:00 llamafile : ppc64le MMA INT8 implementation (#10912)
0d52a69e4bf0d6181beec7853307bdcdeec9905b 02f04301417e7fb44fa1025bc1b0aef866e2ca89 Georgi Gerganov ggerganov@gmail.com 2025-01-08T11:29:34+02:00 GitHub noreply@github.com 2025-01-08T11:29:34+02:00 ci : fix cmake option (#11125)
02f04301417e7fb44fa1025bc1b0aef866e2ca89 bec2183f2c8d37cf1278c11d1adb9311e9eaa242 Mathieu Baudier mbaudier@argeo.org 2025-01-08T09:18:13+01:00 GitHub noreply@github.com 2025-01-08T09:18:13+01:00 Disable GL_KHR_cooperative_matrix Vulkan extension if not available. (#11117)
bec2183f2c8d37cf1278c11d1adb9311e9eaa242 53ff6b9b9fb25ed0ec0a213e05534fe7c3d0040f ag2s20150909 19373730+ag2s20150909@users.noreply.github.com 2025-01-08T16:17:29+08:00 GitHub noreply@github.com 2025-01-08T09:17:29+01:00 fix: Vulkan shader gen binary path when Cross-compiling (#11096)
53ff6b9b9fb25ed0ec0a213e05534fe7c3d0040f 017cc5f446863316d05522a87f25ec48713a9492 Johannes Gäßler johannesg@5d6.de 2025-01-07T18:01:58+01:00 GitHub noreply@github.com 2025-01-07T18:01:58+01:00 GGUF: C++ refactor, backend support, misc fixes (#11030)
017cc5f446863316d05522a87f25ec48713a9492 a3d50bc022bedd6c7754c24749a1fef4d2d60c7c Diego Devesa slarengh@gmail.com 2025-01-07T16:11:57+01:00 GitHub noreply@github.com 2025-01-07T16:11:57+01:00 ggml-backend : only offload from host buffers (fix) (#11124)
a3d50bc022bedd6c7754c24749a1fef4d2d60c7c a4dd490069a66ae56b42127048f06757fc4de4f7 Diego Devesa slarengh@gmail.com 2025-01-07T12:38:05+01:00 GitHub noreply@github.com 2025-01-07T12:38:05+01:00 ggml-backend : only offload from host buffers (#11120)
a4dd490069a66ae56b42127048f06757fc4de4f7 c0d6f790d07aa78be15584ec394ac20739ade93b Radoslav Gerganov rgerganov@gmail.com 2025-01-07T08:37:02+02:00 GitHub noreply@github.com 2025-01-07T08:37:02+02:00 rpc : code cleanup (#11107)
c0d6f790d07aa78be15584ec394ac20739ade93b dc7cef9f373f2a24b851f0df7a618c5209e593fa Akarshan Biswas akarshan.biswas@gmail.com 2025-01-07T11:56:07+05:30 GitHub noreply@github.com 2025-01-07T14:26:07+08:00 SYCL: Use get_multi_ptr instead of deprecated get_pointer in wkv6 (#11087)
dc7cef9f373f2a24b851f0df7a618c5209e593fa ecebbd292d741ac084cf248146b2cfb17002aa1d Eric Curtin ecurtin@redhat.com 2025-01-06T22:45:28Z GitHub noreply@github.com 2025-01-06T23:45:28+01:00 llama-run : fix context size (#11094)
ecebbd292d741ac084cf248146b2cfb17002aa1d 96be8c32649378a23031630a48c440f3a5d0839b Georgi Gerganov ggerganov@gmail.com 2025-01-06T17:52:35+02:00 GitHub noreply@github.com 2025-01-06T17:52:35+02:00 llama : remove unused headers (#11109)
96be8c32649378a23031630a48c440f3a5d0839b e6e7c75d94adf4d39e846d30807c531ff22865e7 Xuan Son Nguyen thichthat@gmail.com 2025-01-06T16:34:49+01:00 GitHub noreply@github.com 2025-01-06T16:34:49+01:00 github : add cmd line field to bug report (#11090)
e6e7c75d94adf4d39e846d30807c531ff22865e7 09186fabbe05236f2b9446ba6c643cb737540d10 Georgi Gerganov ggerganov@gmail.com 2025-01-06T15:36:08+02:00 GitHub noreply@github.com 2025-01-06T15:36:08+02:00 server : fix extra BOS in infill endpoint (#11106)
09186fabbe05236f2b9446ba6c643cb737540d10 96a1dc27c3f09bf1ed83a26292d571795bcf27fa Xuan Son Nguyen thichthat@gmail.com 2025-01-06T13:41:12+01:00 GitHub noreply@github.com 2025-01-06T13:41:12+01:00 llama : remove check flash_attn with lora (#11104)
96a1dc27c3f09bf1ed83a26292d571795bcf27fa 6369f867a410416239d9f20ec27c2b1d6a9fee52 Asghar Ghorbani a-ghorbani@users.noreply.github.com 2025-01-06T12:21:46+01:00 GitHub noreply@github.com 2025-01-06T13:21:46+02:00 llama : prevent system info string accumulation across calls (#11101)
6369f867a410416239d9f20ec27c2b1d6a9fee52 47182dd03fe04a4ffda5d7f4c8a109ae0056cf56 Daniel Bevenius daniel.bevenius@gmail.com 2025-01-06T10:28:17+01:00 GitHub noreply@github.com 2025-01-06T11:28:17+02:00 llama : rename missed batch params/vars to ubatch (#10059)
47182dd03fe04a4ffda5d7f4c8a109ae0056cf56 3e6e7a6bc2c4b980a0cf0fcb5cb3b79a965b5f14 Georgi Gerganov ggerganov@gmail.com 2025-01-06T10:55:18+02:00 GitHub noreply@github.com 2025-01-06T10:55:18+02:00 llama : update llama_model API names (#11063)
3e6e7a6bc2c4b980a0cf0fcb5cb3b79a965b5f14 ae2f606bb598b287f5fb69c9fdfc98b86598c6cc Georgi Gerganov ggerganov@gmail.com 2025-01-06T10:54:25+02:00 GitHub noreply@github.com 2025-01-06T10:54:25+02:00 tokenize : escape the prompt (#11058)
ae2f606bb598b287f5fb69c9fdfc98b86598c6cc 727368c60f2ebf2d6a7473a4a9f80957ab063a8e Georgi Gerganov ggerganov@gmail.com 2025-01-06T10:52:38+02:00 GitHub noreply@github.com 2025-01-06T10:52:38+02:00 mmap : fix fileno macro clash (#11076)
727368c60f2ebf2d6a7473a4a9f80957ab063a8e 5047dd3546951dea3d65c02257d06c46c8662338 Georgi Gerganov ggerganov@gmail.com 2025-01-06T10:52:15+02:00 GitHub noreply@github.com 2025-01-06T10:52:15+02:00 llama : use LLAMA_TOKEN_NULL (#11062)
5047dd3546951dea3d65c02257d06c46c8662338 46e3556e01b824e52395fb050b29804b6cff2a7c Georgi Gerganov ggerganov@gmail.com 2025-01-06T10:52:01+02:00 GitHub noreply@github.com 2025-01-06T10:52:01+02:00 llama : use _impl suffix instead of _internal (#11060)
46e3556e01b824e52395fb050b29804b6cff2a7c b56f079e28fda692f11a8b59200ceb815b05d419 Johannes Gäßler johannesg@5d6.de 2025-01-06T02:33:52+01:00 GitHub noreply@github.com 2025-01-06T02:33:52+01:00 CUDA: add BF16 support (#11093)
b56f079e28fda692f11a8b59200ceb815b05d419 9394bbd484f802ce80d2858033583af3ef700d25 0cc4m picard12@live.de 2025-01-04T21:09:59+01:00 GitHub noreply@github.com 2025-01-04T21:09:59+01:00 Vulkan: Add device-specific blacklist for coopmat for the AMD proprietary driver (#11074)
9394bbd484f802ce80d2858033583af3ef700d25 f922a9c542ee117550a168395c63ea79261f5c99 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2025-01-04T21:06:11+01:00 GitHub noreply@github.com 2025-01-04T21:06:11+01:00 llama : Add support for DeepSeek V3 (#11049)
f922a9c542ee117550a168395c63ea79261f5c99 46be942214e295cd34660bbbd6b846155d1c36a0 matt23654 matthew.webber@protonmail.com 2025-01-04T16:10:30Z GitHub noreply@github.com 2025-01-04T17:10:30+01:00 [GGML][RPC] Support for models with non-512-aligned tensors over RPC. (#11047)
46be942214e295cd34660bbbd6b846155d1c36a0 78c678517530d411b4263341cdb4dc28c9d117c8 DAN™ dranger003@gmail.com 2025-01-04T09:33:31-05:00 GitHub noreply@github.com 2025-01-04T16:33:31+02:00 llama : add support for the cohere2 model architecture (#10900)
78c678517530d411b4263341cdb4dc28c9d117c8 5e3b08d606b5b0caaea16541b504c3bba8f3ec1d Georgi Gerganov ggerganov@gmail.com 2025-01-04T10:54:01+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-04T16:09:53+02:00 sync : ggml
5e3b08d606b5b0caaea16541b504c3bba8f3ec1d db68c93b57bfdf6da1fbdae81080382d6998cbc9 Georgi Gerganov ggerganov@gmail.com 2025-01-04T10:53:54+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-04T16:09:53+02:00 ggml : do not install metal source when embed library (ggml/1054)
db68c93b57bfdf6da1fbdae81080382d6998cbc9 c31fc8b966817b2f0b277fd28e04a189e388972a Daniel Bevenius daniel.bevenius@gmail.com 2024-12-19T03:50:12+01:00 Georgi Gerganov ggerganov@gmail.com 2025-01-04T16:09:53+02:00 ggml : improve inputs log sched_print_assignments (ggml/1053)
c31fc8b966817b2f0b277fd28e04a189e388972a 4b0c638b9a68f577cb2066b638c9f622d91ee661 Gilad S. 7817232+giladgd@users.noreply.github.com 2025-01-04T10:17:31+02:00 GitHub noreply@github.com 2025-01-04T09:17:31+01:00 fix: Vulkan shader gen binary path (#11037)
4b0c638b9a68f577cb2066b638c9f622d91ee661 e7da954eccdf39ee795a6135bdb86f0978902681 Molly Sophia mollysophia379@gmail.com 2025-01-03T20:13:18+08:00 GitHub noreply@github.com 2025-01-03T14:13:18+02:00 common : disable KV cache shifting automatically for unsupported models (#11053)
e7da954eccdf39ee795a6135bdb86f0978902681 f66f5829276650cd83a087ab2cfed1a760183ea1 Georgi Gerganov ggerganov@gmail.com 2025-01-03T11:26:14+02:00 Georgi Gerganov ggerganov@gmail.com 2025-01-03T11:26:14+02:00 metal : avoid uint (#11019)
f66f5829276650cd83a087ab2cfed1a760183ea1 2f0ee84b9b02d2a98742308026f060ebdc2423f1 Georgi Gerganov ggerganov@gmail.com 2025-01-03T10:18:53+02:00 GitHub noreply@github.com 2025-01-03T10:18:53+02:00 llama : refactor `src/llama.cpp` (#10902)
2f0ee84b9b02d2a98742308026f060ebdc2423f1 0da5d860266c6928b8c9408efbd264ae59fedda6 Pierrick Hymbert pierrick.hymbert@gmail.com 2025-01-02T18:06:12+01:00 GitHub noreply@github.com 2025-01-02T18:06:12+01:00 server: bench: minor fixes (#10765)
0da5d860266c6928b8c9408efbd264ae59fedda6 a45433ba209ee0b33d02c7dc4c31f29894ad83a6 Xuan Son Nguyen thichthat@gmail.com 2025-01-02T15:05:18+01:00 GitHub noreply@github.com 2025-01-02T15:05:18+01:00 server : allow using LoRA adapters per-request (#10994)
a45433ba209ee0b33d02c7dc4c31f29894ad83a6 0827b2c1da299805288abbd556d869318f2b121e Benson Wong mostlygeek@gmail.com 2025-01-01T23:14:54-08:00 GitHub noreply@github.com 2025-01-02T09:14:54+02:00 readme : add llama-swap to infrastructure section (#11032)
0827b2c1da299805288abbd556d869318f2b121e 45095a61bfd164e87563a0dc0fbd7b0e9891590b Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-12-31T19:53:33+05:30 GitHub noreply@github.com 2024-12-31T15:23:33+01:00 ggml : fixes for AVXVNNI instruction set with MSVC and Clang (#11027)
45095a61bfd164e87563a0dc0fbd7b0e9891590b 5896c65232c7dc87d78426956b16f63fbf58dcf6 Xuan Son Nguyen thichthat@gmail.com 2024-12-31T15:22:01+01:00 GitHub noreply@github.com 2024-12-31T15:22:01+01:00 server : clean up built-in template detection (#11026)
5896c65232c7dc87d78426956b16f63fbf58dcf6 bc7b1f86324279a3dabb705c04ad754a2b27df16 Xuan Son Nguyen thichthat@gmail.com 2024-12-31T12:34:13+01:00 GitHub noreply@github.com 2024-12-31T12:34:13+01:00 server : add OAI compat for /v1/completions (#10974)
bc7b1f86324279a3dabb705c04ad754a2b27df16 6e1531aca5ed17f078973b4700fcdadbda4a34a5 ymcki 84055651+ymcki@users.noreply.github.com 2024-12-31T19:04:48+08:00 GitHub noreply@github.com 2024-12-31T13:04:48+02:00 convert : fix Llama-3_1-Nemotron-51B rope settings (#11008)
6e1531aca5ed17f078973b4700fcdadbda4a34a5 716bd6dec3e044e5c325386b5b0483392b24cefe Peter peter277@users.noreply.github.com 2024-12-31T11:46:06+11:00 GitHub noreply@github.com 2024-12-31T01:46:06+01:00 common, examples, ggml : fix MSYS2 GCC compiler errors and warnings when building with LLAMA_CURL=ON and GGML_OPENCL=ON (#11013)
716bd6dec3e044e5c325386b5b0483392b24cefe c250ecb3157f3bae0a45f44c3c953b5414d4c2f7 Jeff Bolz jbolz@nvidia.com 2024-12-30T11:27:11-06:00 GitHub noreply@github.com 2024-12-30T18:27:11+01:00 vulkan: optimize mul_mat for small values of N (#10991)
c250ecb3157f3bae0a45f44c3c953b5414d4c2f7 a813badbbdf0d38705f249df7a0c99af5cdee678 ag2s20150909 19373730+ag2s20150909@users.noreply.github.com 2024-12-30T20:35:13+08:00 GitHub noreply@github.com 2024-12-30T14:35:13+02:00 android : fix llama_batch free (#11014)
a813badbbdf0d38705f249df7a0c99af5cdee678 fdd21889123bec62b1db3b2fc22b5a4abab32174 Jeff Bolz jbolz@nvidia.com 2024-12-29T03:16:34-06:00 GitHub noreply@github.com 2024-12-29T10:16:34+01:00 vulkan: im2col and matmul optimizations for stable diffusion (#10942)
fdd21889123bec62b1db3b2fc22b5a4abab32174 f865ea149d71ef883e3780fced8a20a1464eccf4 Jeff Bolz jbolz@nvidia.com 2024-12-29T02:35:11-06:00 GitHub noreply@github.com 2024-12-29T09:35:11+01:00 vulkan: Use push constant offset to handle misaligned descriptors (#10987)
f865ea149d71ef883e3780fced8a20a1464eccf4 16cdce7b68218959e0658e2f95b4572573d5008e Isaac McFadyen isaac@imcf.me 2024-12-28T10:09:19-05:00 GitHub noreply@github.com 2024-12-28T16:09:19+01:00 server: added more docs for response_fields field (#10995)
16cdce7b68218959e0658e2f95b4572573d5008e d79d8f39b4da6deca4aea8bf130c6034c482b320 Alexey Parfenov zxed@alkatrazstudio.net 2024-12-28T15:08:54Z GitHub noreply@github.com 2024-12-28T16:08:54+01:00 server : fix token duplication when streaming with stop strings (#10997)
d79d8f39b4da6deca4aea8bf130c6034c482b320 d283d02bf254a7f2991e1502066330cc0d4321a6 Eve 139727413+netrunnereve@users.noreply.github.com 2024-12-26T10:54:44-05:00 GitHub noreply@github.com 2024-12-26T16:54:44+01:00 vulkan: multi-row k quants (#10846)
d283d02bf254a7f2991e1502066330cc0d4321a6 9ba399dfa7f115effc63d48e6860a94c9faa31b2 Peter peter277@users.noreply.github.com 2024-12-27T00:59:11+11:00 GitHub noreply@github.com 2024-12-26T14:59:11+01:00 examples, ggml : fix GCC compiler warnings (#10983)
9ba399dfa7f115effc63d48e6860a94c9faa31b2 2cd43f4900ba0e34124fdcbf02a7f9df25a10a3d Reza Kakhki rezakakhki.de@gmail.com 2024-12-24T21:33:04+01:00 GitHub noreply@github.com 2024-12-24T21:33:04+01:00 server : add support for "encoding_format": "base64" to the */embeddings endpoints (#10967)
2cd43f4900ba0e34124fdcbf02a7f9df25a10a3d 09fe2e76137dde850b13313f720e7ffa17efdefa Djip007 3705339+Djip007@users.noreply.github.com 2024-12-24T18:54:49+01:00 GitHub noreply@github.com 2024-12-24T18:54:49+01:00 ggml : more perfo with llamafile tinyblas on x86_64 (#10714)
09fe2e76137dde850b13313f720e7ffa17efdefa 30caac3a68a54de8396b21e20ba972554c587230 NeverLucky 92274250+nvrxq@users.noreply.github.com 2024-12-24T19:39:49+03:00 GitHub noreply@github.com 2024-12-24T17:39:49+01:00 server: allow filtering llama server response fields (#10940)
30caac3a68a54de8396b21e20ba972554c587230 60cfa728e27c28537657d4e627ed432508eb9537 Georgi Gerganov ggerganov@gmail.com 2024-12-24T09:44:20+02:00 GitHub noreply@github.com 2024-12-24T09:44:20+02:00 llama : the WPM vocabs use the CLS token as BOS (#10930)
60cfa728e27c28537657d4e627ed432508eb9537 3327bb0f8dea381118f8e66c18ea14db56d3b942 Diego Devesa slarengh@gmail.com 2024-12-24T04:05:27+01:00 GitHub noreply@github.com 2024-12-24T04:05:27+01:00 ggml : use wstring for backend search paths (#10960)
3327bb0f8dea381118f8e66c18ea14db56d3b942 32d6ee6385b3fc908b283f509b845f757a6e7206 Diego Devesa slarengh@gmail.com 2024-12-24T04:05:17+01:00 GitHub noreply@github.com 2024-12-24T04:05:17+01:00 ggml : fix arm enabled features check (#10961)
32d6ee6385b3fc908b283f509b845f757a6e7206 14b699ecde8f1e9e251ebff9eca39ebc5603b83b Diego Devesa slarengh@gmail.com 2024-12-23T20:25:52+01:00 GitHub noreply@github.com 2024-12-23T20:25:52+01:00 ggml : fix const usage in SSE path (#10962)
14b699ecde8f1e9e251ebff9eca39ebc5603b83b 485dc01214f266afff7004bc702498b491abc404 Xuan Son Nguyen thichthat@gmail.com 2024-12-23T12:52:25+01:00 GitHub noreply@github.com 2024-12-23T12:52:25+01:00 server : fix missing model id in /model endpoint (#10957)
485dc01214f266afff7004bc702498b491abc404 86bf31cfe684849157f0875b4f0ebccac7034547 Xuan Son Nguyen thichthat@gmail.com 2024-12-23T12:02:44+01:00 GitHub noreply@github.com 2024-12-23T12:02:44+01:00 server : add system_fingerprint to chat/completion (#10917)
86bf31cfe684849157f0875b4f0ebccac7034547 b92a14a841fb4dfaf27b29d982ec8ba5289a3bff Radoslav Gerganov rgerganov@gmail.com 2024-12-23T10:39:30+02:00 GitHub noreply@github.com 2024-12-23T10:39:30+02:00 rpc-server : add support for the SYCL backend (#10934)
b92a14a841fb4dfaf27b29d982ec8ba5289a3bff 6f0c9e034bb398915a6617ee4acc62adb87d387d Yun Dou dixyes@gmail.com 2024-12-23T08:35:44+08:00 GitHub noreply@github.com 2024-12-23T01:35:44+01:00 llama : support InfiniAI Megrez 3b (#10893)
6f0c9e034bb398915a6617ee4acc62adb87d387d dab76c92cc63072d9495ba87f2f3f3a4872d4f57 ymcki 84055651+ymcki@users.noreply.github.com 2024-12-23T08:22:33+08:00 GitHub noreply@github.com 2024-12-23T01:22:33+01:00 llama : support for Llama-3_1-Nemotron-51B (#10669)
dab76c92cc63072d9495ba87f2f3f3a4872d4f57 7024d59e6a572730626cb11896829d115043a1b1 Eric Curtin ecurtin@redhat.com 2024-12-23T00:21:40Z GitHub noreply@github.com 2024-12-23T01:21:40+01:00 llama-run : include temperature option (#10899)
7024d59e6a572730626cb11896829d115043a1b1 7c0e28585843b366864b43b48f92425e2ea17df6 yuri@FreeBSD yurivict@users.noreply.github.com 2024-12-22T16:20:11-08:00 GitHub noreply@github.com 2024-12-23T01:20:11+01:00 ggml : fix run-time on FreeBSD in get_executable_path() (#10948)
7c0e28585843b366864b43b48f92425e2ea17df6 7ae33a616f44ecc081f3dcb589be20962d1d4a92 Rudi Servo rudiservo@gmail.com 2024-12-22T21:22:58-01:00 GitHub noreply@github.com 2024-12-22T23:22:58+01:00 devops : add docker-multi-stage builds (#10832)
7ae33a616f44ecc081f3dcb589be20962d1d4a92 ebdee9478ca7ba65497b9b96f7457698c6ee5115 Billel Mokeddem billel.mokeddem.ml@gmail.com 2024-12-23T01:09:58+03:00 GitHub noreply@github.com 2024-12-23T00:09:58+02:00 llama : add Falcon3 support (#10883)
ebdee9478ca7ba65497b9b96f7457698c6ee5115 5cd85b5e008de2ec398d6596e240187d627561e3 Jeff Bolz jbolz@nvidia.com 2024-12-22T03:44:01-06:00 GitHub noreply@github.com 2024-12-22T10:44:01+01:00 vulkan: build fixes for 32b (#10927)
5cd85b5e008de2ec398d6596e240187d627561e3 a91a41364b25705dbb81ae996bc35c3440c63b35 Georgi Gerganov ggerganov@gmail.com 2024-12-21T10:10:18+02:00 GitHub noreply@github.com 2024-12-21T10:10:18+02:00 convert : add BertForMaskedLM (#10919)
a91a41364b25705dbb81ae996bc35c3440c63b35 e34c5af43f941f0ddb92466776339897295aca11 Jeff Bolz jbolz@nvidia.com 2024-12-21T01:04:45-06:00 GitHub noreply@github.com 2024-12-21T08:04:45+01:00 vulkan: optimize coopmat2 dequant functions (#10855)
e34c5af43f941f0ddb92466776339897295aca11 eb5c3dc64bd967f2e23c87d9dec195f45468de60 Adrien Gallouët angt@huggingface.co 2024-12-21T00:33:37+01:00 GitHub noreply@github.com 2024-12-21T00:33:37+01:00 ggml-cpu: replace NEON asm with intrinsics in ggml_gemv_q4_0_4x8_q8_0() (#10874)
eb5c3dc64bd967f2e23c87d9dec195f45468de60 0ca416c91aea4549d9c77e3efeca403e15aa6c75 Akarshan Biswas akarshan.biswas@gmail.com 2024-12-20T21:01:28+05:30 GitHub noreply@github.com 2024-12-20T23:31:28+08:00 SYCL: Migrate away from deprecated ggml_tensor->backend (#10840)
0ca416c91aea4549d9c77e3efeca403e15aa6c75 21ae3b9be83820565d1a720999b7f63ce95b4920 Xuan Son Nguyen thichthat@gmail.com 2024-12-20T14:12:06+01:00 GitHub noreply@github.com 2024-12-20T14:12:06+01:00 server : (UI) fix copy to clipboard function (#10916)
21ae3b9be83820565d1a720999b7f63ce95b4920 0a11f8b7b5c39fdf6e91ef9674bc68ff08681af7 Diego Devesa slarengh@gmail.com 2024-12-20T13:31:28+01:00 GitHub noreply@github.com 2024-12-20T13:31:28+01:00 ggml : add test for SVE and disable when it fails (#10906)
0a11f8b7b5c39fdf6e91ef9674bc68ff08681af7 d408bb9268a988c5a60a5746d3a6430386e7604d Molly Sophia mollysophia379@gmail.com 2024-12-20T17:44:58+08:00 GitHub noreply@github.com 2024-12-20T11:44:58+02:00 convert : fix RWKV v6 model conversion (#10913)
d408bb9268a988c5a60a5746d3a6430386e7604d 5cab3e4aaa892df4620b720f20a503a1bbbe7a52 Georgi Gerganov ggerganov@gmail.com 2024-12-19T18:47:15+02:00 GitHub noreply@github.com 2024-12-19T18:47:15+02:00 clip : disable GPU support (#10896)
5cab3e4aaa892df4620b720f20a503a1bbbe7a52 36319dec5d75a7dfe3e3de37b9ca2a76cd52b7b2 Georgi Gerganov ggerganov@gmail.com 2024-12-19T17:42:13+02:00 GitHub noreply@github.com 2024-12-19T17:42:13+02:00 llama : minor grammar refactor (#10897)
36319dec5d75a7dfe3e3de37b9ca2a76cd52b7b2 57bb2c40cd94c5a09f5210ed8264cc93b21c4b7e Georgi Gerganov ggerganov@gmail.com 2024-12-19T17:35:15+02:00 GitHub noreply@github.com 2024-12-19T17:35:15+02:00 tts : small QoL for easy model fetch (#10903)
57bb2c40cd94c5a09f5210ed8264cc93b21c4b7e a3c33b1dce2d4f25040b75f66629104bd1e40128 Xuan Son Nguyen thichthat@gmail.com 2024-12-19T15:40:08+01:00 GitHub noreply@github.com 2024-12-19T15:40:08+01:00 server : fix logprobs, make it OAI-compatible (#10783)
a3c33b1dce2d4f25040b75f66629104bd1e40128 2fffc52b50992ac5bc64db19d33c39cbc06f52cf Adrien Gallouët angt@huggingface.co 2024-12-19T14:20:41+01:00 GitHub noreply@github.com 2024-12-19T14:20:41+01:00 ggml: fix arm build with gcc (#10895)
2fffc52b50992ac5bc64db19d33c39cbc06f52cf 7585edbdebd02861e0994dae67c9338731fb3fc5 Sukriti Sharma Ssukriti@users.noreply.github.com 2024-12-19T06:04:51-07:00 GitHub noreply@github.com 2024-12-19T15:04:51+02:00 llama : fix Roberta embeddings (#10856)
7585edbdebd02861e0994dae67c9338731fb3fc5 cd920d0ac38ec243605a5a57c50941140a193f9e fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-12-19T10:37:12+01:00 GitHub noreply@github.com 2024-12-19T10:37:12+01:00 convert : Add support for Microsoft Phi-4 model (#10817)
cd920d0ac38ec243605a5a57c50941140a193f9e 7909e8588ddf70820adf1f325490eb3f67b32875 Johannes Gäßler johannesg@5d6.de 2024-12-19T08:53:58+01:00 GitHub noreply@github.com 2024-12-19T08:53:58+01:00 tests: disable GGUF test for bad value size (#10886)
7909e8588ddf70820adf1f325490eb3f67b32875 9177484f589d770ffc4e655b9819124d6a22c1d9 Eric Curtin ecurtin@redhat.com 2024-12-19T02:58:00Z GitHub noreply@github.com 2024-12-19T03:58:00+01:00 llama-run : improve progress bar (#10821)
9177484f589d770ffc4e655b9819124d6a22c1d9 0bf2d10c5514ff61b99897a4a5054f846e384e1e Diego Devesa slarengh@gmail.com 2024-12-18T23:21:42+01:00 GitHub noreply@github.com 2024-12-18T23:21:42+01:00 ggml : fix arm build (#10890)
0bf2d10c5514ff61b99897a4a5054f846e384e1e 7bbb5acf125d1d2840cac7d31b9aaa72210dd5ec Georgi Gerganov ggerganov@gmail.com 2024-12-18T19:27:21+02:00 GitHub noreply@github.com 2024-12-18T19:27:21+02:00 tts : add OuteTTS support (#10784)
7bbb5acf125d1d2840cac7d31b9aaa72210dd5ec 152610eda91217ac409342cd976d05f5114ad39f Gaetan Bisson gaetan@fenua.org 2024-12-18T04:00:07-10:00 GitHub noreply@github.com 2024-12-18T15:00:07+01:00 server: avoid overwriting Authorization header (#10878)
152610eda91217ac409342cd976d05f5114ad39f 0e70ba686e6c717a0aa41d88284e2a392c2bd0cd Georgi Gerganov ggerganov@gmail.com 2024-12-18T13:01:41+02:00 GitHub noreply@github.com 2024-12-18T13:01:41+02:00 server : output embeddings for all tokens when pooling = none (#10861)
0e70ba686e6c717a0aa41d88284e2a392c2bd0cd 46828872c31b25df16169cbbf5c2225fa9cb0675 Georgi Gerganov ggerganov@gmail.com 2024-12-18T11:05:29+02:00 GitHub noreply@github.com 2024-12-18T11:05:29+02:00 server : add "tokens" output (#10853)
46828872c31b25df16169cbbf5c2225fa9cb0675 6b064c92b4c0228e333193c167f2c98d2ec8d9bc Xuan Son Nguyen thichthat@gmail.com 2024-12-18T09:55:09+01:00 GitHub noreply@github.com 2024-12-18T10:55:09+02:00 server : (embeddings) using same format for "input" and "content" (#10872)
6b064c92b4c0228e333193c167f2c98d2ec8d9bc 4da69d1abd15263061aff94c10f205836a96a4bc redbeard bharrington@alticon.net 2024-12-18T00:35:00-08:00 GitHub noreply@github.com 2024-12-18T10:35:00+02:00 docs: Fix HIP (née hipBLAS) in README (#10880)
4da69d1abd15263061aff94c10f205836a96a4bc d62b532c52e0118323277eaa5f442e11ce6505ed Diego Devesa slarengh@gmail.com 2024-12-18T01:36:46+01:00 GitHub noreply@github.com 2024-12-18T01:36:46+01:00 Revert "llama : add Falcon3 support (#10864)" (#10876)
d62b532c52e0118323277eaa5f442e11ce6505ed 081b29bd2a3d91e7772e3910ce223dd63b8d7d26 DAN™ dranger003@gmail.com 2024-12-17T17:24:22-05:00 GitHub noreply@github.com 2024-12-17T23:24:22+01:00 Use model->gguf_kv for loading the template instead of using the C API. (#10868)
081b29bd2a3d91e7772e3910ce223dd63b8d7d26 5437d4aaf5132c879acda0bb67f2f8f71da4c9fe Johannes Gäßler johannesg@5d6.de 2024-12-17T19:09:35+01:00 GitHub noreply@github.com 2024-12-17T19:09:35+01:00 tests: add tests for GGUF (#10830)
5437d4aaf5132c879acda0bb67f2f8f71da4c9fe 78f766768d94e9c8b30ce10ca76f568e710d84f7 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:36:02+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:36:02+02:00 sync : ggml
78f766768d94e9c8b30ce10ca76f568e710d84f7 8dd19a48129d578972ea3d98896edbbf492891a9 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:34:32+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:35:49+02:00 cmake : fix "amd64" processor string (whisper/2638)
8dd19a48129d578972ea3d98896edbbf492891a9 130d0c90bd26b25e3a713b17a61a5d4eb0a66405 gn64 yukikaze.jp@gmail.com 2024-12-16T19:34:38+09:00 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:35:49+02:00 vulkan : fix soft_max.comp division by zero (whisper/2633)
130d0c90bd26b25e3a713b17a61a5d4eb0a66405 3919da8e335dbfd0741d239932a9b9e72061bf27 Daniel Bevenius daniel.bevenius@gmail.com 2024-12-14T03:23:08+01:00 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:35:49+02:00 ggml : remove return from ggml_gallocr_allocate_node (ggml/1048)
3919da8e335dbfd0741d239932a9b9e72061bf27 0006f5a74a59945f22ff966e723c3d566b3ca8d0 Daniel Bevenius daniel.bevenius@gmail.com 2024-12-13T08:19:38+01:00 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:35:48+02:00 ggml : add check for grad_accs (ggml/1046)
0006f5a74a59945f22ff966e723c3d566b3ca8d0 05c3a444b8b017b01b366b725ba22c740aae6b38 Georgi Gerganov ggerganov@gmail.com 2024-12-17T18:35:42+02:00 GitHub noreply@github.com 2024-12-17T18:35:42+02:00 ggml : update ggml_backend_cpu_device_supports_op (#10867)
05c3a444b8b017b01b366b725ba22c740aae6b38 382bc7f2e8ffd0b89f23e840d097e21f301197ba krystiancha krystian@krystianch.com 2024-12-17T16:00:24Z GitHub noreply@github.com 2024-12-17T18:00:24+02:00 server : fill usage info in embeddings and rerank responses (#10852)
382bc7f2e8ffd0b89f23e840d097e21f301197ba 4f51968aca049080dc77e26603aa0681ea77fe45 Billel Mokeddem billel.mokeddem.ml@gmail.com 2024-12-17T19:24:56+04:00 GitHub noreply@github.com 2024-12-17T17:24:56+02:00 llama : add Falcon3 support (#10864)
4f51968aca049080dc77e26603aa0681ea77fe45 227d7c5a7f4cc7734fde8a4ef4382d613486d3c8 Ruan 47767371+ruanych@users.noreply.github.com 2024-12-17T17:47:20+08:00 GitHub noreply@github.com 2024-12-17T11:47:20+02:00 readme : update typos (#10863)
227d7c5a7f4cc7734fde8a4ef4382d613486d3c8 7b1ec53f56bb72c49e4c8434157895f94d3709c2 Xuan Son Nguyen thichthat@gmail.com 2024-12-17T09:52:09+01:00 GitHub noreply@github.com 2024-12-17T09:52:09+01:00 server : (UI) fix missing async generator on safari (#10857)
7b1ec53f56bb72c49e4c8434157895f94d3709c2 160bc039c862c10b9938269a90ddb09d794a7b0d Eve 139727413+netrunnereve@users.noreply.github.com 2024-12-17T05:52:55Z GitHub noreply@github.com 2024-12-17T06:52:55+01:00 vulkan: bugfixes for small subgroup size systems + llvmpipe test (#10809)
160bc039c862c10b9938269a90ddb09d794a7b0d 08ea539df211e46bb4d0dd275e541cb591d5ebc8 Zhiyuan Li uniartisan2017@gmail.com 2024-12-17T05:00:46+08:00 GitHub noreply@github.com 2024-12-16T22:00:46+01:00 rwkv6: add wkv6 support for Vulkan backend (#10829)
08ea539df211e46bb4d0dd275e541cb591d5ebc8 644fd71b44c4cdbfc6482fbf0353d289c3bc29e6 Georgi Gerganov ggerganov@gmail.com 2024-12-16T12:31:45+02:00 GitHub noreply@github.com 2024-12-16T12:31:45+02:00 unicode : improve naming style (#10838)
644fd71b44c4cdbfc6482fbf0353d289c3bc29e6 4ddd199f6f6b980e0a7ed9f9b44efeae2fbdf5c4 Georgi Gerganov ggerganov@gmail.com 2024-12-16T12:31:14+02:00 GitHub noreply@github.com 2024-12-16T12:31:14+02:00 sampling : refactor + optimize penalties sampler (#10803)
4ddd199f6f6b980e0a7ed9f9b44efeae2fbdf5c4 a0974156f334acf8af5858d7ede5ab7d7490d415 Bartowski ckealty1182@gmail.com 2024-12-15T15:43:25-05:00 GitHub noreply@github.com 2024-12-15T21:43:25+01:00 llava : Allow locally downloaded models for QwenVL (#10833)
a0974156f334acf8af5858d7ede5ab7d7490d415 87cf323cef80f6aa530f047ab05b539ebc6b7e3c Valentin Mamedov 45292985+Inf1delis@users.noreply.github.com 2024-12-16T00:02:46+07:00 GitHub noreply@github.com 2024-12-15T19:02:46+02:00 llama : add Deepseek MoE v1 & GigaChat models (#10827)
87cf323cef80f6aa530f047ab05b539ebc6b7e3c 5478bbcd173e7027af7689493c7421719f5c43df Georgi Gerganov ggerganov@gmail.com 2024-12-15T18:44:47+02:00 GitHub noreply@github.com 2024-12-15T18:44:47+02:00 scripts : change build path to "build-bench" for compare-commits.sh (#10836)
5478bbcd173e7027af7689493c7421719f5c43df b5ae1ddff93403300fc79c7ab5ee73b8cfbb3457 Vinesh Janarthanan 36610342+VJHack@users.noreply.github.com 2024-12-15T05:55:54-06:00 GitHub noreply@github.com 2024-12-15T12:55:54+01:00 server: (UI) add syntax highlighting and latex math rendering (#10808)
b5ae1ddff93403300fc79c7ab5ee73b8cfbb3457 89d604f2c87af9db657d8a27a1528bc4b7579c29 Georgi Gerganov ggerganov@gmail.com 2024-12-15T13:16:42+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-15T13:16:42+02:00 gguf-py : bump to v0.13.0
89d604f2c87af9db657d8a27a1528bc4b7579c29 e52aba537a34d51a65cddec6bc6dafc9031edc63 Michelle Tan 41475767+MichelleTanPY@users.noreply.github.com 2024-12-14T22:29:45Z GitHub noreply@github.com 2024-12-14T23:29:45+01:00 server: Fix `has_next_line` in JSON response (#10818)
e52aba537a34d51a65cddec6bc6dafc9031edc63 ba1cb19cdd0d92e012e0f6e009e0620f854b6afd Evgeny Kurnevsky kurnevsky@gmail.com 2024-12-14T18:17:36Z GitHub noreply@github.com 2024-12-14T10:17:36-08:00 nix: allow to override rocm gpu targets (#10794)
ba1cb19cdd0d92e012e0f6e009e0620f854b6afd 56eea0781cbd2608ed8ff524955495569b9264be HimariO dsfhe49854@gmail.com 2024-12-14T20:43:46+08:00 GitHub noreply@github.com 2024-12-14T14:43:46+02:00 llama : add Qwen2VL support + multimodal RoPE (#10361)
56eea0781cbd2608ed8ff524955495569b9264be a76c56fa1a3d27467eb97468d8c3b2fe1243b61a cduk 19917266+cduk@users.noreply.github.com 2024-12-13T23:21:49+01:00 GitHub noreply@github.com 2024-12-13T23:21:49+01:00 Removes spurious \r in output that causes logging in journalctl to treat lines as binary and therefore hidden by default (#10771)
a76c56fa1a3d27467eb97468d8c3b2fe1243b61a c27ac678dd393af0da9b8acf10266e760c8a0912 lhez quic_lih@quicinc.com 2024-12-13T12:23:52-08:00 GitHub noreply@github.com 2024-12-13T12:23:52-08:00 Introducing experimental OpenCL backend with support for Qualcomm Adreno GPUs (#10693)
c27ac678dd393af0da9b8acf10266e760c8a0912 11e07fd63bac1cb642380a7a3eac03fd8703e948 Eric Curtin ecurtin@redhat.com 2024-12-13T18:34:25Z GitHub noreply@github.com 2024-12-13T19:34:25+01:00 Opt class for positional argument handling (#10508)
11e07fd63bac1cb642380a7a3eac03fd8703e948 4601a8bb6784d2ab8b4b605354b51979fbeea1d3 Corentin REGAL corentin.regal@gmail.com 2024-12-13T18:23:50+01:00 GitHub noreply@github.com 2024-12-13T18:23:50+01:00 fix: graceful shutdown for Docker images (#10815)
4601a8bb6784d2ab8b4b605354b51979fbeea1d3 9f35e44592a7646a5803620eb6a3f0ed5ac90553 Jett Janiak jettjaniak@gmail.com 2024-12-13T15:48:44+01:00 GitHub noreply@github.com 2024-12-13T16:48:44+02:00 gguf-py : numpy 2 newbyteorder fix (#9772)
9f35e44592a7646a5803620eb6a3f0ed5ac90553 64ae0655114f84f11a724bc6878c6f8f4a55560b 谢乃闻 sienaiwun@users.noreply.github.com 2024-12-13T12:56:07Z GitHub noreply@github.com 2024-12-13T13:56:07+01:00 Fix crash caused by ggml_backend_load_all when launching on Android Activity (#10812)
64ae0655114f84f11a724bc6878c6f8f4a55560b 83ed24a97b500ccdb32b90b94e6f9621ad8db79e Eve 139727413+netrunnereve@users.noreply.github.com 2024-12-13T08:42:04Z GitHub noreply@github.com 2024-12-13T09:42:04+01:00 vulkan: small mul_mat_vec optimizations (#10665)
83ed24a97b500ccdb32b90b94e6f9621ad8db79e d583cd03f663701858ba152a334cbb467fabe342 Akarshan Biswas akarshan.biswas@gmail.com 2024-12-13T12:12:15+05:30 GitHub noreply@github.com 2024-12-13T12:12:15+05:30 SYCL: Reduce most of the compiler warnings (#10748)
d583cd03f663701858ba152a334cbb467fabe342 adffa6ffd59997da59f62b72d2b79fc37e085e84 Karol Kontny 82021046+kkontny@users.noreply.github.com 2024-12-13T01:04:19+01:00 GitHub noreply@github.com 2024-12-13T01:04:19+01:00 ggml : Fix compilation issues on ARM platform when building without fp16 (#10811)
adffa6ffd59997da59f62b72d2b79fc37e085e84 274ec65af6e54039eb95cb44904af5c945dca1fa Xuan Son Nguyen thichthat@gmail.com 2024-12-12T22:53:05+01:00 GitHub noreply@github.com 2024-12-12T22:53:05+01:00 common : improve -ctv -ctk CLI arguments (#10806)
274ec65af6e54039eb95cb44904af5c945dca1fa 8faa1d4dd42f6cb26088ce7f5bbca5996b921685 Xuan Son Nguyen thichthat@gmail.com 2024-12-12T20:52:28+01:00 GitHub noreply@github.com 2024-12-12T20:52:28+01:00 contrib : add ngxson as codeowner (#10804)
8faa1d4dd42f6cb26088ce7f5bbca5996b921685 cb13ef85a444eb52a3f1b82dce198ceb25606583 a3sh 38979186+A3shTnT@users.noreply.github.com 2024-12-13T02:09:50+08:00 GitHub noreply@github.com 2024-12-12T19:09:50+01:00 CUDA: faster non-contiguous concat (#10760)
cb13ef85a444eb52a3f1b82dce198ceb25606583 4064c0e3b6c27440f5d12b7caaf90b4088c28c61 Diego Devesa slarengh@gmail.com 2024-12-12T19:02:49+01:00 GitHub noreply@github.com 2024-12-12T19:02:49+01:00 remove CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS (#10797)
4064c0e3b6c27440f5d12b7caaf90b4088c28c61 dc5301d565b7d0b2c691f7df13099aab3997479c 0cc4m picard12@live.de 2024-12-12T18:36:00+01:00 GitHub noreply@github.com 2024-12-12T18:36:00+01:00 Vulkan: Use improved q4_k and q5_k dequant code in dequant shaders (#10798)
dc5301d565b7d0b2c691f7df13099aab3997479c 9fdb1243049aa7e8211693f116daf2052d47507d 0cc4m picard12@live.de 2024-12-12T18:35:37+01:00 GitHub noreply@github.com 2024-12-12T18:35:37+01:00 Vulkan: Add VK_EXT_subgroup_size_control support to ensure full subgroups for coopmats (#10721)
9fdb1243049aa7e8211693f116daf2052d47507d 5555c0c1f66d766c544c30699190dec0285bcbfc Xuan Son Nguyen thichthat@gmail.com 2024-12-12T16:57:32+01:00 GitHub noreply@github.com 2024-12-12T16:57:32+01:00 common : add missing env var for speculative (#10801)
5555c0c1f66d766c544c30699190dec0285bcbfc 973f328b1e92a6406030442dfd15b29449e89747 CentricStorm CentricStorm@users.noreply.github.com 2024-12-11T22:40:40Z GitHub noreply@github.com 2024-12-11T23:40:40+01:00 docs: update server streaming mode documentation (#9519)
973f328b1e92a6406030442dfd15b29449e89747 235f6e14bf0ed0211c51aeff14139038ae1000aa fb18934a97425c42c8b32a1baaa94f0080eb051d Georgi Gerganov ggerganov@gmail.com 2024-12-11T23:14:46+02:00 GitHub noreply@github.com 2024-12-11T23:14:46+02:00 Merge pull request #10788 from ggerganov/gg/gguf-py-0.11.0
fb18934a97425c42c8b32a1baaa94f0080eb051d 235f6e14bf0ed0211c51aeff14139038ae1000aa Georgi Gerganov ggerganov@gmail.com 2024-12-11T23:13:31+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-11T23:13:31+02:00 gguf-py : bump version to 0.11.0
235f6e14bf0ed0211c51aeff14139038ae1000aa 1a31d0dc00ba946d448e16ecc915ce5e8355994e Xuan Son Nguyen thichthat@gmail.com 2024-12-11T20:52:14+01:00 GitHub noreply@github.com 2024-12-11T20:52:14+01:00 server : (UI) add tok/s, get rid of completion.js (#10786)
1a31d0dc00ba946d448e16ecc915ce5e8355994e 92f77a640f763c0af73554fb810a85a7d4c85e5e qingy1337 qxli2@students.everettcc.edu 2024-12-11T07:16:32-08:00 GitHub noreply@github.com 2024-12-11T16:16:32+01:00 Update README.md (#10772)
92f77a640f763c0af73554fb810a85a7d4c85e5e 484d2f31aed34ff9f096e3961125762e81d9b7d6 Xuan Son Nguyen thichthat@gmail.com 2024-12-11T14:59:41+01:00 GitHub noreply@github.com 2024-12-11T14:59:41+01:00 ci : pin nodejs to 22.11.0 (#10779)
484d2f31aed34ff9f096e3961125762e81d9b7d6 4b4d92b0986e3b627b9a9ef4782973108bf47691 kallewoof kalle.alm@gmail.com 2024-12-11T22:48:04+09:00 GitHub noreply@github.com 2024-12-11T14:48:04+01:00 bug-fix: snprintf prints NULL in place of the last character (#10419)
4b4d92b0986e3b627b9a9ef4782973108bf47691 43041d2eb32623d5b6ca734313327abe96f73146 CentricStorm CentricStorm@users.noreply.github.com 2024-12-11T10:47:43Z GitHub noreply@github.com 2024-12-11T11:47:43+01:00 docs: fix server documentation formatting (#10776)
43041d2eb32623d5b6ca734313327abe96f73146 b685daf3867c54e42a9db484d7b92619021d4510 Gilad S. 7817232+giladgd@users.noreply.github.com 2024-12-11T02:47:21+02:00 GitHub noreply@github.com 2024-12-11T01:47:21+01:00 ggml: load all backends from a user-provided search path (#10699)
b685daf3867c54e42a9db484d7b92619021d4510 dafae66cc242eb766797194d3c85c5e502625623 Jeff Bolz jbolz@nvidia.com 2024-12-10T14:23:17-06:00 GitHub noreply@github.com 2024-12-10T21:23:17+01:00 vulkan: request round-to-even for fp16 in im2col/rope_head (#10767)
dafae66cc242eb766797194d3c85c5e502625623 ae4b922614d452477cf5d2fb8cad247c9c12596c Eve 139727413+netrunnereve@users.noreply.github.com 2024-12-10T19:33:23Z GitHub noreply@github.com 2024-12-10T20:33:23+01:00 vulkan: dynamic subgroup size for the remaining k quants (#10745)
ae4b922614d452477cf5d2fb8cad247c9c12596c 750cb3e246f7e544124cf18cb0c5e0c8b7e38738 Bartowski ckealty1182@gmail.com 2024-12-10T12:23:50-05:00 GitHub noreply@github.com 2024-12-10T18:23:50+01:00 imatrix : Add imatrix to --no-context-shift (#10766)
750cb3e246f7e544124cf18cb0c5e0c8b7e38738 a86ad841f103e471ac0fd7ee8852d1eb5015ce89 Andreas Kieslinger 47689530+aendk@users.noreply.github.com 2024-12-10T18:23:24+01:00 GitHub noreply@github.com 2024-12-10T18:23:24+01:00 CUDA: rename macros to avoid conflicts with WinAPI (#10736)
a86ad841f103e471ac0fd7ee8852d1eb5015ce89 a05e2afcc241c1ecd38ec5cb4c579d90cdf3f918 Yüg eugeniosegalaweb@gmail.com 2024-12-10T17:22:34Z GitHub noreply@github.com 2024-12-10T18:22:34+01:00 server : add flag to disable the web-ui (#10762) (#10751)
a05e2afcc241c1ecd38ec5cb4c579d90cdf3f918 26a8406ba9198eb6fdd8329fa717555b4f77f05f Jeff Bolz jbolz@nvidia.com 2024-12-10T11:22:20-06:00 GitHub noreply@github.com 2024-12-10T18:22:20+01:00 vulkan: disable spirv-opt for coopmat shaders (#10763)
26a8406ba9198eb6fdd8329fa717555b4f77f05f c37fb4cf62ddf0d33562c4c4a4d6fb45e32ad3b6 Johannes Gäßler johannesg@5d6.de 2024-12-09T20:07:12+01:00 GitHub noreply@github.com 2024-12-09T20:07:12+01:00 CUDA: fix shared memory access condition for mmv (#10740)
c37fb4cf62ddf0d33562c4c4a4d6fb45e32ad3b6 3d98b4cb226c3140bd1ae6c65ed126b7d90332fa Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-12-09T23:10:19+05:30 GitHub noreply@github.com 2024-12-09T09:40:19-08:00 Changes to CMakePresets.json to add ninja clang target on windows (#10668)
3d98b4cb226c3140bd1ae6c65ed126b7d90332fa 1a05004743e00aca400833be625f0ec8cce176a7 Jeff Bolz jbolz@nvidia.com 2024-12-09T01:24:01-06:00 GitHub noreply@github.com 2024-12-09T08:24:01+01:00 vulkan: fix compile warnings (#10731)
1a05004743e00aca400833be625f0ec8cce176a7 ce8784bdb153ff7794dde5a50b0ebfa51baa6171 Borislav Stanimirov b@ibob.bg 2024-12-09T09:15:13+02:00 GitHub noreply@github.com 2024-12-09T09:15:13+02:00 cmake : simplify msvc charsets (#10672)
ce8784bdb153ff7794dde5a50b0ebfa51baa6171 e52522b8694ae73abf12feb18d29168674aa1c1b Xuan Son Nguyen thichthat@gmail.com 2024-12-08T23:04:29+01:00 GitHub noreply@github.com 2024-12-08T23:04:29+01:00 server : fix format_infill (#10724)
e52522b8694ae73abf12feb18d29168674aa1c1b 06d70147e6480c021e493c442ae0f0d83ae366de Xuan Son Nguyen thichthat@gmail.com 2024-12-08T20:38:51+01:00 GitHub noreply@github.com 2024-12-08T20:38:51+01:00 server : bring back info of final chunk in stream mode (#10722)
06d70147e6480c021e493c442ae0f0d83ae366de 43ed389a3f102517e6f7d5620d8e451e88afbf27 stduhpf stephduh@live.fr 2024-12-08T19:19:19+01:00 GitHub noreply@github.com 2024-12-08T19:19:19+01:00 Vulkan: fix NaN in tanh.comp with AMD proprietary driver on Windows (#10723)
43ed389a3f102517e6f7d5620d8e451e88afbf27 ecc93d0558fc3ecb8a5af69d2ece02fae4710ade Diego Devesa slarengh@gmail.com 2024-12-08T12:14:54+01:00 GitHub noreply@github.com 2024-12-08T13:14:54+02:00 llama : use cmake for swift build (#10525)
ecc93d0558fc3ecb8a5af69d2ece02fae4710ade 62e84d984875372f4b0fb89a67658e012ff0cc9f Jeff Bolz jbolz@nvidia.com 2024-12-08T02:05:55-06:00 GitHub noreply@github.com 2024-12-08T09:05:55+01:00 vulkan: compile a test shader in cmake to check for coopmat2 support (#10713)
62e84d984875372f4b0fb89a67658e012ff0cc9f 3573fa8e7b7f0865638b52b4e9b4d2006f0558a2 Robert Collins roberto.tomas.cuentas@gmail.com 2024-12-07T16:12:27-05:00 GitHub noreply@github.com 2024-12-07T23:12:27+02:00 llama : add 128k yarn context for Qwen (#10698)
3573fa8e7b7f0865638b52b4e9b4d2006f0558a2 d9c3ba2b7749c00df477599aa141a98b4521aa2c Xuan Son Nguyen thichthat@gmail.com 2024-12-07T20:21:09+01:00 GitHub noreply@github.com 2024-12-07T20:21:09+01:00 server : (refactor) no more json in server_task input (#10691)
d9c3ba2b7749c00df477599aa141a98b4521aa2c ce4a7b849388b67d45ad420bdd82d5efcd55647a Georgi Gerganov ggerganov@gmail.com 2024-12-07T18:38:15+02:00 GitHub noreply@github.com 2024-12-07T18:38:15+02:00 ggml : disable iq4_nl interleave size 8 (#10709)
ce4a7b849388b67d45ad420bdd82d5efcd55647a 19d8762ab61df8286367588a80b9c7db4cb568db Georgi Gerganov ggerganov@gmail.com 2024-12-07T18:02:05+02:00 GitHub noreply@github.com 2024-12-07T18:02:05+02:00 server : various fixes (#10704)
19d8762ab61df8286367588a80b9c7db4cb568db c2a16c0bdbe2e51adf318918bad82f0c3e3d6f3b Djip007 3705339+Djip007@users.noreply.github.com 2024-12-07T13:37:50+01:00 GitHub noreply@github.com 2024-12-07T14:37:50+02:00 ggml : refactor online repacking (#10446)
c2a16c0bdbe2e51adf318918bad82f0c3e3d6f3b 3df784b3050f657ea681f804187ce5bddb433e88 Georgi Gerganov ggerganov@gmail.com 2024-12-07T11:52:44+02:00 GitHub noreply@github.com 2024-12-07T11:52:44+02:00 server : fix free of spec context and batch (#10651)
3df784b3050f657ea681f804187ce5bddb433e88 86a1934978ce5daffc7e5b4251f6540ee5e7b47b 0cc4m picard12@live.de 2024-12-07T10:24:15+01:00 GitHub noreply@github.com 2024-12-07T10:24:15+01:00 Vulkan: VK_KHR_cooperative_matrix support to speed up prompt processing (#10597)
86a1934978ce5daffc7e5b4251f6540ee5e7b47b 784a14aa496a66cc43e76014a1bf4333b4a2a55a Robert Ormandi 52251610+ormandi@users.noreply.github.com 2024-12-07T01:55:01-06:00 GitHub noreply@github.com 2024-12-07T09:55:01+02:00 metal : Extend how Llama.cpp locates metal resources (#10676)
784a14aa496a66cc43e76014a1bf4333b4a2a55a c5ede3849fc021174862f9c0bf8273808d8f0d39 Sukriti Sharma Ssukriti@users.noreply.github.com 2024-12-07T00:02:14-07:00 GitHub noreply@github.com 2024-12-07T09:02:14+02:00 convert : add support for Roberta embeddings (#10695)
c5ede3849fc021174862f9c0bf8273808d8f0d39 f162d45a21b27f7613f14539f9a4932d6ff3ca48 Georgi Gerganov ggerganov@gmail.com 2024-12-06T21:33:15+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-06T21:33:49+02:00 convert : add custom attention mapping
f162d45a21b27f7613f14539f9a4932d6ff3ca48 6c5bc0625fae6909cb40def15bc4bb45db6f7f4d Xuan Son Nguyen thichthat@gmail.com 2024-12-06T13:29:05+01:00 GitHub noreply@github.com 2024-12-06T13:29:05+01:00 common : bring back --no-warmup to server (#10686)
6c5bc0625fae6909cb40def15bc4bb45db6f7f4d 7736837d62efed1dbebfe579472fca041eda12d6 Xuan Son Nguyen thichthat@gmail.com 2024-12-06T11:14:32+01:00 GitHub noreply@github.com 2024-12-06T11:14:32+01:00 server : (refactoring) do not rely on JSON internally (#10643)
7736837d62efed1dbebfe579472fca041eda12d6 c9c6e01daedac542b174c235872569fce5385982 Plamen Minev pacominev@gmail.com 2024-12-05T23:36:41+02:00 GitHub noreply@github.com 2024-12-05T22:36:41+01:00 fix(server) : not show alert when DONE is received (#10674)
c9c6e01daedac542b174c235872569fce5385982 6fe624783166e7355cec915de0094e63cd3558eb Jeff Bolz jbolz@nvidia.com 2024-12-05T13:15:05-06:00 GitHub noreply@github.com 2024-12-05T20:15:05+01:00 vulkan: Add VK_NV_cooperative_matrix2 support for mul_mat and flash attention (#10206)
6fe624783166e7355cec915de0094e63cd3558eb 0cd182ebcc2c45907240e727e80e8fbf0f5fdee9 Riccardo Orlando Riccorl@users.noreply.github.com 2024-12-05T19:30:59+01:00 GitHub noreply@github.com 2024-12-05T20:30:59+02:00 llama : add Minerva 7B model support (#10673)
0cd182ebcc2c45907240e727e80e8fbf0f5fdee9 a8cbab201dcf4c76c30b8028427494d71f02bb57 Georgi Gerganov ggerganov@gmail.com 2024-12-05T13:27:42+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-05T13:27:42+02:00 sync : ggml
a8cbab201dcf4c76c30b8028427494d71f02bb57 c2082d93a82d66dca112098c63775d7a97e6d47b PAB pierreantoine.bannier@gmail.com 2024-12-04T09:19:30+01:00 Georgi Gerganov ggerganov@gmail.com 2024-12-05T13:27:33+02:00 ggml: add `GGML_SET` Metal kernel + i32 CPU kernel (ggml/1037)
c2082d93a82d66dca112098c63775d7a97e6d47b d405804be84cfdac936f28b3446ce8cb988408d8 PAB pierreantoine.bannier@gmail.com 2024-12-03T20:20:04+01:00 Georgi Gerganov ggerganov@gmail.com 2024-12-05T13:27:31+02:00 ggml : add `GGML_PAD_REFLECT_1D` operation (ggml/1034)
d405804be84cfdac936f28b3446ce8cb988408d8 f112d198cd9953b633ac662c2705d6e423438717 Daniel Bevenius daniel.bevenius@gmail.com 2024-12-05T08:47:55+01:00 GitHub noreply@github.com 2024-12-05T09:47:55+02:00 py : update outdated copy-paste instructions [no ci] (#10667)
f112d198cd9953b633ac662c2705d6e423438717 1da7b765692764a8b33b08da61cbee63812a7bd9 aryantandon01 80969509+aryantandon01@users.noreply.github.com 2024-12-05T03:49:20+05:30 GitHub noreply@github.com 2024-12-04T23:19:20+01:00 Update deprecation-warning.cpp (#10619)
1da7b765692764a8b33b08da61cbee63812a7bd9 59f4db10883a4f3e855cffbf2c3ab68430e95272 Georgi Gerganov ggerganov@gmail.com 2024-12-04T22:38:20+02:00 GitHub noreply@github.com 2024-12-04T22:38:20+02:00 server : fix speculative decoding with context shift (#10641)
59f4db10883a4f3e855cffbf2c3ab68430e95272 2803540814bf0a4e44d0960ff6afda6bac971c17 Diego Devesa slarengh@gmail.com 2024-12-04T14:45:40+01:00 GitHub noreply@github.com 2024-12-04T14:45:40+01:00 ggml : add predefined list of CPU backend variants to build (#10626)
2803540814bf0a4e44d0960ff6afda6bac971c17 253b7fde910731104670724391bfbcb94d97d0c3 Diego Devesa slarengh@gmail.com 2024-12-04T14:40:44+01:00 GitHub noreply@github.com 2024-12-04T14:40:44+01:00 ggml-cpu : fix HWCAP2_I8MM value (#10646)
253b7fde910731104670724391bfbcb94d97d0c3 8d0cfd554a9ae545ff94d27e04458f537b4e8c0e ltoniazzi 61414566+ltoniazzi@users.noreply.github.com 2024-12-04T09:45:48Z GitHub noreply@github.com 2024-12-04T10:45:48+01:00 Fix HF repo commit to clone lora test models (#10649)
8d0cfd554a9ae545ff94d27e04458f537b4e8c0e 2759916d86b70e7aceaed4d0b4e7ed126f0f9e51 JFLFY2255 JFLFY2255@163.com 2024-12-04T17:42:50+08:00 GitHub noreply@github.com 2024-12-04T11:42:50+02:00 llama: Support MiniCPM-1B (with & w/o longrope) (#10559)
2759916d86b70e7aceaed4d0b4e7ed126f0f9e51 40c6d79fb52f995f47507fedfeaae2ac05d9b35c Jeff Bolz jbolz@nvidia.com 2024-12-04T01:28:59-06:00 GitHub noreply@github.com 2024-12-04T08:28:59+01:00 vulkan: Implement "fast divide" (mul+shift) for unary ops like copy (#10642)
40c6d79fb52f995f47507fedfeaae2ac05d9b35c 98036d5670f21e9b9a99d5e3dbb3bf7589f5c4e3 Nicolò Scipione nicolo.scipione@codeplay.com 2024-12-04T02:29:20+01:00 GitHub noreply@github.com 2024-12-04T09:29:20+08:00 SYCL : Move to compile time oneMKL interface backend selection for NVIDIA backend (#10584)
98036d5670f21e9b9a99d5e3dbb3bf7589f5c4e3 cd2f37b304f8e88b9de8424b31078b97f9cf7c60 Wang Ran (汪然) wangr@smail.nju.edu.cn 2024-12-04T09:22:50+08:00 GitHub noreply@github.com 2024-12-04T02:22:50+01:00 fix typo of README.md (#10605)
cd2f37b304f8e88b9de8424b31078b97f9cf7c60 da6aac91f150a3b0bcc26d3fd50288accb15f179 Frankie Robertson frankier@users.noreply.github.com 2024-12-04T02:41:37+02:00 GitHub noreply@github.com 2024-12-04T01:41:37+01:00 Avoid using __fp16 on ARM with old nvcc (#10616)
da6aac91f150a3b0bcc26d3fd50288accb15f179 01e6d9bb71eb71fe1f811f2fdef15753232cd0f2 Benson Wong mostlygeek@gmail.com 2024-12-03T16:40:36-08:00 GitHub noreply@github.com 2024-12-04T01:40:36+01:00 Add docs for creating a static build (#10268) (#10630)
01e6d9bb71eb71fe1f811f2fdef15753232cd0f2 cc98896db858df7aa40d0e16a505883ef196a482 piDack 104877312+piDack@users.noreply.github.com 2024-12-04T08:26:37+08:00 GitHub noreply@github.com 2024-12-04T01:26:37+01:00 clip : add sycl support (#10574)
cc98896db858df7aa40d0e16a505883ef196a482 91c36c269bca75b2d08119c653512cd20b4ea2ba Jeff Bolz jbolz@nvidia.com 2024-12-03T13:29:54-06:00 GitHub noreply@github.com 2024-12-03T20:29:54+01:00 vulkan: optimize and reenable split_k (#10637)
91c36c269bca75b2d08119c653512cd20b4ea2ba 1cd3df46bd49a0c1c78da8b68c956448a73b7476 Xuan Son Nguyen thichthat@gmail.com 2024-12-03T19:38:44+01:00 GitHub noreply@github.com 2024-12-03T19:38:44+01:00 server : (web ui) Various improvements, now use vite as bundler (#10599)
1cd3df46bd49a0c1c78da8b68c956448a73b7476 c5054718575d37f43cc4e6f61ea7c4014ad2c0cf Georgi Gerganov ggerganov@gmail.com 2024-12-03T19:42:30+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-03T20:04:49+02:00 scripts : remove amx sync
c5054718575d37f43cc4e6f61ea7c4014ad2c0cf e9e661bd59364e5d4fce035834b6cadcadf8c2ef Georgi Gerganov ggerganov@gmail.com 2024-12-03T19:40:25+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-03T20:04:49+02:00 sync : ggml
e9e661bd59364e5d4fce035834b6cadcadf8c2ef efb6ae963031709fc331e6e48cc4606ac8f9c3a7 mahorozte 41834471+mahorozte@users.noreply.github.com 2024-12-03T21:11:43+08:00 Georgi Gerganov ggerganov@gmail.com 2024-12-03T20:04:49+02:00 CUDA: remove unnecessary warp reduce in FA (ggml/1032)
efb6ae963031709fc331e6e48cc4606ac8f9c3a7 667d70d1704dfa6977505f5d01d4638669b90dce PAB pierreantoine.bannier@gmail.com 2024-12-02T19:27:24+01:00 Georgi Gerganov ggerganov@gmail.com 2024-12-03T20:04:49+02:00 feat: add `GGML_UNARY_OP_ARGMAX` Metal kernel (ggml/1019)
667d70d1704dfa6977505f5d01d4638669b90dce 3b4f2e33e2cbfca621e623c4b92b88da57a8c2f4 PAB pierreantoine.bannier@gmail.com 2024-11-28T09:25:06+01:00 Georgi Gerganov ggerganov@gmail.com 2024-12-03T20:04:49+02:00 metal : add `GGML_OP_CONV_TRANSPOSE_1D` kernels (ggml/1026)
3b4f2e33e2cbfca621e623c4b92b88da57a8c2f4 82bca2257b3cec72676abb26011f1b99fcdab29d Xuan Son Nguyen thichthat@gmail.com 2024-12-03T12:54:30+01:00 GitHub noreply@github.com 2024-12-03T12:54:30+01:00 llama : add missing LLAMA_API for llama_chat_builtin_templates (#10636)
82bca2257b3cec72676abb26011f1b99fcdab29d 0115df2f65ac7c64dd0e5915c72ecc4a9343a130 Nikolaos Pothitos pothitos@di.uoa.gr 2024-12-03T12:50:08+02:00 GitHub noreply@github.com 2024-12-03T12:50:08+02:00 readme : add option, update default value, fix formatting (#10271)
0115df2f65ac7c64dd0e5915c72ecc4a9343a130 515d4e53727924e48774f45ecb15bdacbf851e13 Georgi Gerganov ggerganov@gmail.com 2024-12-03T11:52:33+02:00 GitHub noreply@github.com 2024-12-03T11:52:33+02:00 metal : small-batch mat-mul kernels (#10581)
515d4e53727924e48774f45ecb15bdacbf851e13 844e2e1feec887c803845a3b8762f3b15979b095 Georgi Gerganov ggerganov@gmail.com 2024-12-03T11:21:43+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-03T11:21:43+02:00 github : minify link [no ci] (revert)
844e2e1feec887c803845a3b8762f3b15979b095 70b98fadbc8c07a0144f3b50a4d7ab7e2aeff878 Georgi Gerganov ggerganov@gmail.com 2024-12-03T11:20:35+02:00 Georgi Gerganov ggerganov@gmail.com 2024-12-03T11:20:35+02:00 github : minify link [no ci]
70b98fadbc8c07a0144f3b50a4d7ab7e2aeff878 642330ac7cea11dc1f9d3df2b8f3dbd10b5e3f3e Georgi Gerganov ggerganov@gmail.com 2024-12-03T11:20:00+02:00 GitHub noreply@github.com 2024-12-03T11:20:00+02:00 server : fix default draft model parameters (#10586)
642330ac7cea11dc1f9d3df2b8f3dbd10b5e3f3e 8648c521010620c2daccfa1d26015c668ba2c717 Xuan Son Nguyen thichthat@gmail.com 2024-12-02T22:10:19+01:00 GitHub noreply@github.com 2024-12-02T22:10:19+01:00 llama : add enum for built-in chat templates (#10623)
8648c521010620c2daccfa1d26015c668ba2c717 64ed2091b24b2f9747148fdf49a34ed5938762c3 Georgi Gerganov ggerganov@gmail.com 2024-12-02T21:22:53+02:00 GitHub noreply@github.com 2024-12-02T21:22:53+02:00 make : deprecate (#10514)
64ed2091b24b2f9747148fdf49a34ed5938762c3 991f8aabeec89d801300bb179e52013fb0eb0584 haopeng 657407891@qq.com 2024-12-02T21:45:54+08:00 GitHub noreply@github.com 2024-12-02T14:45:54+01:00 server: Add "tokens per second" information in the backend (#10548)
991f8aabeec89d801300bb179e52013fb0eb0584 4cb003dd8d1f37523120a21e4b1a50a2adcb8c84 Akarshan Biswas akarshan.biswas@gmail.com 2024-12-02T12:34:11+05:30 GitHub noreply@github.com 2024-12-02T15:04:11+08:00 SYCL: Fix and switch to GGML_LOG system instead of fprintf (#10579)
4cb003dd8d1f37523120a21e4b1a50a2adcb8c84 917786f43d0f29b7c77a0c56767c0fa4df68b1c5 Georgi Gerganov ggerganov@gmail.com 2024-12-02T08:53:27+02:00 GitHub noreply@github.com 2024-12-02T08:53:27+02:00 contrib : refresh (#10593)
917786f43d0f29b7c77a0c56767c0fa4df68b1c5 5e1ed95583ca552a98d8528b73e1ff81249c2bf9 Juk Armstrong 69222624+jukofyork@users.noreply.github.com 2024-12-01T22:09:49Z GitHub noreply@github.com 2024-12-01T23:09:49+01:00 Add `mistral-v1`, `mistral-v3`, `mistral-v3-tekken` and `mistral-v7` chat template types (#10572)
5e1ed95583ca552a98d8528b73e1ff81249c2bf9 5c7a5aa0c32eb19ce03e178560797db5875d7692 Georgi Gerganov ggerganov@gmail.com 2024-12-01T21:37:54+02:00 GitHub noreply@github.com 2024-12-01T21:37:54+02:00 grammars : add English-only grammar (#10612)
5c7a5aa0c32eb19ce03e178560797db5875d7692 3420909dffa50e70660524797a1e715a717684d2 Wang Qin 37098874+wangqin0@users.noreply.github.com 2024-12-01T10:11:42-08:00 GitHub noreply@github.com 2024-12-01T20:11:42+02:00 ci: add error handling for Python venv creation in run.sh (#10608)
3420909dffa50e70660524797a1e715a717684d2 86dc11c5bcf34db2749d8bd8d4fa07a542c94f84 Diego Devesa slarengh@gmail.com 2024-12-01T16:12:41+01:00 GitHub noreply@github.com 2024-12-01T16:12:41+01:00 ggml : automatic selection of best CPU backend (#10606)
86dc11c5bcf34db2749d8bd8d4fa07a542c94f84 6acce3971098772a8aacb10fe8550b4119110581 alek3y 44779186+alek3y@users.noreply.github.com 2024-12-01T12:33:12+01:00 GitHub noreply@github.com 2024-12-01T13:33:12+02:00 server : bind to any port when specified (#10590)
6acce3971098772a8aacb10fe8550b4119110581 43957ef203b4c9ceaee42c176b3ef44ea4359c85 Georgi Gerganov ggerganov@gmail.com 2024-12-01T11:25:17+02:00 GitHub noreply@github.com 2024-12-01T11:25:17+02:00 readme : update the usage section with examples (#10596)
43957ef203b4c9ceaee42c176b3ef44ea4359c85 0c39f44d70d058940fe2afe50cfc789e3e44d756 Wang Qin 37098874+wangqin0@users.noreply.github.com 2024-11-30T19:19:44-08:00 GitHub noreply@github.com 2024-12-01T04:19:44+01:00 build: update Makefile comments for C++ version change (#10598)
0c39f44d70d058940fe2afe50cfc789e3e44d756 3e0ba0e604b1ac5b2cbca9a3f38f91f2be4ef1cd Adrien Gallouët adrien@gallouet.fr 2024-11-30T18:13:18+01:00 GitHub noreply@github.com 2024-11-30T09:13:18-08:00 ggml-cpu: replace AArch64 NEON assembly with intrinsics in ggml_gemv_q4_0_4x4_q8_0() (#10567)
3e0ba0e604b1ac5b2cbca9a3f38f91f2be4ef1cd abadba05be52cccf6c0da49534e37f6062ce8ded Georgi Gerganov ggerganov@gmail.com 2024-11-30T10:09:21+02:00 GitHub noreply@github.com 2024-11-30T10:09:21+02:00 readme : remove old badge
abadba05be52cccf6c0da49534e37f6062ce8ded 0533e7fb3842a523f64dc533bd7bd7147ec2c63a Georgi Gerganov ggerganov@gmail.com 2024-11-30T09:47:07+02:00 GitHub noreply@github.com 2024-11-30T09:47:07+02:00 readme : refresh (#10587)
0533e7fb3842a523f64dc533bd7bd7147ec2c63a 7cc2d2c88908fc92b97b28acafb82f7d6e425b85 Eve 139727413+netrunnereve@users.noreply.github.com 2024-11-30T07:00:02Z GitHub noreply@github.com 2024-11-30T08:00:02+01:00 vulkan: Dynamic subgroup size support for Q6_K mat_vec (#10536)
7cc2d2c88908fc92b97b28acafb82f7d6e425b85 b782e5c7d453b3f1fa8dc6c34cde7e2fa946af93 Diego Devesa slarengh@gmail.com 2024-11-29T21:54:58+01:00 GitHub noreply@github.com 2024-11-29T21:54:58+01:00 ggml : move AMX to the CPU backend (#10570)
b782e5c7d453b3f1fa8dc6c34cde7e2fa946af93 3a8e9af402f7893423bdab444aa16c5d9a2d429a Xuan Son Nguyen thichthat@gmail.com 2024-11-29T21:48:56+01:00 GitHub noreply@github.com 2024-11-29T21:48:56+01:00 server : add more test cases (#10569)
3a8e9af402f7893423bdab444aa16c5d9a2d429a a3a3048e7a0f9464d0d625a29257d8bce5da5090 Robert Collins roberto.tomas.cuentas@gmail.com 2024-11-29T12:21:37-05:00 GitHub noreply@github.com 2024-11-29T19:21:37+02:00 imatrix : support combine-only (#10492)
a3a3048e7a0f9464d0d625a29257d8bce5da5090 f0678c5ff4cb8873d6ff48801475ff270db656fa Diego Devesa slarengh@gmail.com 2024-11-29T17:45:08+01:00 GitHub noreply@github.com 2024-11-29T17:45:08+01:00 cleanup UI link list (#10577)
f0678c5ff4cb8873d6ff48801475ff270db656fa 4b3242bbea172ac0980378496fbc676d44c4f459 Georgi Gerganov ggerganov@gmail.com 2024-11-29T16:25:39+02:00 GitHub noreply@github.com 2024-11-29T16:25:39+02:00 ggml : fix I8MM Q4_1 scaling factor conversion (#10562)
4b3242bbea172ac0980378496fbc676d44c4f459 0f77aae5608f16780a49926b67be6d56ec4b09bd Shupei Fan dymarkfan@outlook.com 2024-11-29T21:49:02+08:00 GitHub noreply@github.com 2024-11-29T14:49:02+01:00 ggml-cpu: fix typo in gemv/gemm iq4_nl_4_4 (#10580)
0f77aae5608f16780a49926b67be6d56ec4b09bd 266b8519ee6d21e7ba2bf56f5629e20a181fee8b Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-11-29T12:38:45Z GitHub noreply@github.com 2024-11-29T20:38:45+08:00 sycl : offload of get_rows set to 0 (#10432)
266b8519ee6d21e7ba2bf56f5629e20a181fee8b 938f6087421889a3af7d0786c64406ced2be81b8 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-11-29T09:49:43Z GitHub noreply@github.com 2024-11-29T09:49:43Z sycl : Reroute permuted mul_mats through oneMKL (#10408)
938f6087421889a3af7d0786c64406ced2be81b8 f095a649ec390e04dfab1b04e646ae8549dafaef Chenguang Li 87689256+noemotiovon@users.noreply.github.com 2024-11-29T14:46:55+08:00 GitHub noreply@github.com 2024-11-29T14:46:55+08:00 CANN: RoPE operator optimization (#10563)
f095a649ec390e04dfab1b04e646ae8549dafaef 678d7994f4da0af3d29046be99950ac999ee9762 Jeff Bolz jbolz@nvidia.com 2024-11-29T00:18:02-06:00 GitHub noreply@github.com 2024-11-29T07:18:02+01:00 vulkan: get the first command buffer submitted sooner (#10499)
678d7994f4da0af3d29046be99950ac999ee9762 dc22344088a7ee81a1e4f096459b03a72f24ccdc Ting Lou louting@189.cn 2024-11-29T08:09:46+08:00 GitHub noreply@github.com 2024-11-29T01:09:46+01:00 llava: return false instead of exit (#10546)
dc22344088a7ee81a1e4f096459b03a72f24ccdc 4c0a95b1074907ce7efe6f5bb6ae3351c01429ab Georgi Gerganov ggerganov@gmail.com 2024-11-28T20:46:40+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-28T20:46:40+02:00 ggml : remove redundant copyright notice + update authors
4c0a95b1074907ce7efe6f5bb6ae3351c01429ab 6c595676899013102fdb0aa4b06a49954300c94a Georgi Gerganov ggerganov@gmail.com 2024-11-28T20:45:07+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-28T20:45:07+02:00 llama : add missing model types
6c595676899013102fdb0aa4b06a49954300c94a 890719311b6535e572f15965c6d7ec4ac2537f60 Xuan Son Nguyen thichthat@gmail.com 2024-11-28T19:17:49+01:00 GitHub noreply@github.com 2024-11-28T19:17:49+01:00 server : (tests) don't use thread for capturing stdout/stderr, bump openai client library (#10568)
890719311b6535e572f15965c6d7ec4ac2537f60 7281cf13addfae9b64bb2be87e3b5b1914505d63 Johannes Gäßler johannesg@5d6.de 2024-11-28T18:15:25+01:00 GitHub noreply@github.com 2024-11-28T18:15:25+01:00 common: fix warning message when no GPU found (#10564)
7281cf13addfae9b64bb2be87e3b5b1914505d63 e90688edd004fdb7063f463bd18408ba9ae008dd Random Fly renfei8@live.cn 2024-11-28T23:03:11+08:00 GitHub noreply@github.com 2024-11-28T16:03:11+01:00 docs: fix outdated usage of llama-simple (#10565)
e90688edd004fdb7063f463bd18408ba9ae008dd 76b27d29c22af03172cf211a8a31025c7c828a57 Diego Devesa slarengh@gmail.com 2024-11-28T15:58:54+01:00 GitHub noreply@github.com 2024-11-28T15:58:54+01:00 ci : fix tag name in cuda and hip releases (#10566)
76b27d29c22af03172cf211a8a31025c7c828a57 eea986f215e1dc490654d012ccf2ab62fe8f606d Georgi Gerganov ggerganov@gmail.com 2024-11-28T14:56:37+02:00 GitHub noreply@github.com 2024-11-28T14:56:37+02:00 ggml : fix row condition for i8mm kernels (#10561)
eea986f215e1dc490654d012ccf2ab62fe8f606d c202cef1686182a78f8f4e253ab8d0c0ffe2fcc8 Georgi Gerganov ggerganov@gmail.com 2024-11-28T14:56:23+02:00 GitHub noreply@github.com 2024-11-28T14:56:23+02:00 cmake : fix ARM feature detection (#10543)
c202cef1686182a78f8f4e253ab8d0c0ffe2fcc8 2025fa67e94358deda4740a74fe9803916cb2f60 Shupei Fan dymarkfan@outlook.com 2024-11-28T20:52:03+08:00 GitHub noreply@github.com 2024-11-28T13:52:03+01:00 ggml-cpu: support IQ4_NL_4_4 by runtime repack (#10541)
2025fa67e94358deda4740a74fe9803916cb2f60 c6bc73951ed52466392b1abda98c28ecbe522c7f Sergio López slp@redhat.com 2024-11-28T12:51:38+01:00 GitHub noreply@github.com 2024-11-28T12:51:38+01:00 kompute : improve backend to pass test_backend_ops (#10542)
c6bc73951ed52466392b1abda98c28ecbe522c7f 605fa66c509f9f117bd654cf0b9b3ea08bb86e80 Ruixin Huang 18860020911@163.com 2024-11-28T15:27:11+08:00 GitHub noreply@github.com 2024-11-28T15:27:11+08:00 CANN: Update cann.md to display correctly in CLion (#10538)
605fa66c509f9f117bd654cf0b9b3ea08bb86e80 b7420131bf8ab3e067bc660439ab1ab18be7edbd leo-pony nengjunma@outlook.com 2024-11-28T15:25:24+08:00 GitHub noreply@github.com 2024-11-28T15:25:24+08:00 CANN: Fix SOC_TYPE compile bug (#10519)
b7420131bf8ab3e067bc660439ab1ab18be7edbd 9f912511bc9414fa7a3c521378b6388cd932b58d Chenguang Li 87689256+noemotiovon@users.noreply.github.com 2024-11-28T14:24:46+08:00 GitHub noreply@github.com 2024-11-28T14:24:46+08:00 CANN: ROPE operator optimization (#10540)
9f912511bc9414fa7a3c521378b6388cd932b58d 3ad5451f3b75809e3033e4e577b9f60bcaf6676a Xuan Son Nguyen thichthat@gmail.com 2024-11-27T22:30:52+01:00 GitHub noreply@github.com 2024-11-27T22:30:52+01:00 common : fix duplicated file name with hf_repo and hf_file (#10550)
3ad5451f3b75809e3033e4e577b9f60bcaf6676a 46c69e0e752ff16206347bb12f96ed69f4a01abf uvos devnull@uvos.xyz 2024-11-27T17:10:08+01:00 GitHub noreply@github.com 2024-11-27T17:10:08+01:00 Add some minimal optimizations for CDNA (#10498)
46c69e0e752ff16206347bb12f96ed69f4a01abf 9e2301f4a4ef1690bd99360c11de43fe830b1c8d Diego Devesa slarengh@gmail.com 2024-11-27T11:03:25+01:00 GitHub noreply@github.com 2024-11-27T11:03:25+01:00 ci : faster CUDA toolkit installation method and use ccache (#10537)
9e2301f4a4ef1690bd99360c11de43fe830b1c8d fee824a1a1e35b5c49d482f654613addade61764 Georgi Gerganov ggerganov@gmail.com 2024-11-27T11:22:14+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-27T11:22:14+02:00 metal : fix group_norm support condition (#0)
fee824a1a1e35b5c49d482f654613addade61764 9150f8fef95327474d39ccd6c6e30787e85f3529 Georgi Gerganov ggerganov@gmail.com 2024-11-27T11:10:42+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-27T11:10:42+02:00 sync : ggml
9150f8fef95327474d39ccd6c6e30787e85f3529 c31ed2abfce05c38a2a5189586bfae45a139a547 Frankie Robertson frankier@users.noreply.github.com 2024-11-26T15:50:26+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-27T11:10:27+02:00 Do not include arm_neon.h when compiling CUDA code (ggml/1028)
c31ed2abfce05c38a2a5189586bfae45a139a547 5b3466bedfa84aa29c6871c7254467550186ecc6 Jeff Bolz jbolz@nvidia.com 2024-11-27T01:32:54-06:00 GitHub noreply@github.com 2024-11-27T08:32:54+01:00 vulkan: define all quant data structures in types.comp (#10440)
5b3466bedfa84aa29c6871c7254467550186ecc6 249a7902ec710c8d027b9cc0ed10219d2b4184f8 Jeff Bolz jbolz@nvidia.com 2024-11-27T01:30:27-06:00 GitHub noreply@github.com 2024-11-27T08:30:27+01:00 vulkan: Handle GPUs with less shared memory (#10468)
249a7902ec710c8d027b9cc0ed10219d2b4184f8 71a64989a5d2e25c13507efada145f12cf358914 Jeff Bolz jbolz@nvidia.com 2024-11-27T01:21:59-06:00 GitHub noreply@github.com 2024-11-27T08:21:59+01:00 vulkan: further optimize q5_k mul_mat_vec (#10479)
71a64989a5d2e25c13507efada145f12cf358914 4a57d362e1948ada50af997a92c3cbff9711e78b Jeff Bolz jbolz@nvidia.com 2024-11-27T01:08:54-06:00 GitHub noreply@github.com 2024-11-27T08:08:54+01:00 vulkan: skip integer div/mod in get_offsets for batch_idx==0 (#10506)
4a57d362e1948ada50af997a92c3cbff9711e78b c9b00a70b080d5c0668608024afc3e0e2fed822f Jeff Bolz jbolz@nvidia.com 2024-11-27T01:00:50-06:00 GitHub noreply@github.com 2024-11-27T08:00:50+01:00 vulkan: optimize Q2_K and Q3_K mul_mat_vec (#10459)
c9b00a70b080d5c0668608024afc3e0e2fed822f de5097351caffb3deaea3393633609df49ef41d0 Diego Devesa slarengh@gmail.com 2024-11-26T22:12:10+01:00 GitHub noreply@github.com 2024-11-26T22:12:10+01:00 ci : fix cuda releases (#10532)
de5097351caffb3deaea3393633609df49ef41d0 5a349f2809dc825960dfcfdf8f76b19cd0345be7 Shane A shanea@allenai.org 2024-11-26T12:55:29-08:00 GitHub noreply@github.com 2024-11-26T21:55:29+01:00 Add OLMo 2 model in docs (#10530)
5a349f2809dc825960dfcfdf8f76b19cd0345be7 30ec39832165627dd6ed98938df63adfc6e6a21a Diego Devesa slarengh@gmail.com 2024-11-26T21:13:54+01:00 GitHub noreply@github.com 2024-11-26T21:13:54+01:00 ci : remove nix workflows (#10526)
30ec39832165627dd6ed98938df63adfc6e6a21a be0e350c8b69632b27d5fb41fa064fa256dd7fbf Diego Devesa slarengh@gmail.com 2024-11-26T21:01:47+01:00 GitHub noreply@github.com 2024-11-26T21:01:47+01:00 llama : disable warnings for 3rd party sha1 dependency (#10527)
be0e350c8b69632b27d5fb41fa064fa256dd7fbf 249cd93da3df9c8fa78869b0522526d1625aca91 Tristan Druyen tristan@vault81.mozmail.com 2024-11-26T19:27:28+01:00 GitHub noreply@github.com 2024-11-26T19:27:28+01:00 Fix HIP flag inconsistency & build docs (#10524)
249cd93da3df9c8fa78869b0522526d1625aca91 904109ed0d97c9b656a5e8bf612925f739bb8166 R0CKSTAR xiaodong.ye@mthreads.com 2024-11-27T00:00:41+08:00 GitHub noreply@github.com 2024-11-26T17:00:41+01:00 mtgpu: Add MUSA_DOCKER_ARCH in Dockerfiles && update cmake and make (#10516)
904109ed0d97c9b656a5e8bf612925f739bb8166 45abe0f74ee281aea6e5283c1e738061256cfcae Jeff Bolz jbolz@nvidia.com 2024-11-26T09:45:05-06:00 GitHub noreply@github.com 2024-11-26T16:45:05+01:00 vulkan: fix group_norm (#10496)
45abe0f74ee281aea6e5283c1e738061256cfcae 0bbd2262a3263f37385297b30de37941836e57f7 Xuan Son Nguyen thichthat@gmail.com 2024-11-26T16:20:18+01:00 GitHub noreply@github.com 2024-11-26T16:20:18+01:00 server : replace behave with pytest (#10416)
0bbd2262a3263f37385297b30de37941836e57f7 ab96610b1e58684bc5e8b810130c4cf6d8252e21 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-11-26T21:43:47+08:00 GitHub noreply@github.com 2024-11-26T21:43:47+08:00 restore the condistion to build & update pacakge when merge (#10507)
ab96610b1e58684bc5e8b810130c4cf6d8252e21 7db3846a94ce7683b3e8120abe427457edf840c9 Georgi Gerganov ggerganov@gmail.com 2024-11-26T14:18:08+02:00 GitHub noreply@github.com 2024-11-26T14:18:08+02:00 cmake : enable warnings in llama (#10474)
7db3846a94ce7683b3e8120abe427457edf840c9 c6807b3f28cc3dbfda3ec390bcb87e69fb5634e2 Diego Devesa slarengh@gmail.com 2024-11-26T13:05:20+01:00 GitHub noreply@github.com 2024-11-26T13:05:20+01:00 ci : publish the docker images created during scheduled runs (#10515)
c6807b3f28cc3dbfda3ec390bcb87e69fb5634e2 25669aa92caaddff09f39b54a5173e5cb2680fa3 Diego Devesa slarengh@gmail.com 2024-11-26T13:05:07+01:00 GitHub noreply@github.com 2024-11-26T13:05:07+01:00 ci : add ubuntu cuda build, build with one arch on windows (#10456)
25669aa92caaddff09f39b54a5173e5cb2680fa3 84e1c33cde9e0a7aafcda2d4f21ba51c300482d7 Charles Xu charles.xu@arm.com 2024-11-26T12:37:05+01:00 GitHub noreply@github.com 2024-11-26T13:37:05+02:00 ggml-cpu: cmake add arm64 cpu feature check for macos (#10487)
84e1c33cde9e0a7aafcda2d4f21ba51c300482d7 811872a59daefb25fc0c4326bcb6d8ae893c2f7c Georgi Gerganov ggerganov@gmail.com 2024-11-26T13:36:40+02:00 GitHub noreply@github.com 2024-11-26T13:36:40+02:00 server : fix parallel speculative decoding (#10513)
811872a59daefb25fc0c4326bcb6d8ae893c2f7c 9a4b79bcfa4338b922fa8cf903bd5ac058aaf46f Georgi Gerganov ggerganov@gmail.com 2024-11-26T12:29:38+02:00 GitHub noreply@github.com 2024-11-26T12:29:38+02:00 speculative : simplify the implementation (#10504)
9a4b79bcfa4338b922fa8cf903bd5ac058aaf46f 7066b4cce2898993e943ad6af5d8f1de5840c8e9 Shanshan Shen 467638484@qq.com 2024-11-26T18:08:37+08:00 GitHub noreply@github.com 2024-11-26T18:08:37+08:00 CANN: Improve the Inferencing Performance for Ascend NPU Device (#10454)
7066b4cce2898993e943ad6af5d8f1de5840c8e9 0eb4e12beebabae46d37b78742f4c5d4dbe52dc1 Chenguang Li 87689256+noemotiovon@users.noreply.github.com 2024-11-26T17:31:05+08:00 GitHub noreply@github.com 2024-11-26T17:31:05+08:00 CANN: RoPE and CANCAT operator optimization (#10488)
0eb4e12beebabae46d37b78742f4c5d4dbe52dc1 0cc63754b831d3a6c37bc5d721d12ce9540ffe76 Junil Kim logyourself@gmail.com 2024-11-26T10:47:20+09:00 GitHub noreply@github.com 2024-11-26T01:47:20Z vulkan: Fix a vulkan-shaders-gen arugment parsing error (#10484)
0cc63754b831d3a6c37bc5d721d12ce9540ffe76 50d5cecbda3b0d03344eed326287adc1f6c7f3ef Eric Curtin ecurtin@redhat.com 2024-11-25T16:56:24-05:00 GitHub noreply@github.com 2024-11-25T22:56:24+01:00 Introduce llama-run (#10291)
50d5cecbda3b0d03344eed326287adc1f6c7f3ef 9fd8c2687f5aa2f095ac6e12a376e1c0583888e8 Diego Devesa slarengh@gmail.com 2024-11-25T22:05:39+01:00 GitHub noreply@github.com 2024-11-25T22:05:39+01:00 ci : build docker images only once daily (#10503)
9fd8c2687f5aa2f095ac6e12a376e1c0583888e8 47f931c8f9a26c072d71224bc8013cc66ea9e445 Georgi Gerganov ggerganov@gmail.com 2024-11-25T22:28:27+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-25T22:28:59+02:00 server : add more information about error (#10455)
47f931c8f9a26c072d71224bc8013cc66ea9e445 106964e3d266740f571b5aad7b57545b4a901ac9 Georgi Gerganov ggerganov@gmail.com 2024-11-25T21:50:07+02:00 GitHub noreply@github.com 2024-11-25T21:50:07+02:00 server : enable cache_prompt by default (#10501)
106964e3d266740f571b5aad7b57545b4a901ac9 80acb7b430d826a7685326603a07342e9abc1b45 Georgi Gerganov ggerganov@gmail.com 2024-11-25T21:49:31+02:00 GitHub noreply@github.com 2024-11-25T21:49:31+02:00 metal : enable mat-vec kernels for bs <= 4 (#10491)
80acb7b430d826a7685326603a07342e9abc1b45 10bce0450f0c4d80087e06312b9dbbab3e87f16b Shane A shanea@allenai.org 2024-11-25T10:36:09-08:00 GitHub noreply@github.com 2024-11-25T19:36:09+01:00 Rename Olmo1124 to Olmo2 (#10500)
10bce0450f0c4d80087e06312b9dbbab3e87f16b 1f922254f0c984a8fb9fbaa0c390d7ffae49aedb Diego Devesa slarengh@gmail.com 2024-11-25T19:30:06+01:00 GitHub noreply@github.com 2024-11-25T19:30:06+01:00 llama : accept a list of devices to use to offload a model (#10497)
1f922254f0c984a8fb9fbaa0c390d7ffae49aedb a9a678a6b2264e5895533217b0cf8f250534cc58 Johannes Gäßler johannesg@5d6.de 2024-11-25T19:18:37+01:00 GitHub noreply@github.com 2024-11-25T19:18:37+01:00 Github: update issue templates [no ci] (#10489)
a9a678a6b2264e5895533217b0cf8f250534cc58 9ca2e677626fce759d5d95c407c03677b9c87a26 brucepro git@brucepro.net 2024-11-25T08:11:55-08:00 GitHub noreply@github.com 2024-11-25T17:11:55+01:00 Add download chat feature to server chat (#10481)
9ca2e677626fce759d5d95c407c03677b9c87a26 5931c1f233c616083d64e41a228249d58e039aa5 Georgi Gerganov ggerganov@gmail.com 2024-11-25T16:31:38+02:00 GitHub noreply@github.com 2024-11-25T16:31:38+02:00 server : add speculative decoding support (#10455)
5931c1f233c616083d64e41a228249d58e039aa5 f6d12e7df8fe64384f1939976871252e6422a01e Diego Devesa slarengh@gmail.com 2024-11-25T15:13:39+01:00 GitHub noreply@github.com 2024-11-25T15:13:39+01:00 ggml : add support for dynamic loading of backends (#10469)
f6d12e7df8fe64384f1939976871252e6422a01e b756441104ca8384640d6df22ba4ea6dab7ad799 Georgi Gerganov ggerganov@gmail.com 2024-11-25T15:17:32+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-25T15:17:32+02:00 tests : fix compile warning
b756441104ca8384640d6df22ba4ea6dab7ad799 5a8987793f3e7c1fbfa6806bfcd17d578071b6c9 Georgi Gerganov ggerganov@gmail.com 2024-11-25T15:08:04+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-25T15:08:04+02:00 metal : minor code formatting
5a8987793f3e7c1fbfa6806bfcd17d578071b6c9 d9d54e498d38ec99bbc0031022f9c92711e97bbc Neo Zhang Jianyu jianyu.zhang@intel.com 2024-11-25T17:31:10+08:00 GitHub noreply@github.com 2024-11-25T17:31:10+08:00 [SYCL] Fix building Win package for oneAPI 2025.0 update (#10483)
d9d54e498d38ec99bbc0031022f9c92711e97bbc cce5a9007572c6e9fa522296b77571d2e5071357 Georgi Gerganov ggerganov@gmail.com 2024-11-25T09:58:41+02:00 GitHub noreply@github.com 2024-11-25T09:58:41+02:00 speculative : refactor and add a simpler example (#10362)
cce5a9007572c6e9fa522296b77571d2e5071357 dc39012cbaf8752fabecaeb60af78ccdd1dfb73b Georgi Gerganov ggerganov@gmail.com 2024-11-24T18:03:25+02:00 GitHub noreply@github.com 2024-11-24T08:03:25-08:00 flake.lock: Update (#10470)
dc39012cbaf8752fabecaeb60af78ccdd1dfb73b 9336db462c0c34bbe2055413fe4e16442626c38b Diego Devesa slarengh@gmail.com 2024-11-24T16:10:26+01:00 GitHub noreply@github.com 2024-11-24T16:10:26+01:00 llama : fix op mul check with command-r-plus (#10476)
9336db462c0c34bbe2055413fe4e16442626c38b 96fa2c5e2d6cb5319f34c3a7fb0cec05694b22f1 Gabe Goodhart ghart@us.ibm.com 2024-11-24T02:02:34-07:00 GitHub noreply@github.com 2024-11-24T11:02:34+02:00 convert : XLMRoberta Type Vocab Size (#10458)
96fa2c5e2d6cb5319f34c3a7fb0cec05694b22f1 55ed008b2de01592659b9eba068ea01bb2f72160 momonga 146910567+mmngays@users.noreply.github.com 2024-11-24T09:09:22+09:00 GitHub noreply@github.com 2024-11-24T01:09:22+01:00 fix gguf-py: Conversion error when multiple licenses are configured (#9807)
55ed008b2de01592659b9eba068ea01bb2f72160 6dfcfef0787e9902df29f510b63621f60a09a50b Diego Devesa slarengh@gmail.com 2024-11-23T14:41:12+01:00 GitHub noreply@github.com 2024-11-23T14:41:12+01:00 ggml : do not use ARM features not included in the build (#10457)
6dfcfef0787e9902df29f510b63621f60a09a50b 599b3e0cd40432cd1975a8906f3db70bbe53b627 蕭澧邦 45505768+shou692199@users.noreply.github.com 2024-11-22T17:44:08+08:00 GitHub noreply@github.com 2024-11-22T10:44:08+01:00 ci: Update oneAPI runtime dll packaging (#10428)
599b3e0cd40432cd1975a8906f3db70bbe53b627 c18610b4ee29ca056bb4f2d375a4ad1b16f44ef7 Johannes Gäßler johannesg@5d6.de 2024-11-22T08:32:40+01:00 GitHub noreply@github.com 2024-11-22T08:32:40+01:00 GitHub: ask for more info in issue templates (#10426)
c18610b4ee29ca056bb4f2d375a4ad1b16f44ef7 a5e47592b6171ae21f3eaa1aba6fb2b707875063 leo-pony nengjunma@outlook.com 2024-11-22T14:07:20+08:00 GitHub noreply@github.com 2024-11-22T14:07:20+08:00 CANN: Support Ascend310P to accelerate F32 and F16 Model (#10216)
a5e47592b6171ae21f3eaa1aba6fb2b707875063 1bb30bf28cb5a7adf111bc41c935bdaf128397e7 Diego Devesa slarengh@gmail.com 2024-11-21T18:18:50+01:00 GitHub noreply@github.com 2024-11-21T18:18:50+01:00 cuda : optimize argmax (#10441)
1bb30bf28cb5a7adf111bc41c935bdaf128397e7 87a533be57e602f8ca469d14ad15ee851265b655 Georgi Gerganov ggerganov@gmail.com 2024-11-21T10:22:47+02:00 GitHub noreply@github.com 2024-11-21T10:22:47+02:00 llama : handle KV shift for recurrent models (#10402)
87a533be57e602f8ca469d14ad15ee851265b655 59b917282236eadfb82bf1f46a31eb119941da08 Georgi Gerganov ggerganov@gmail.com 2024-11-21T09:22:11+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-21T09:22:11+02:00 sync : ggml
59b917282236eadfb82bf1f46a31eb119941da08 02e4eaf22f229a114054b053a9eff61483653670 slaren slarengh@gmail.com 2024-11-20T13:25:08+01:00 Georgi Gerganov ggerganov@gmail.com 2024-11-21T09:22:05+02:00 ggml/sched : do not skip views in pre-assignments
02e4eaf22f229a114054b053a9eff61483653670 9abe9eeae98b11fa93b82632b264126a010225ff Johannes Gäßler johannesg@5d6.de 2024-11-20T14:56:04+01:00 Georgi Gerganov ggerganov@gmail.com 2024-11-21T09:22:02+02:00 ggml-opt: fix data corruption (ggml/1022)
9abe9eeae98b11fa93b82632b264126a010225ff f95caa79546271722ada703da20ffb1cfcd21fed Jeff Bolz jbolz@nvidia.com 2024-11-20T13:47:36-06:00 GitHub noreply@github.com 2024-11-20T20:47:36+01:00 vulkan: predicate max operation in soft_max shaders/soft_max (#10437)
f95caa79546271722ada703da20ffb1cfcd21fed fab5d30ff6729ff6ff615c41e8c0215d6bc30393 bandoti 141645996+bandoti@users.noreply.github.com 2024-11-20T12:22:19-04:00 GitHub noreply@github.com 2024-11-20T17:22:19+01:00 cmake: add link dependencies to cmake find pkg (#10433)
fab5d30ff6729ff6ff615c41e8c0215d6bc30393 8fd4b7fa29c3061b2e02e897d818dfcbc593430a Diego Devesa slarengh@gmail.com 2024-11-20T12:57:53+01:00 GitHub noreply@github.com 2024-11-20T12:57:53+01:00 llama : add .clang-format file (#10415)
8fd4b7fa29c3061b2e02e897d818dfcbc593430a 1bacb9f62514b520bdf74ed6feb46c80508dad38 Jeff Bolz jbolz@nvidia.com 2024-11-20T01:40:18-06:00 GitHub noreply@github.com 2024-11-20T08:40:18+01:00 vulkan: copy iq4_nl LUT into shared memory (#10409)
1bacb9f62514b520bdf74ed6feb46c80508dad38 ad21c9e1f14d82b8c15ae369a8839019e3d498b4 Jeff Bolz jbolz@nvidia.com 2024-11-20T01:11:00-06:00 GitHub noreply@github.com 2024-11-20T08:11:00+01:00 vulkan: further optimize mul_mat_vec using larger loads (#10387)
ad21c9e1f14d82b8c15ae369a8839019e3d498b4 3952a221af54b8a6549bc2bd4a7363ef7ad3081e Neo Zhang Jianyu jianyu.zhang@intel.com 2024-11-20T13:54:25+08:00 GitHub noreply@github.com 2024-11-20T13:54:25+08:00 update rel to 4040 (#10395)
3952a221af54b8a6549bc2bd4a7363ef7ad3081e 42ae10bbcd7b56f29a302c86796542a6dadf46c9 Anthony Van de Gejuchte anthonyvdgent@gmail.com 2024-11-19T23:18:17+01:00 GitHub noreply@github.com 2024-11-19T23:18:17+01:00 Fix missing file renames in Makefile due to changes in commit ae8de6d50a (#10413)
42ae10bbcd7b56f29a302c86796542a6dadf46c9 9fe0fb062630728e3c21b5839e3bce87bff2440a haopeng 657407891@qq.com 2024-11-20T04:10:31+08:00 GitHub noreply@github.com 2024-11-19T21:10:31+01:00 add cmake rvv support (#10411)
9fe0fb062630728e3c21b5839e3bce87bff2440a 611fabd7922050e1e99bd276d3544527cd46047b Georgi Gerganov ggerganov@gmail.com 2024-11-19T19:15:50+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-19T20:03:21+02:00 sync : ggml
611fabd7922050e1e99bd276d3544527cd46047b 12b0ad953a59563ea8d973708760d747321d8432 Plamen Minev pacominev@gmail.com 2024-11-18T15:02:27+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-19T20:03:21+02:00 metal : fox offset integer overflows in im2col (ggml/1015)
12b0ad953a59563ea8d973708760d747321d8432 342397dc7edb311e0373205134d0d3a928b891b3 PAB pierreantoine.bannier@gmail.com 2024-11-18T10:02:49+01:00 Georgi Gerganov ggerganov@gmail.com 2024-11-19T20:03:21+02:00 metal : add `GGML_UNARY_OP_ELU` kernel (ggml/1018)
342397dc7edb311e0373205134d0d3a928b891b3 2a11b6b0946c1abab2ab150725610e5ee736b3af 蕭澧邦 45505768+shou692199@users.noreply.github.com 2024-11-20T01:42:00+08:00 GitHub noreply@github.com 2024-11-19T18:42:00+01:00 cmake: force MSVC compiler charset to utf-8 (#9989)
2a11b6b0946c1abab2ab150725610e5ee736b3af 3ee6382d48b07b31e64983969c16019490e19740 bandoti 141645996+bandoti@users.noreply.github.com 2024-11-19T12:10:30-04:00 GitHub noreply@github.com 2024-11-19T17:10:30+01:00 Add required ggml-base and backend libs to cmake pkg (#10407)
3ee6382d48b07b31e64983969c16019490e19740 8e752a777b272606f22cb741b03e062de4ddb8fe Diego Devesa slarengh@gmail.com 2024-11-19T14:29:38+01:00 GitHub noreply@github.com 2024-11-19T14:29:38+01:00 cuda : fix CUDA_FLAGS not being applied (#10403)
8e752a777b272606f22cb741b03e062de4ddb8fe a88ad007de3eb5d92cb538fd269ff94c4bf0c8d2 Georgi Gerganov ggerganov@gmail.com 2024-11-19T13:29:26+02:00 GitHub noreply@github.com 2024-11-19T13:29:26+02:00 llama : add check for KV cache shifts (#10401)
a88ad007de3eb5d92cb538fd269ff94c4bf0c8d2 2a1507c1629975d9d20a503d6a14f44eff292c25 Shane A shanea@allenai.org 2024-11-19T01:04:08-08:00 GitHub noreply@github.com 2024-11-19T11:04:08+02:00 llama : add OLMo November 2024 support (#10394)
2a1507c1629975d9d20a503d6a14f44eff292c25 b3e585988fc65d3a8083c6d94dfc0629f9ce226d Romain Biessy romain.biessy@codeplay.com 2024-11-19T09:02:23+01:00 GitHub noreply@github.com 2024-11-19T08:02:23Z sycl : Add option to set the SYCL architecture for all targets (#10266)
b3e585988fc65d3a8083c6d94dfc0629f9ce226d 557924f22237c76387a39c4db5abae154d57e754 Jeff Bolz jbolz@nvidia.com 2024-11-19T01:25:17-06:00 GitHub noreply@github.com 2024-11-19T08:25:17+01:00 vulkan: Optimize soft_max (#10301)
557924f22237c76387a39c4db5abae154d57e754 d3481e631661b5e9517f78908cdd58cee63c4903 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-11-19T00:50:04Z GitHub noreply@github.com 2024-11-19T08:50:04+08:00 sycl: Revert MUL_MAT_OP support changes (#10385)
d3481e631661b5e9517f78908cdd58cee63c4903 531cb1c233800e6acb021dc56d69595e314db072 Diego Devesa slarengh@gmail.com 2024-11-18T18:43:40+01:00 GitHub noreply@github.com 2024-11-18T18:43:40+01:00 cuda : only use native when supported by cmake (#10389)
531cb1c233800e6acb021dc56d69595e314db072 f139d2ea611c5604395c95160d3c53f7c4eaf220 bandoti 141645996+bandoti@users.noreply.github.com 2024-11-18T11:23:58-04:00 GitHub noreply@github.com 2024-11-18T16:23:58+01:00 Skip searching root path for cross-compile builds (#10383)
f139d2ea611c5604395c95160d3c53f7c4eaf220 2eb76b2a5e4ea395b971a419c95b473ab6f253e4 Jeff Bolz jbolz@nvidia.com 2024-11-18T08:28:42-06:00 GitHub noreply@github.com 2024-11-18T08:28:42-06:00 vulkan: remove use of null initializer (#10372)
2eb76b2a5e4ea395b971a419c95b473ab6f253e4 9b75f03cd2ec9cc482084049d87a0f08f9f01517 Georgi Gerganov ggerganov@gmail.com 2024-11-18T16:08:20+02:00 GitHub noreply@github.com 2024-11-18T06:08:20-08:00 flake.lock: Update (#10346)
9b75f03cd2ec9cc482084049d87a0f08f9f01517 75207b3a887f91f813de1eb6e9fd135d3cb2b8c6 0cc4m picard12@live.de 2024-11-18T11:02:43+01:00 GitHub noreply@github.com 2024-11-18T11:02:43+01:00 Vulkan: Fix device info output format specifiers (#10366)
75207b3a887f91f813de1eb6e9fd135d3cb2b8c6 76e9e58b7847112848aa1f40b65d1cbcd6d5f5a3 Johannes Gäßler johannesg@5d6.de 2024-11-18T00:21:53+01:00 GitHub noreply@github.com 2024-11-18T00:21:53+01:00 docker: use GGML_NATIVE=OFF (#10368)
76e9e58b7847112848aa1f40b65d1cbcd6d5f5a3 ce2e59ba107cf71ed566040ff20a15d1c58e09c2 Johannes Gäßler johannesg@5d6.de 2024-11-17T23:20:42+01:00 GitHub noreply@github.com 2024-11-17T23:20:42+01:00 CUDA: fix MMV kernel being used for FP16 src1 (#10357)
ce2e59ba107cf71ed566040ff20a15d1c58e09c2 be5caccef945546ee9fd25a151330a88d785faf9 Johannes Gäßler johannesg@5d6.de 2024-11-17T12:59:38+01:00 GitHub noreply@github.com 2024-11-17T12:59:38+01:00 CMake: fix typo in comment [no ci] (#10360)
be5caccef945546ee9fd25a151330a88d785faf9 20a780c7b64c38071fe70ad23e16e80c23c0147b Diego Devesa slarengh@gmail.com 2024-11-17T12:25:45+01:00 GitHub noreply@github.com 2024-11-17T12:25:45+01:00 llama : only use default buffer types for the KV cache (#10358)
20a780c7b64c38071fe70ad23e16e80c23c0147b cf32a9b93ad859ea592c31785b6bd3b4b2121463 Georgi Gerganov ggerganov@gmail.com 2024-11-17T13:12:22+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T13:12:22+02:00 gitignore : ignore local run scripts [no ci]
cf32a9b93ad859ea592c31785b6bd3b4b2121463 a43178299c2f74f14bcfc659bfd9fd32d931d1f4 Georgi Gerganov ggerganov@gmail.com 2024-11-17T11:23:01+02:00 GitHub noreply@github.com 2024-11-17T11:23:01+02:00 metal : refactor kernel args into structs (#10238)
a43178299c2f74f14bcfc659bfd9fd32d931d1f4 c3ea58aca406911eb4d409cdbfc76683393442b6 FirstTimeEZ 179362031+FirstTimeEZ@users.noreply.github.com 2024-11-17T21:39:22+13:00 GitHub noreply@github.com 2024-11-17T10:39:22+02:00 ggml : fix undefined reference to 'getcpu' (#10354)
c3ea58aca406911eb4d409cdbfc76683393442b6 467576b6cc7d2b9220f55bc635aa51469cf26fb5 Johannes Gäßler johannesg@5d6.de 2024-11-17T09:09:55+01:00 GitHub noreply@github.com 2024-11-17T09:09:55+01:00 CUDA: remove DMMV, consolidate F16 mult mat vec (#10318)
467576b6cc7d2b9220f55bc635aa51469cf26fb5 eda7e1d4f54711de1c9b40502d6c88bbc217da60 Johannes Gäßler johannesg@5d6.de 2024-11-17T09:06:34+01:00 GitHub noreply@github.com 2024-11-17T09:06:34+01:00 CMake: default to -arch=native for CUDA build (#10320)
eda7e1d4f54711de1c9b40502d6c88bbc217da60 24203e9dd7355a4a10bc32d959fd0148d37bf666 Diego Devesa slarengh@gmail.com 2024-11-17T07:31:17+01:00 GitHub noreply@github.com 2024-11-17T08:31:17+02:00 ggml : fix possible buffer use after free in sched reserve (#9930)
24203e9dd7355a4a10bc32d959fd0148d37bf666 5d9e59979c2fd91c99d03c23e8df9c50c9d55a85 Georgi Gerganov ggerganov@gmail.com 2024-11-16T23:40:39+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T08:30:29+02:00 ggml : inttypes.h -> cinttypes (#0)
5d9e59979c2fd91c99d03c23e8df9c50c9d55a85 a4200cafadebb7576a9a3905039858f5e73ce4cc Georgi Gerganov ggerganov@gmail.com 2024-11-16T21:38:01+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T08:30:29+02:00 ggml : adapt AMX to tensor->grad removal (#0)
a4200cafadebb7576a9a3905039858f5e73ce4cc 84274a10c399d843cff39feb2b86dc8224f613d8 Georgi Gerganov ggerganov@gmail.com 2024-11-16T21:35:31+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T08:30:29+02:00 make : add ggml-opt (#0)
84274a10c399d843cff39feb2b86dc8224f613d8 68fcb4759c0ca2874b59d9c1e6a35ceca1cc04ce Georgi Gerganov ggerganov@gmail.com 2024-11-16T21:34:03+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T08:30:29+02:00 tests : remove test-grad0
68fcb4759c0ca2874b59d9c1e6a35ceca1cc04ce 8a43e940ab0daaff198809bf9277289994ec62f5 Georgi Gerganov ggerganov@gmail.com 2024-11-16T21:32:41+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T08:30:29+02:00 ggml : fix compile warnings (#0)
8a43e940ab0daaff198809bf9277289994ec62f5 5c9a8b22b10132db620529435e3cfa49304b65cc Johannes Gäßler johannesg@5d6.de 2024-11-16T22:17:59+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T08:30:29+02:00 ggml: new optimization interface (ggml/988)
5c9a8b22b10132db620529435e3cfa49304b65cc 0fff7fd79818980763a601660f25b01a0cf4b87a Georgi Gerganov ggerganov@gmail.com 2024-11-16T22:16:04+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-17T08:30:29+02:00 scripts : update sync
0fff7fd79818980763a601660f25b01a0cf4b87a 4e54be0ec6cb5cd6ed56e52e927b80b2796ec844 FirstTimeEZ 179362031+FirstTimeEZ@users.noreply.github.com 2024-11-17T12:29:18+13:00 GitHub noreply@github.com 2024-11-17T00:29:18+01:00 docs : vulkan build instructions to use git bash mingw64 (#10303)
4e54be0ec6cb5cd6ed56e52e927b80b2796ec844 db4cfd5dbc31c90f0d5c413a2e182d068b8ee308 Johannes Gäßler johannesg@5d6.de 2024-11-16T23:00:41+01:00 GitHub noreply@github.com 2024-11-16T23:00:41+01:00 llama/ex: remove --logdir argument (#10339)
db4cfd5dbc31c90f0d5c413a2e182d068b8ee308 8ee0d09ae6928d0501765cfc4e430b9236730caf Georgi Gerganov ggerganov@gmail.com 2024-11-16T17:58:56+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-16T20:36:26+02:00 llamafile : fix include path (#0)
8ee0d09ae6928d0501765cfc4e430b9236730caf bcdb7a23862b61aa307fc462fadfe1e2e653d010 Georgi Gerganov ggerganov@gmail.com 2024-11-16T17:58:32+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-16T20:36:26+02:00 make : auto-determine dependencies (#0)
bcdb7a23862b61aa307fc462fadfe1e2e653d010 f245cc28d4eb900efad0bc740145f58d713c6e4f MaggotHATE clay1326@gmail.com 2024-11-16T18:26:54+05:00 GitHub noreply@github.com 2024-11-16T14:26:54+01:00 server: (web UI) Add samplers sequence customization (#10255)
f245cc28d4eb900efad0bc740145f58d713c6e4f 772703c8fffdd83d2e28f60119e83525f1189412 Georgi Gerganov ggerganov@gmail.com 2024-11-16T10:32:50+02:00 GitHub noreply@github.com 2024-11-16T10:32:50+02:00 scripts : fix missing key in compare-llama-bench.py (#10332)
772703c8fffdd83d2e28f60119e83525f1189412 dd3a6ce9f84d21ba05fe98af9f983bdea0398e6c Jeff Bolz jbolz@nvidia.com 2024-11-16T00:26:57-06:00 GitHub noreply@github.com 2024-11-16T07:26:57+01:00 vulkan: Optimize some mat-vec mul quant shaders (#10296)
dd3a6ce9f84d21ba05fe98af9f983bdea0398e6c 1e58ee1318429a3e97aa66f3034cdfd65ffc6c34 FirstTimeEZ 179362031+FirstTimeEZ@users.noreply.github.com 2024-11-16T14:59:33+13:00 GitHub noreply@github.com 2024-11-16T02:59:33+01:00 vulkan : add cmake preset debug/release (#10306)
1e58ee1318429a3e97aa66f3034cdfd65ffc6c34 89e4caaaf081f4712af61a3e08cb67b406c02b80 Dan Johansson dan.johansson@arm.com 2024-11-16T01:53:37+01:00 GitHub noreply@github.com 2024-11-16T01:53:37+01:00 ggml : optimize Q4_0 into Q4_0_X_Y repack (#10324)
89e4caaaf081f4712af61a3e08cb67b406c02b80 74d73dc85cc2057446bf63cc37ff649ae7cebd80 FirstTimeEZ 179362031+FirstTimeEZ@users.noreply.github.com 2024-11-16T13:42:13+13:00 GitHub noreply@github.com 2024-11-16T01:42:13+01:00 llama : save number of parameters and the size in llama_model (#10286)
74d73dc85cc2057446bf63cc37ff649ae7cebd80 4047be74da398acb8717a4d21b77b929ad7ed4f7 Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-11-16T02:57:00+05:30 GitHub noreply@github.com 2024-11-15T22:27:00+01:00 Make updates to fix issues with clang-cl builds while using AVX512 flags (#10314)
4047be74da398acb8717a4d21b77b929ad7ed4f7 883d206fbd2c5b2b9b589a9328503b9005e146c9 Johannes Gäßler johannesg@5d6.de 2024-11-15T21:19:03+01:00 GitHub noreply@github.com 2024-11-15T21:19:03+01:00 scripts: update compare-llama-bench.py (#10319)
883d206fbd2c5b2b9b589a9328503b9005e146c9 09ecbcb596ed8fa97d503d7440f0b3eff872e8f1 slaren slarengh@gmail.com 2024-11-15T20:20:54+01:00 Georgi Gerganov ggerganov@gmail.com 2024-11-15T21:45:32+02:00 ggml : fix some build issues
09ecbcb596ed8fa97d503d7440f0b3eff872e8f1 3225008973579cc6a784890c237e1bfc9de41819 Georgi Gerganov ggerganov@gmail.com 2024-11-15T15:35:22+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-15T15:44:06+02:00 cmake : fix ppc64 check (whisper/0)
3225008973579cc6a784890c237e1bfc9de41819 cbf5541a82952bcd7c4fceb55f5e332cafbf1720 thewh1teagle 61390950+thewh1teagle@users.noreply.github.com 2024-11-15T15:33:53+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-15T15:44:06+02:00 ggml : vulkan logs (whisper/2547)
cbf5541a82952bcd7c4fceb55f5e332cafbf1720 18429220bdb344da1bc7df9bc580c7b41b3cd57b Georgi Gerganov ggerganov@gmail.com 2024-11-15T15:31:16+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-15T15:44:06+02:00 sync : ggml
18429220bdb344da1bc7df9bc580c7b41b3cd57b f0204a0ec70d50ca60e07bc0096ec1d6508ab0c7 Eve 139727413+netrunnereve@users.noreply.github.com 2024-11-15T11:47:58Z GitHub noreply@github.com 2024-11-15T12:47:58+01:00 AVX BF16 and single scale quant optimizations (#10212)
f0204a0ec70d50ca60e07bc0096ec1d6508ab0c7 57f8355b29a8c7dfcd1fb6094758ad85644f8535 R0CKSTAR xiaodong.ye@mthreads.com 2024-11-15T19:47:25+08:00 GitHub noreply@github.com 2024-11-15T12:47:25+01:00 ci: build test musa with cmake (#10298)
57f8355b29a8c7dfcd1fb6094758ad85644f8535 9901068ac78838745e604fffb4601d315a610456 Romain Biessy romain.biessy@codeplay.com 2024-11-15T12:10:45+01:00 GitHub noreply@github.com 2024-11-15T13:10:45+02:00 sycl: Update Intel docker images to use DPC++ 2025.0 (#10305)
9901068ac78838745e604fffb4601d315a610456 231f9360d94446cd083b6b116f63991b1328c484 Xuan Son Nguyen thichthat@gmail.com 2024-11-15T05:48:49-04:00 GitHub noreply@github.com 2024-11-15T10:48:49+01:00 server : (web UI) add copy button for code block, fix api key (#10242)
231f9360d94446cd083b6b116f63991b1328c484 4802ad350b8e19cbc7a77269b4494c896f6e0896 Chenguang Li 87689256+noemotiovon@users.noreply.github.com 2024-11-15T15:09:35+08:00 GitHub noreply@github.com 2024-11-15T15:09:35+08:00 cann: dockerfile and doc adjustment (#10302)
4802ad350b8e19cbc7a77269b4494c896f6e0896 5a54af4d4f588f109f31e456483fdf77096399d9 Georgi Gerganov ggerganov@gmail.com 2024-11-15T08:38:43+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-15T08:38:43+02:00 scripts : fix regex in sync [no ci]
5a54af4d4f588f109f31e456483fdf77096399d9 1607a5e5b08f4e55f118af3d7de325949d8f1835 Romain Biessy romain.biessy@codeplay.com 2024-11-15T04:09:12+01:00 GitHub noreply@github.com 2024-11-15T11:09:12+08:00 sycl: Use syclcompat::dp4a (#10267)
1607a5e5b08f4e55f118af3d7de325949d8f1835 ae8de6d50a09d49545e0afab2e50cc4acfb280e2 Charles Xu charles.xu@arm.com 2024-11-15T01:28:50+01:00 GitHub noreply@github.com 2024-11-15T01:28:50+01:00 backend cpu: add online flow for aarch64 Q4_0 GEMV/GEMM kernels (#9921)
ae8de6d50a09d49545e0afab2e50cc4acfb280e2 4a8ccb37ad9c9027cbcfd5548c19cdffe48d5197 Diego Devesa slarengh@gmail.com 2024-11-14T18:04:35+01:00 GitHub noreply@github.com 2024-11-14T18:04:35+01:00 ggml : build backends as libraries (#10256)
4a8ccb37ad9c9027cbcfd5548c19cdffe48d5197 2a82891a853db908679f7b24b04586e6f6393fe0 Johannes Gäßler johannesg@5d6.de 2024-11-14T13:00:15+01:00 GitHub noreply@github.com 2024-11-14T13:00:15+01:00 CUDA: no -sm row for very small matrices (#10185)
2a82891a853db908679f7b24b04586e6f6393fe0 af148c9386da825a60c7038549c121c35ca56b50 Georgi Gerganov ggerganov@gmail.com 2024-11-14T11:44:15+02:00 GitHub noreply@github.com 2024-11-14T11:44:15+02:00 speculative : fix out-of-bounds access (#10289)
af148c9386da825a60c7038549c121c35ca56b50 66798e42fbe636f1cb6236e4bc30939d23ef7c25 Jeff Bolz jbolz@nvidia.com 2024-11-13T23:22:55-06:00 GitHub noreply@github.com 2024-11-14T06:22:55+01:00 vulkan: Optimize binary ops (#10270)
66798e42fbe636f1cb6236e4bc30939d23ef7c25 fb4a0ec0833c71cff5a1a367ba375447ce6106eb Jeff Bolz jbolz@nvidia.com 2024-11-13T14:59:47-06:00 GitHub noreply@github.com 2024-11-13T21:59:47+01:00 vulkan: Use macros to make the mat mul pipeline creation more concise (#10259)
fb4a0ec0833c71cff5a1a367ba375447ce6106eb 5ea926dad7f62ebccff7b24784bd1e01a06d13ae Michael Podvitskiy podvitskiymichael@gmail.com 2024-11-13T20:00:35+02:00 GitHub noreply@github.com 2024-11-13T20:00:35+02:00 llama : propagate the results of `graph_compute` (#9525)
5ea926dad7f62ebccff7b24784bd1e01a06d13ae 1ee9eea094fe5846c7d8d770aa7caa749d246b23 Georgi Gerganov ggerganov@gmail.com 2024-11-13T18:11:54+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-13T18:11:54+02:00 sync : ggml
1ee9eea094fe5846c7d8d770aa7caa749d246b23 ff7fb670d0a62897c5662536aeb53422c549fbe8 Small Grass Forest zixuanxcl@gmail.com 2024-11-13T19:17:10+08:00 GitHub noreply@github.com 2024-11-13T13:17:10+02:00 docs : update bindings list (#10261)
ff7fb670d0a62897c5662536aeb53422c549fbe8 0e712a5acbbdd1593e5aeb86d4f6b896a11b438c Alexey Parfenov zxed@alkatrazstudio.net 2024-11-13T11:16:30Z GitHub noreply@github.com 2024-11-13T13:16:30+02:00 server : add missing docs (#10269)
0e712a5acbbdd1593e5aeb86d4f6b896a11b438c a0ec17b32ec6077f5ca22fe833ebdc9b86795a4d Jhen-Jie Hong iainst0409@gmail.com 2024-11-13T19:15:23+08:00 GitHub noreply@github.com 2024-11-13T13:15:23+02:00 server : fix incorrect res in validate_model_chat_template (#10272)
a0ec17b32ec6077f5ca22fe833ebdc9b86795a4d 2e82ffa4af29f87e7d3d6dff8060a2a79613b72f Brian mofosyne@gmail.com 2024-11-13T21:10:38+11:00 GitHub noreply@github.com 2024-11-13T21:10:38+11:00 metadata: Detailed Dataset Authorship Metadata (#8875)
2e82ffa4af29f87e7d3d6dff8060a2a79613b72f 80dd7ff22fd050fed58b552cc8001aaf968b7ebf Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-11-13T09:40:57Z GitHub noreply@github.com 2024-11-13T09:40:57Z sycl : Fixes to broken builds and test-backend-ops (#10257)
80dd7ff22fd050fed58b552cc8001aaf968b7ebf 54ef9cfc726a799e6f454ac22c4815d037716eda Jeff Bolz jbolz@nvidia.com 2024-11-13T00:58:57-06:00 GitHub noreply@github.com 2024-11-13T07:58:57+01:00 vulkan: Optimize contiguous copies (#10254)
54ef9cfc726a799e6f454ac22c4815d037716eda b0cefea58a20020754b7431e3c725625274372a5 Jeff Bolz jbolz@nvidia.com 2024-11-11T11:13:51-06:00 GitHub noreply@github.com 2024-11-11T18:13:51+01:00 vulkan: Throttle the number of shader compiles during the build step. (#10222)
b0cefea58a20020754b7431e3c725625274372a5 b141e5f6efbab3a00633df88c4f9425bfe8b78ab Georgi Gerganov ggerganov@gmail.com 2024-11-11T08:39:13+02:00 GitHub noreply@github.com 2024-11-11T08:39:13+02:00 metal : more precise Q*K in FA vec kernel (#10247)
b141e5f6efbab3a00633df88c4f9425bfe8b78ab 4b3a9212b602be3d4e2e3ca26efd796cef13c55e Georgi Gerganov ggerganov@gmail.com 2024-11-11T08:38:43+02:00 GitHub noreply@github.com 2024-11-11T08:38:43+02:00 server : enable KV cache defrag by default (#10233)
4b3a9212b602be3d4e2e3ca26efd796cef13c55e 505f33274d60676320216b662a97672a76ec600e Georgi Gerganov ggerganov@gmail.com 2024-11-10T21:45:25+02:00 GitHub noreply@github.com 2024-11-10T11:45:25-08:00 flake.lock: Update (#10243)
505f33274d60676320216b662a97672a76ec600e 160687b3ed002eee83a04de83a9cd752f928ced1 MaggotHATE clay1326@gmail.com 2024-11-11T00:42:25+05:00 GitHub noreply@github.com 2024-11-10T15:42:25-04:00 server : (web UI) Add back sampler settings (#10239)
160687b3ed002eee83a04de83a9cd752f928ced1 6423c65aa8be1b98f990cf207422505ac5a441a1 Jeff Bolz jbolz@nvidia.com 2024-11-10T05:37:56-06:00 GitHub noreply@github.com 2024-11-10T12:37:56+01:00 vulkan: Fix newly added tests for permuted mul_mat and 1D im2col (#10226)
6423c65aa8be1b98f990cf207422505ac5a441a1 39a334a9aaf2000f93a899d9f43d889e460640ee Georgi Gerganov ggerganov@gmail.com 2024-11-09T11:53:13+02:00 GitHub noreply@github.com 2024-11-09T11:53:13+02:00 metal : reorder write loop in mul mat kernel + style (#10231)
39a334a9aaf2000f93a899d9f43d889e460640ee bb38cdd8baf37de1fadab3e867c6ba4ae452eff6 Georgi Gerganov ggerganov@gmail.com 2024-11-09T11:53:02+02:00 GitHub noreply@github.com 2024-11-09T11:53:02+02:00 metal : fix build and some more comments (#10229)
bb38cdd8baf37de1fadab3e867c6ba4ae452eff6 f018acba22095b8995bf6c5ef815b16a3ce4cf1b Georgi Gerganov ggerganov@gmail.com 2024-11-09T11:52:45+02:00 GitHub noreply@github.com 2024-11-09T11:52:45+02:00 metal : fix F32 accumulation in FA vec kernel (#10232)
f018acba22095b8995bf6c5ef815b16a3ce4cf1b 46323fa9efd5e6c8aeef8d6eb6c332ee0d95eb13 Georgi Gerganov ggerganov@gmail.com 2024-11-09T11:26:34+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-09T11:26:34+02:00 llama : fix Qwen model type strings
46323fa9efd5e6c8aeef8d6eb6c332ee0d95eb13 5b359bb1e3585de45bec79fd6c18934897662cdf Georgi Gerganov ggerganov@gmail.com 2024-11-09T11:21:49+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-09T11:21:49+02:00 metal : hide debug messages from normal log
5b359bb1e3585de45bec79fd6c18934897662cdf e89213492d3e01705739789733f0f2d250b4c449 SXX sxx1136965276@gmail.com 2024-11-09T15:35:46+08:00 GitHub noreply@github.com 2024-11-09T08:35:46+01:00 ggml: fix zero division in ‘dne’ calculation in CUDA COUNT_EQUAL operator when ‘ne’ is small (#10213)
e89213492d3e01705739789733f0f2d250b4c449 8fc393f246c550d2481e53323a47644a94e8d01f amritahs-ibm amritahs@linux.vnet.ibm.com 2024-11-09T12:47:50+05:30 GitHub noreply@github.com 2024-11-09T09:17:50+02:00 ggml : optimize llamafile cpu matrix multiplication for ppc64le (#10156)
8fc393f246c550d2481e53323a47644a94e8d01f ec450d3bbf9fdb3cd06b27c00c684fd1861cb0cf haopeng 657407891@qq.com 2024-11-09T15:06:54+08:00 GitHub noreply@github.com 2024-11-09T09:06:54+02:00 scripts : fix pattern and get n_tokens in one go (#10221)
ec450d3bbf9fdb3cd06b27c00c684fd1861cb0cf 695ad752b2631af84ba321177656705b30c6e401 Georgi Gerganov ggerganov@gmail.com 2024-11-08T21:59:46+02:00 GitHub noreply@github.com 2024-11-08T21:59:46+02:00 metal : opt-in compile flag for BF16 (#10218)
695ad752b2631af84ba321177656705b30c6e401 841f27abdbbcecc9daac14dc540ba6202e4ffe40 Georgi Gerganov ggerganov@gmail.com 2024-11-08T18:37:41+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-08T18:37:41+02:00 metal : improve clarity (minor) (#10171)
841f27abdbbcecc9daac14dc540ba6202e4ffe40 d05b3127bd30515955aa4ee2bacdb68ebafe88f4 Georgi Gerganov ggerganov@gmail.com 2024-11-08T13:47:22+02:00 GitHub noreply@github.com 2024-11-08T13:47:22+02:00 metal : optimize FA kernels (#10171)
d05b3127bd30515955aa4ee2bacdb68ebafe88f4 76c6e7f10551960e4ec9e14e0535b72081f1c7ad Jhen-Jie Hong iainst0409@gmail.com 2024-11-08T17:34:06+08:00 GitHub noreply@github.com 2024-11-08T11:34:06+02:00 swift : exclude ggml-metal-embed.metal (#10211)
76c6e7f10551960e4ec9e14e0535b72081f1c7ad a71d81cf8c1afb26b166f897c94ee1581f9fac7d Xuan Son Nguyen thichthat@gmail.com 2024-11-07T18:44:38-04:00 GitHub noreply@github.com 2024-11-07T18:44:38-04:00 server : minor UI fix (#10207)
a71d81cf8c1afb26b166f897c94ee1581f9fac7d eec4d71737b32f312e0082b671629a0368e1a20d Xuan Son Nguyen thichthat@gmail.com 2024-11-07T17:31:10-04:00 GitHub noreply@github.com 2024-11-07T17:31:10-04:00 server : revamp chat UI with vuejs and daisyui (#10175)
eec4d71737b32f312e0082b671629a0368e1a20d 3b08828674f561c78af182d47fc0636fc3ccd1e9 Georgi Gerganov ggerganov@gmail.com 2024-11-07T23:11:36+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-07T23:11:36+02:00 scripts : add amx to sync-ggml.sh [no ci]
3b08828674f561c78af182d47fc0636fc3ccd1e9 a2c6fd747c77fe183e2f556a4a2f1fb0a0be4c7b Georgi Gerganov ggerganov@gmail.com 2024-11-07T23:08:24+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-07T23:08:24+02:00 sync : ggml
a2c6fd747c77fe183e2f556a4a2f1fb0a0be4c7b 97404c4a0374cac45c8c34a32d13819de1dd023d Georgi Gerganov ggerganov@gmail.com 2024-11-07T23:07:55+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-07T23:07:55+02:00 scripts : sync update
97404c4a0374cac45c8c34a32d13819de1dd023d 60e17ce23c2740369af6304113a2dfa0454eaf26 Diego Devesa slarengh@gmail.com 2024-11-07T18:16:08+01:00 GitHub noreply@github.com 2024-11-07T18:16:08+01:00 ggml : add ggml-cpu.h to the public headers (#10204)
60e17ce23c2740369af6304113a2dfa0454eaf26 5107e8cea35be46a27cfc940e6841c0cf81c0525 Faisal Zaghloul quic_fzaghlou@quicinc.com 2024-11-07T11:46:12-05:00 GitHub noreply@github.com 2024-11-07T08:46:12-08:00 Remove identical wte/etw logic for jais (#10203)
5107e8cea35be46a27cfc940e6841c0cf81c0525 2319126a70b541f8670225a04a38202bbdccbedb wwoodsTM 104587230+wwoodsTM@users.noreply.github.com 2024-11-07T08:20:25-07:00 GitHub noreply@github.com 2024-11-07T16:20:25+01:00 DRY: Fixes clone functionality (#10192)
2319126a70b541f8670225a04a38202bbdccbedb 3bcd40b3c593d14261fb2abfabad3c0fb5b9e318 snadampal 87143774+snadampal@users.noreply.github.com 2024-11-07T02:02:08-06:00 GitHub noreply@github.com 2024-11-07T09:02:08+01:00 fix q4_0_8_8 format for corrupted tokens issue (#10198)
3bcd40b3c593d14261fb2abfabad3c0fb5b9e318 5c333e014059122245c318e7ed4ec27d1085573c Zhiyuan Li lizhiyuan@uniartisan.com 2024-11-07T18:19:10+11:00 GitHub noreply@github.com 2024-11-07T15:19:10+08:00 Optimize RWKV6 Operator Naming and Implement Multi-core CPU/ SYCL Acceleration (#10133)
5c333e014059122245c318e7ed4ec27d1085573c b11f9ba9b8ce319f04b88afe40d264e6b7f4ba46 Georgi Gerganov ggerganov@gmail.com 2024-11-06T19:53:51+02:00 GitHub noreply@github.com 2024-11-06T19:53:51+02:00 metal : add BF16 support (#8439)
b11f9ba9b8ce319f04b88afe40d264e6b7f4ba46 94d8cb8be13b7c4d04eeca5a2b956b9148e6f222 Georgi Gerganov ggerganov@gmail.com 2024-11-06T13:29:01+02:00 GitHub noreply@github.com 2024-11-06T13:29:01+02:00 server : remove hack for extra parallel slot (#10187)
94d8cb8be13b7c4d04eeca5a2b956b9148e6f222 1dc04b2deed2d2f2ae3aff9b14ae29674dee1fb8 Diego Devesa slarengh@gmail.com 2024-11-06T12:10:07+01:00 GitHub noreply@github.com 2024-11-06T12:10:07+01:00 metal : fix from ptr buffer name (#10189)
1dc04b2deed2d2f2ae3aff9b14ae29674dee1fb8 a1eaf6a9600bb1608753420ba886a3b0a208ffc0 Georgi Gerganov ggerganov@gmail.com 2024-11-06T11:20:10+02:00 GitHub noreply@github.com 2024-11-06T11:20:10+02:00 ggml : adjust is_first_call init value (#10193)
a1eaf6a9600bb1608753420ba886a3b0a208ffc0 b8deef0ec0af5febac1d2cfd9119ff330ed0b762 Georgi Gerganov ggerganov@gmail.com 2024-11-06T10:24:23+02:00 GitHub noreply@github.com 2024-11-06T10:24:23+02:00 metal : add quantized FA support (#10149)
b8deef0ec0af5febac1d2cfd9119ff330ed0b762 a9e8a9a0306a8093eef93b0022d9f45510490072 Gabe Goodhart ghart@us.ibm.com 2024-11-05T05:23:04-07:00 GitHub noreply@github.com 2024-11-05T14:23:04+02:00 llama : add <|tool_call|> formatting to Granite template (#10177)
a9e8a9a0306a8093eef93b0022d9f45510490072 340736477651095a98a3b10e19b038ec62593a1d Diego Devesa slarengh@gmail.com 2024-11-04T23:17:01+01:00 GitHub noreply@github.com 2024-11-04T23:17:01+01:00 ggml : fix arch check in bf16_to_fp32 (#10164)
340736477651095a98a3b10e19b038ec62593a1d d5a409e57fe8bd24fef597ab8a31110d390a6392 Eve 139727413+netrunnereve@users.noreply.github.com 2024-11-04T22:06:31Z GitHub noreply@github.com 2024-11-04T23:06:31+01:00 Q6_K AVX improvements (#10118)
d5a409e57fe8bd24fef597ab8a31110d390a6392 401558b7ba7a08175c153cd3607230f63c8a528e Diego Devesa slarengh@gmail.com 2024-11-04T20:06:58+01:00 GitHub noreply@github.com 2024-11-04T20:06:58+01:00 ggml : fix gelu tables initialization (#10172)
401558b7ba7a08175c153cd3607230f63c8a528e 9e0ecfb697d297355e43c20559d29bcc71beb0c3 Diego Devesa slarengh@gmail.com 2024-11-04T17:34:08+01:00 GitHub noreply@github.com 2024-11-04T17:34:08+01:00 ggml : fix q4xx mat mul, increase ggml_aligned_malloc alignment (#10167)
9e0ecfb697d297355e43c20559d29bcc71beb0c3 6a066b9978533e2ab9890b7f4f8c0262d91798b3 Xuan Son Nguyen thichthat@gmail.com 2024-11-04T16:33:29+01:00 GitHub noreply@github.com 2024-11-04T16:33:29+01:00 server : clarify /slots endpoint, add is_processing (#10162)
6a066b9978533e2ab9890b7f4f8c0262d91798b3 ea02c753ebf9342114cb173f10b3ffc2af1e7d04 snadampal 87143774+snadampal@users.noreply.github.com 2024-11-04T09:08:33-06:00 GitHub noreply@github.com 2024-11-04T16:08:33+01:00 fix build break on arm64 linux (#10166)
ea02c753ebf9342114cb173f10b3ffc2af1e7d04 05697f670b1ea28b80c39854832ea53527f75c55 Diego Devesa slarengh@gmail.com 2024-11-04T13:10:23+01:00 GitHub noreply@github.com 2024-11-04T13:10:23+01:00 cuda : clear error after changing peer access (#10153)
05697f670b1ea28b80c39854832ea53527f75c55 f8e58135cff1c373df2934306f9c9da99673c2ed Georgi Gerganov ggerganov@gmail.com 2024-11-04T13:49:34+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-04T13:49:34+02:00 metal : simplify f16 and f32 dequant kernels (#0)
f8e58135cff1c373df2934306f9c9da99673c2ed 329ed914c959c510d076fb06b43eeb3f7b804d6f Georgi Gerganov ggerganov@gmail.com 2024-11-04T13:43:32+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-04T13:44:06+02:00 metal : move dequantize templates to beginning of MSL source (#0)
329ed914c959c510d076fb06b43eeb3f7b804d6f ce027adfb3b131f0d2368294fc276bb0e342b3f6 leo-pony nengjunma@outlook.com 2024-11-04T19:08:22+08:00 GitHub noreply@github.com 2024-11-04T19:08:22+08:00 CANN: adjust backend registry refactor. (#10158)
ce027adfb3b131f0d2368294fc276bb0e342b3f6 284e5b0275cc1292096e72e808e41d17e8cdf019 Georgi Gerganov ggerganov@gmail.com 2024-11-04T10:33:37+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-04T10:33:37+02:00 sync : ggml
284e5b0275cc1292096e72e808e41d17e8cdf019 e2292aaa17cf8530b0d0d899909588c3a095799d Yuri Khrustalev ykhrustalev@users.noreply.github.com 2024-11-02T05:09:12-04:00 Georgi Gerganov ggerganov@gmail.com 2024-11-04T10:33:11+02:00 cmake : make it possible linking ggml as external lib (ggml/1003)
e2292aaa17cf8530b0d0d899909588c3a095799d 9f409893519b4a6def46ef80cd6f5d05ac0fb157 Plamen Minev pacominev@gmail.com 2024-11-01T16:55:10+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-04T10:33:10+02:00 metal : fix minor string leaks (ggml/1004)
9f409893519b4a6def46ef80cd6f5d05ac0fb157 08828a6d7d0006a487c9655ba8ace0ebe35ecad1 Diego Devesa slarengh@gmail.com 2024-11-03T19:34:08+01:00 GitHub noreply@github.com 2024-11-03T19:34:08+01:00 ggml : move CPU backend to a separate file (#10144)
08828a6d7d0006a487c9655ba8ace0ebe35ecad1 1839f69130151ceeac4d01c0ef8964e1fb43bba6 Georgi Gerganov ggerganov@gmail.com 2024-11-03T15:18:40+02:00 GitHub noreply@github.com 2024-11-03T15:18:40+02:00 metal : minor fixup in FA kernel (#10143)
1839f69130151ceeac4d01c0ef8964e1fb43bba6 9830b6923b61f1e652a35afeac77aa5f886dad09 Georgi Gerganov ggerganov@gmail.com 2024-11-03T15:14:15+02:00 GitHub noreply@github.com 2024-11-03T05:14:15-08:00 flake.lock: Update (#10146)
9830b6923b61f1e652a35afeac77aa5f886dad09 42cadc74bda60afafb45b71b1a39d150ede0ed4d Christian Köhnenkamp cvk5@me.com 2024-11-02T23:35:31+01:00 GitHub noreply@github.com 2024-11-02T15:35:31-07:00 Add apple arm to presets (#10134)
42cadc74bda60afafb45b71b1a39d150ede0ed4d 45950415ed985830c59bf42cf9c9216b20cf08ef sasha0552 admin@sasha0552.org 2024-11-02T16:34:56Z GitHub noreply@github.com 2024-11-02T18:34:56+02:00 server : fix slot selection by lru (#10126)
45950415ed985830c59bf42cf9c9216b20cf08ef 1926d6e39d6f6358bc1a4c52316a560178be7233 Georgi Gerganov ggerganov@gmail.com 2024-11-02T18:34:00+02:00 GitHub noreply@github.com 2024-11-02T18:34:00+02:00 server : fix endpoint checks (#10135)
1926d6e39d6f6358bc1a4c52316a560178be7233 b634f8a26fef65210fd9fb2f87e83a2809535e89 Georgi Gerganov ggerganov@gmail.com 2024-11-02T15:18:56+02:00 GitHub noreply@github.com 2024-11-02T15:18:56+02:00 llama : adjust default context size + print warnings (#10136)
b634f8a26fef65210fd9fb2f87e83a2809535e89 7554aa4655f44b33a29068f2b18c5976fae45f9d Diego Devesa slarengh@gmail.com 2024-11-02T13:08:53+01:00 GitHub noreply@github.com 2024-11-02T13:08:53+01:00 simple-chat : only add bos on first prompt (#10129)
7554aa4655f44b33a29068f2b18c5976fae45f9d a6744e43e80f4be6398fc7733a01642c846dce1d Xuan Son Nguyen thichthat@gmail.com 2024-11-02T12:53:17+01:00 GitHub noreply@github.com 2024-11-02T12:53:17+01:00 convert-lora : make `--base` optional (#10110)
a6744e43e80f4be6398fc7733a01642c846dce1d e991e3127ff71a29e61fe1de5dd1cbd2e1df1858 Diego Devesa slarengh@gmail.com 2024-11-01T23:50:59+01:00 GitHub noreply@github.com 2024-11-01T23:50:59+01:00 llama : add simple-chat example (#10124)
e991e3127ff71a29e61fe1de5dd1cbd2e1df1858 418f5eef262cea07c2af4f45ee6a88d882221fcb Diego Devesa slarengh@gmail.com 2024-11-01T23:48:26+01:00 GitHub noreply@github.com 2024-11-01T23:48:26+01:00 llama : use smart pointers for ggml resources (#10117)
418f5eef262cea07c2af4f45ee6a88d882221fcb ba6f62eb793d6617892d252f5c04d7685d908a38 Shupei Fan dymarkfan@outlook.com 2024-11-02T02:33:14+08:00 GitHub noreply@github.com 2024-11-01T19:33:14+01:00 vulkan : improve ggml_vk_create_buffer error handling (#9898)
ba6f62eb793d6617892d252f5c04d7685d908a38 d865d1478cd4e403f82d793c2afcd0f943412f05 Georgi Gerganov ggerganov@gmail.com 2024-11-01T17:31:51+02:00 GitHub noreply@github.com 2024-11-01T17:31:51+02:00 readme : update hot topics
d865d1478cd4e403f82d793c2afcd0f943412f05 1804adb0cfee4811eaf633741503d683a46e4c77 sasha0552 admin@sasha0552.org 2024-11-01T13:33:14Z GitHub noreply@github.com 2024-11-01T14:33:14+01:00 server : fix smart selection of available slot (#10120)
1804adb0cfee4811eaf633741503d683a46e4c77 815fe72adcea5ec79d358db6a4c479191f396b3c Georgi Gerganov ggerganov@gmail.com 2024-11-01T12:58:45+02:00 GitHub noreply@github.com 2024-11-01T12:58:45+02:00 ggml : remove ggml_scratch (#10121)
815fe72adcea5ec79d358db6a4c479191f396b3c f221d56220899f38f0126e683b2432bc79d1e3f6 Georgi Gerganov ggerganov@gmail.com 2024-11-01T10:28:24+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-01T10:28:24+02:00 sync : ggml
f221d56220899f38f0126e683b2432bc79d1e3f6 e597e50794f07ec8dc24b9efb18f94ec6386fda0 Georgi Gerganov ggerganov@gmail.com 2024-11-01T10:23:05+02:00 Georgi Gerganov ggerganov@gmail.com 2024-11-01T10:24:50+02:00 ggml : alloc ggml_contexts on the heap (whisper/2525)
e597e50794f07ec8dc24b9efb18f94ec6386fda0 85679d37f34f66783cc04664a06c405b28e8e035 Zhenwei Jin 109658203+kylo5aby@users.noreply.github.com 2024-11-01T11:09:59+08:00 GitHub noreply@github.com 2024-11-01T11:09:59+08:00 build: fix build error in Windows env with OneAPI setup (#10107)
85679d37f34f66783cc04664a06c405b28e8e035 1e9f94994ef908d964cf81069f03d9d3668beb7d Diego Devesa slarengh@gmail.com 2024-11-01T00:49:53+01:00 GitHub noreply@github.com 2024-11-01T00:49:53+01:00 llama : improve output buffer type selection (#10098)
1e9f94994ef908d964cf81069f03d9d3668beb7d c02e5ab2a675c8bc1abc8b1e4cb6a93b26bdcce7 Diego Devesa slarengh@gmail.com 2024-11-01T00:45:34+01:00 GitHub noreply@github.com 2024-11-01T00:45:34+01:00 quantize : fix --keep-split (#10114)
c02e5ab2a675c8bc1abc8b1e4cb6a93b26bdcce7 ab3d71f97f5b2915a229099777af00d3eada1d24 Diego Devesa slarengh@gmail.com 2024-10-31T22:54:23+01:00 GitHub noreply@github.com 2024-10-31T22:54:23+01:00 llama : fix buffer checks for mamba and rwk (#10111)
ab3d71f97f5b2915a229099777af00d3eada1d24 0a683e8088d849626e7471f9e2ed381f7dbdf2e9 Zhenwei Jin 109658203+kylo5aby@users.noreply.github.com 2024-11-01T02:50:39+08:00 GitHub noreply@github.com 2024-10-31T19:50:39+01:00 loader: refactor tensor weights storage (#9935)
0a683e8088d849626e7471f9e2ed381f7dbdf2e9 dea5e86051aadcdf42f7db7a8855a78d8f5ff3d6 Kevin Gibbons bakkot@gmail.com 2024-10-31T06:02:35-07:00 GitHub noreply@github.com 2024-10-31T14:02:35+01:00 server : include scheme when printing URL (#10106)
dea5e86051aadcdf42f7db7a8855a78d8f5ff3d6 1329c0a75e6a7defc5c380eaf80d8e0f66d7da78 Diego Devesa slarengh@gmail.com 2024-10-31T11:40:59+01:00 GitHub noreply@github.com 2024-10-31T11:40:59+01:00 ggml : check tensor name lengths in gguf files (#10100)
1329c0a75e6a7defc5c380eaf80d8e0f66d7da78 61408e7fad082dc44a11c8a9f1398da4837aad44 Sergio López slp@redhat.com 2024-10-31T10:09:52+01:00 GitHub noreply@github.com 2024-10-31T11:09:52+02:00 kompute: add mul_mat_q4_k shader (#10097)
61408e7fad082dc44a11c8a9f1398da4837aad44 b9e02e8184f5e6094a9e87eaf040becd404bfc90 Sergio López slp@redhat.com 2024-10-30T17:01:52+01:00 GitHub noreply@github.com 2024-10-30T17:01:52+01:00 kompute: add backend registry / device interfaces (#10045)
b9e02e8184f5e6094a9e87eaf040becd404bfc90 6763f713bb692910e9b2d9d1a82d6959cee2dcf3 Diego Devesa slarengh@gmail.com 2024-10-30T14:51:21+01:00 GitHub noreply@github.com 2024-10-30T14:51:21+01:00 ggml : fix memory leaks when loading invalid gguf files (#10094)
6763f713bb692910e9b2d9d1a82d6959cee2dcf3 79a2bc042dcacaad59306865208a8c8c3149e3ea Rich Dougherty rich@rd.nz 2024-10-31T01:22:39+13:00 GitHub noreply@github.com 2024-10-30T13:22:39+01:00 readme : more lora detail in main example readme (#10064)
79a2bc042dcacaad59306865208a8c8c3149e3ea fc83a9e58479e4dd70054daa7afe5184c1bbe545 Rich Dougherty rich@rd.nz 2024-10-31T01:22:21+13:00 GitHub noreply@github.com 2024-10-30T13:22:21+01:00 convert : more detailed convert lora usage docs (#10065)
fc83a9e58479e4dd70054daa7afe5184c1bbe545 c5b0f4b5d90297f3e729fca7f78ddb25fcab5ddc xctan axunlei@gmail.com 2024-10-30T15:00:40+08:00 GitHub noreply@github.com 2024-10-30T09:00:40+02:00 ggml : add Q4_0_8_8 RISC-V GEMV and GEMM kernels (#10029)
c5b0f4b5d90297f3e729fca7f78ddb25fcab5ddc 8f275a7c4593aa34147595a90282cf950a853690 Diego Devesa slarengh@gmail.com 2024-10-30T02:01:23+01:00 GitHub noreply@github.com 2024-10-30T02:01:23+01:00 llama : refactor model loader with backend registry (#10026)
8f275a7c4593aa34147595a90282cf950a853690 8d8ff715367480b856ad86ac3888e9742b13a6fa Changyeon Kim cyzero.kim@samsung.com 2024-10-29T17:52:56+09:00 GitHub noreply@github.com 2024-10-29T09:52:56+01:00 ggml: Add POOL2D OP for GPU acceleration to the Vulkan backend in the MobileVLM model. (#9763)
8d8ff715367480b856ad86ac3888e9742b13a6fa 61715d5cc83a28181df6a641846e4f6a740f3c74 Georgi Gerganov ggerganov@gmail.com 2024-10-29T10:42:05+02:00 GitHub noreply@github.com 2024-10-29T10:42:05+02:00 llama : remove Tail-Free sampling (#10071)
61715d5cc83a28181df6a641846e4f6a740f3c74 07028f9d74d895da2ca4a1956624e3f07e04e620 arch-btw 57669023+arch-btw@users.noreply.github.com 2024-10-28T10:45:33-07:00 GitHub noreply@github.com 2024-10-28T18:45:33+01:00 llama : Add IBM granite template (#10013)
07028f9d74d895da2ca4a1956624e3f07e04e620 524afeec9dad7d765ce91f5cf30c73703867cb47 Georgi Gerganov ggerganov@gmail.com 2024-10-28T17:41:24+02:00 GitHub noreply@github.com 2024-10-28T08:41:24-07:00 flake.lock: Update (#10063)
524afeec9dad7d765ce91f5cf30c73703867cb47 8125e6cbfcf2b3b9066e4d923aca9295526730f5 R0CKSTAR xiaodong.ye@mthreads.com 2024-10-28T17:02:48+08:00 GitHub noreply@github.com 2024-10-28T10:02:48+01:00 musa: workaround for Guilty Lockup in cleaning src0 (#10042)
8125e6cbfcf2b3b9066e4d923aca9295526730f5 8841ce3f439de6e770f70319b7e08b6613197ea7 Georgi Gerganov ggerganov@gmail.com 2024-10-28T08:49:32+02:00 GitHub noreply@github.com 2024-10-28T08:49:32+02:00 server : don't overfill the batch during infill (#10018)
8841ce3f439de6e770f70319b7e08b6613197ea7 cc2983d3753c94a630ca7257723914d4c4f6122b Georgi Gerganov ggerganov@gmail.com 2024-10-27T20:59:58+02:00 GitHub noreply@github.com 2024-10-27T20:59:58+02:00 llama : switch KQ multiplication to F32 precision by default (#10015)
cc2983d3753c94a630ca7257723914d4c4f6122b 8c60a8a46261ffb92b6d23a78acfac2fcb6fe233 Georgi Gerganov ggerganov@gmail.com 2024-10-26T10:34:08+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-26T10:34:08+03:00 sync : ggml
8c60a8a46261ffb92b6d23a78acfac2fcb6fe233 9e4a2563eadf34e9432d248224d4f43e8495e8fe bssrdf merlintiger@hotmail.com 2024-10-23T14:34:00-04:00 Georgi Gerganov ggerganov@gmail.com 2024-10-26T10:33:56+03:00 increase cuda_cpy block size (ggml/996)
9e4a2563eadf34e9432d248224d4f43e8495e8fe 668750357e66bfa3d1504b65699f5a0dfe3cb7cb Georgi Gerganov ggerganov@gmail.com 2024-10-26T10:33:31+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-26T10:33:31+03:00 scripts : fix amx sync [no ci]
668750357e66bfa3d1504b65699f5a0dfe3cb7cb ff252ea48e90e6552010fd74584334fb41bdd387 Georgi Gerganov ggerganov@gmail.com 2024-10-25T22:26:15+03:00 GitHub noreply@github.com 2024-10-25T22:26:15+03:00 metal : support permuted matrix multiplicaions (#10033)
ff252ea48e90e6552010fd74584334fb41bdd387 d80fb71f8b8bf69ec095ba281f8248d136d21c76 wwoodsTM 104587230+wwoodsTM@users.noreply.github.com 2024-10-25T10:07:34-06:00 GitHub noreply@github.com 2024-10-25T19:07:34+03:00 llama : add DRY sampler (#9702)
d80fb71f8b8bf69ec095ba281f8248d136d21c76 2f8bd2b90133cf37ae752015e1bfd738cc6d0112 Michael Podvitskiy podvitskiymichael@gmail.com 2024-10-25T17:57:54+02:00 GitHub noreply@github.com 2024-10-25T17:57:54+02:00 llama: string_split fix (#10022)
2f8bd2b90133cf37ae752015e1bfd738cc6d0112 bc5ba007b2c83ac95875e68724dabfc12159fc61 Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-10-25T12:57:41+05:30 GitHub noreply@github.com 2024-10-25T10:27:41+03:00 llamafile : extend sgemm.cpp support for Q5_0 models (#10010)
bc5ba007b2c83ac95875e68724dabfc12159fc61 958367bf530d943a902afa1ce1c342476098576b Georgi Gerganov ggerganov@gmail.com 2024-10-25T10:13:46+03:00 GitHub noreply@github.com 2024-10-25T10:13:46+03:00 server : check that the prompt fits in the slot's context (#10030)
958367bf530d943a902afa1ce1c342476098576b 40f2555797f97314de749873cdc29dc102be66e2 Xuan Son Nguyen thichthat@gmail.com 2024-10-24T21:51:22+02:00 GitHub noreply@github.com 2024-10-24T21:51:22+02:00 server : refactor slot input data, move tokenizer to HTTP thread (#10023)
40f2555797f97314de749873cdc29dc102be66e2 167a515651a4b065a16225ffc69564c5674f3d0f Georgi Gerganov ggerganov@gmail.com 2024-10-24T21:23:33+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-24T21:23:33+03:00 ci : fix cmake flags for SYCL
167a515651a4b065a16225ffc69564c5674f3d0f c39665f589091903396a442a6ee56613303e0350 Johannes Gäßler johannesg@5d6.de 2024-10-24T14:40:23+02:00 GitHub noreply@github.com 2024-10-24T14:40:23+02:00 CUDA: fix insufficient buffer clearing for MMQ (#10032)
c39665f589091903396a442a6ee56613303e0350 0a1c750c80147687df267114c81956757cc14382 Johannes Gäßler johannesg@5d6.de 2024-10-24T11:09:36+02:00 GitHub noreply@github.com 2024-10-24T11:09:36+02:00 CUDA: fix MMQ for non-contiguous src0, add tests (#10021)
0a1c750c80147687df267114c81956757cc14382 190a37d7977eb5bd6a729299bd1e371208c87149 wwoodsTM 104587230+wwoodsTM@users.noreply.github.com 2024-10-23T13:27:51-06:00 GitHub noreply@github.com 2024-10-23T22:27:51+03:00 server : samplers accept the prompt correctly (#10019)
190a37d7977eb5bd6a729299bd1e371208c87149 2d3aba9ee8da9c026d54e8a912a1d64f56809be3 Georgi Gerganov ggerganov@gmail.com 2024-10-23T17:23:55+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-23T17:23:55+03:00 sync : ggml
2d3aba9ee8da9c026d54e8a912a1d64f56809be3 80273a306d07ed95059d6130389deacb3b2d7196 Georgi Gerganov ggerganov@gmail.com 2024-10-23T17:16:56+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-23T17:16:56+03:00 llama.vim : bump generation time limit to 3s [no ci]
80273a306d07ed95059d6130389deacb3b2d7196 c19af0acb1fe6d0fdbecadd8483c1fbe5d68d095 Johannes Gäßler johannesg@5d6.de 2024-10-18T09:24:44+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-23T16:50:02+03:00 CUDA: fix 1D im2col, add tests (ggml/993)
c19af0acb1fe6d0fdbecadd8483c1fbe5d68d095 ac113a0feee0935b2018312f7bc8d7a646b117ed Daniel Bevenius daniel.bevenius@gmail.com 2024-10-16T20:10:01+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-23T16:50:02+03:00 ggml : remove redundant set of contexts used field (ggml/978)
ac113a0feee0935b2018312f7bc8d7a646b117ed 4c9388fb96ac2415fbb1239b7ba8346616606e2e Michael Coppola m18coppola@gmail.com 2024-10-23T07:09:26-04:00 GitHub noreply@github.com 2024-10-23T14:09:26+03:00 llama.vim : add classic vim support (#9995)
4c9388fb96ac2415fbb1239b7ba8346616606e2e 873279b1592e433c4d9eb5065091cc98473c7bee Jun Hee Yoo contact.jhyoo@gmail.com 2024-10-23T19:33:45+09:00 GitHub noreply@github.com 2024-10-23T13:33:45+03:00 metal : add POOL2D and fix IM2COL (#9943)
873279b1592e433c4d9eb5065091cc98473c7bee c8c07d658a6cefc5a50cfdf6be7d726503612303 github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-10-20T00:22:59Z Someone else@someonex.net 2024-10-23T01:28:07Z flake.lock: Update
c8c07d658a6cefc5a50cfdf6be7d726503612303 19d900a7565b8f6b0a708836a57d26966cb9efe2 Xuan Son Nguyen thichthat@gmail.com 2024-10-22T16:59:02+02:00 GitHub noreply@github.com 2024-10-22T16:59:02+02:00 llama : fix empty batch causing llama_batch_allocr to crash (#9966)
19d900a7565b8f6b0a708836a57d26966cb9efe2 11d47057a51f3d9b9231e6b57d0ca36020c0ee99 Daniel Bevenius daniel.bevenius@gmail.com 2024-10-22T15:31:06+02:00 GitHub noreply@github.com 2024-10-22T16:31:06+03:00 llama : rename batch to ubatch (#9950)
11d47057a51f3d9b9231e6b57d0ca36020c0ee99 c421ac072d46172ab18924e1e8be53680b54ed3b Molly Sophia mollysophia379@gmail.com 2024-10-22T21:22:26+08:00 GitHub noreply@github.com 2024-10-22T15:22:26+02:00 Rwkv chat template fix (#10001)
c421ac072d46172ab18924e1e8be53680b54ed3b 4ff7fe1fb36b04ddd158b2de881c348c5f0ff5e4 Xuan Son Nguyen thichthat@gmail.com 2024-10-22T13:08:41+02:00 GitHub noreply@github.com 2024-10-22T13:08:41+02:00 lora : warn user if new token is added in the adapter (#9948)
4ff7fe1fb36b04ddd158b2de881c348c5f0ff5e4 6b8447352df3d662b56280c8fc38d7f092885787 Molly Sophia mollysophia379@gmail.com 2024-10-22T18:33:37+08:00 GitHub noreply@github.com 2024-10-22T13:33:37+03:00 llama : add chat template for RWKV-World + fix EOT (#9968)
6b8447352df3d662b56280c8fc38d7f092885787 674804a99617b4f90292b4080ecab450ea3d30ba leo-pony nengjunma@outlook.com 2024-10-22T16:16:01+08:00 GitHub noreply@github.com 2024-10-22T16:16:01+08:00 [CANN] Adapt to dynamically loadable backends mechanism (#9970)
674804a99617b4f90292b4080ecab450ea3d30ba e94a138d644a9b34da61805f7aeb8af595c61b53 Daniel Bevenius daniel.bevenius@gmail.com 2024-10-22T09:40:02+02:00 GitHub noreply@github.com 2024-10-22T10:40:02+03:00 arg : fix typo in embeddings argument help [no ci] (#9994)
e94a138d644a9b34da61805f7aeb8af595c61b53 e01c67affe450638162a1a457e2e57859ef6ebf0 Georgi Gerganov ggerganov@gmail.com 2024-10-22T00:35:25+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-22T00:37:55+03:00 llama.vim : fix info text display [no ci] (#9787)
e01c67affe450638162a1a457e2e57859ef6ebf0 994cfb1acb9144bc95be0ab319175f30737cc92b Georgi Gerganov ggerganov@gmail.com 2024-10-21T22:52:22+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-21T22:53:18+03:00 llama.vim : move info to the right of screen [no ci] (#9787)
994cfb1acb9144bc95be0ab319175f30737cc92b 94008cc76075fb4a29ee371e7ac255378d1bce6c Asghar Ghorbani a-ghorbani@users.noreply.github.com 2024-10-21T20:20:59+02:00 GitHub noreply@github.com 2024-10-21T21:20:59+03:00 readme : update UI list (#9972)
94008cc76075fb4a29ee371e7ac255378d1bce6c dbd5f2f5736aec6ff8fd63df3b351dae23c43e2f Daniel Bevenius daniel.bevenius@gmail.com 2024-10-21T20:12:52+02:00 GitHub noreply@github.com 2024-10-21T21:12:52+03:00 arg : fix attention non-causal arg value hint (#9985)
dbd5f2f5736aec6ff8fd63df3b351dae23c43e2f f594bc80baf683818f29d8f5d6fb52daab99e572 Georgi Gerganov ggerganov@gmail.com 2024-10-21T20:25:02+03:00 GitHub noreply@github.com 2024-10-21T20:25:02+03:00 llama.vim : plugin for Neovim (#9787)
f594bc80baf683818f29d8f5d6fb52daab99e572 d5ebd79c76abd4887f0283cd6f6f9689122094d0 Georgi Gerganov ggerganov@gmail.com 2024-10-21T16:20:46+03:00 GitHub noreply@github.com 2024-10-21T16:20:46+03:00 ggml : add asserts for type conversion in fattn kernels (#9971)
d5ebd79c76abd4887f0283cd6f6f9689122094d0 55e47786e373c90fc7803e718e3e1dd6d53c3db6 Radoslav Gerganov rgerganov@gmail.com 2024-10-21T13:35:40+03:00 GitHub noreply@github.com 2024-10-21T13:35:40+03:00 rpc : pack only RPC structs (#9959)
55e47786e373c90fc7803e718e3e1dd6d53c3db6 bc219750845a59166d79f0d4ee3da1993b369b8a Georgi Gerganov ggerganov@gmail.com 2024-10-21T09:46:40+03:00 GitHub noreply@github.com 2024-10-21T09:46:40+03:00 llama : default sampling changes + greedy update (#9897)
bc219750845a59166d79f0d4ee3da1993b369b8a 1db8c84fc62857e1e45c1c7ea93bcd5344cb3d31 Georgi Gerganov ggerganov@gmail.com 2024-10-21T09:37:12+03:00 GitHub noreply@github.com 2024-10-21T09:37:12+03:00 speculative : fix handling of some input params (#9963)
1db8c84fc62857e1e45c1c7ea93bcd5344cb3d31 45f097645efb11b6d09a5b4adbbfd7c312ac0126 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-10-21T14:26:09+08:00 GitHub noreply@github.com 2024-10-21T14:26:09+08:00 fix mul_mat_vec_q and *_vec_q error (#9939)
45f097645efb11b6d09a5b4adbbfd7c312ac0126 7cab2083c768dd92c20b105556c4165b59cd8a41 Loïc Carrère loic.carrere@gmail.com 2024-10-20T18:25:41+02:00 GitHub noreply@github.com 2024-10-20T19:25:41+03:00 readme : update bindings list (#9951)
7cab2083c768dd92c20b105556c4165b59cd8a41 cda0e4b648dde8fac162b3430b14a99597d3d74f icppWorld 124377669+icppWorld@users.noreply.github.com 2024-10-20T12:01:34-04:00 GitHub noreply@github.com 2024-10-20T19:01:34+03:00 readme : update infra list (#9942)
cda0e4b648dde8fac162b3430b14a99597d3d74f afd9909a6481402844aecefa8a8908afdd7f52f1 Xuan Son Nguyen thichthat@gmail.com 2024-10-18T23:18:01+02:00 GitHub noreply@github.com 2024-10-18T23:18:01+02:00 llama : remove all_pos_0, all_pos_1, all_seq_id from llama_batch (#9745)
afd9909a6481402844aecefa8a8908afdd7f52f1 87421a23e8c60e00a7b227d501e8aab2a1aff7ce Radoslav Gerganov rgerganov@gmail.com 2024-10-18T14:33:58+03:00 GitHub noreply@github.com 2024-10-18T14:33:58+03:00 rpc : backend refactoring (#9912)
87421a23e8c60e00a7b227d501e8aab2a1aff7ce 60ce97c9d809f4b040e90b597468b839df5728d0 Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-10-18T06:46:16+01:00 GitHub noreply@github.com 2024-10-18T06:46:16+01:00 [SYCL] Add SYCL Backend registry, device and Event Interfaces (#9705)
60ce97c9d809f4b040e90b597468b839df5728d0 8901755ba328643c9ab071c20e1939ea52951a0e Ma Mingfei mingfei.ma@intel.com 2024-10-18T13:34:36+08:00 GitHub noreply@github.com 2024-10-18T13:34:36+08:00 add amx kernel for gemm (#8998)
8901755ba328643c9ab071c20e1939ea52951a0e 6f55bccbb8835d42147add4ee48807450f5ff535 Georgi Gerganov ggerganov@gmail.com 2024-10-18T07:32:19+03:00 GitHub noreply@github.com 2024-10-18T07:32:19+03:00 server : add n_indent parameter for line indentation requirement (#9929)
6f55bccbb8835d42147add4ee48807450f5ff535 17bb9280807cfbb6611b853aa1ef05114bd9efe9 Daniel Bevenius daniel.bevenius@gmail.com 2024-10-18T01:41:51+02:00 GitHub noreply@github.com 2024-10-18T01:41:51+02:00 llama : rename batch_all to batch (#8881)
17bb9280807cfbb6611b853aa1ef05114bd9efe9 9f45fc1e9950a496febc575cdd196cd5cad000cc Georgi Gerganov ggerganov@gmail.com 2024-10-17T23:43:05+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-17T23:43:05+03:00 readme : remove --memory-f32 references (#9925)
9f45fc1e9950a496febc575cdd196cd5cad000cc 99bd4ac28c32cd17c0e337ff5601393b033dc5fc Georgi Gerganov ggerganov@gmail.com 2024-10-17T23:26:32+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-17T23:27:42+03:00 llama : change warning to debug log
99bd4ac28c32cd17c0e337ff5601393b033dc5fc 3752217ed5a6a11864682fbf009bcb36afffd6bc Georgi Gerganov ggerganov@gmail.com 2024-10-17T22:32:47+03:00 GitHub noreply@github.com 2024-10-17T22:32:47+03:00 llama : infill sampling handle very long tokens (#9924)
3752217ed5a6a11864682fbf009bcb36afffd6bc f010b77a372ffcfaf4338c670d6d3ecd89aa4eb6 Tim Wang overocean@gmail.com 2024-10-17T17:57:14+11:00 GitHub noreply@github.com 2024-10-17T09:57:14+03:00 readme : update bindings list (#9918)
f010b77a372ffcfaf4338c670d6d3ecd89aa4eb6 21942002780352b4a54f4bd3e5eefa3bc7f14fe6 Diego Devesa slarengh@gmail.com 2024-10-17T02:46:58+02:00 GitHub noreply@github.com 2024-10-17T02:46:58+02:00 vulkan : add backend registry / device interfaces (#9721)
21942002780352b4a54f4bd3e5eefa3bc7f14fe6 73afe681aa76e818733fc1f30de082c1d6910bcd Gilad S. 7817232+giladgd@users.noreply.github.com 2024-10-17T02:34:22+03:00 GitHub noreply@github.com 2024-10-17T01:34:22+02:00 fix: allocating CPU buffer with size `0` (#9917)
73afe681aa76e818733fc1f30de082c1d6910bcd 9e041024481f6b249ab8918e18b9477f873b5a5e Gilad S. 7817232+giladgd@users.noreply.github.com 2024-10-17T01:36:51+03:00 GitHub noreply@github.com 2024-10-17T00:36:51+02:00 fix: use `vm_allocate` to allocate CPU backend buffer on macOS (#9875)
9e041024481f6b249ab8918e18b9477f873b5a5e dbf18e4de9e7aa496871f1555f9f0c8d84567108 Daniel Bevenius daniel.bevenius@gmail.com 2024-10-16T19:34:28+02:00 GitHub noreply@github.com 2024-10-16T20:34:28+03:00 llama : suppress conversion from 'size_t' to 'int' (#9046)
dbf18e4de9e7aa496871f1555f9f0c8d84567108 66c2c93082289325199ae1f773f3b0ab2e399a47 Daniel Bevenius daniel.bevenius@gmail.com 2024-10-16T19:24:05+02:00 GitHub noreply@github.com 2024-10-16T20:24:05+03:00 llava : fix typo in error message [no ci] (#9884)
66c2c93082289325199ae1f773f3b0ab2e399a47 10433e8b457c4cfd759cbb41fc55fc398db4a5da Joe Eli McIlvain joe.eli.mac@gmail.com 2024-10-16T09:03:24-07:00 GitHub noreply@github.com 2024-10-16T19:03:24+03:00 grammar : fix JSON Schema for string regex with top-level alt. (#9903)
10433e8b457c4cfd759cbb41fc55fc398db4a5da 1f66b699c48cb5ab3265ed72c48e8549b1674291 Molly Sophia mollysophia379@gmail.com 2024-10-16T18:10:21+08:00 GitHub noreply@github.com 2024-10-16T13:10:21+03:00 llama : add tensor name for "result_norm" (#9907)
1f66b699c48cb5ab3265ed72c48e8549b1674291 0e41b300ed28f7fe185d938b2e3d56a0bf7411ed Alexey Parfenov zxed@alkatrazstudio.net 2024-10-16T08:35:53Z GitHub noreply@github.com 2024-10-16T11:35:53+03:00 server : fix the disappearance of the end of the text (#9867)
0e41b300ed28f7fe185d938b2e3d56a0bf7411ed cd60b88bf7ad7785fb6ac9864e360cf10e42faad Georgi Gerganov ggerganov@gmail.com 2024-10-16T11:28:14+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-16T11:28:14+03:00 sync : ggml
cd60b88bf7ad7785fb6ac9864e360cf10e42faad becfd387f6919d99ec34b76c2522f90ac250c489 Daniel Bevenius daniel.bevenius@gmail.com 2024-10-09T16:40:35+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-16T11:28:01+03:00 ggml-alloc : remove buffer_id from leaf_alloc (ggml/987)
becfd387f6919d99ec34b76c2522f90ac250c489 755a9b2bf00fbae988e03a47e852b66eaddd113a leo-pony nengjunma@outlook.com 2024-10-16T08:51:46+08:00 GitHub noreply@github.com 2024-10-16T08:51:46+08:00 [CANN] Fix cann compilation error (#9891)
755a9b2bf00fbae988e03a47e852b66eaddd113a 223c25a72fcc3f65cdfd7f5d57edd5b44b550e18 Georgi Gerganov ggerganov@gmail.com 2024-10-15T16:35:33+03:00 GitHub noreply@github.com 2024-10-15T16:35:33+03:00 llama : add infill sampler (#9896)
223c25a72fcc3f65cdfd7f5d57edd5b44b550e18 fbc98b748e7b075e327bcf13237057f647678049 Georgi Gerganov ggerganov@gmail.com 2024-10-15T16:28:55+03:00 GitHub noreply@github.com 2024-10-15T16:28:55+03:00 server : improve infill context reuse (#9894)
fbc98b748e7b075e327bcf13237057f647678049 dcdd535302fc9702a4709be25f56540d65163a44 MaggotHATE clay1326@gmail.com 2024-10-15T15:54:55+05:00 GitHub noreply@github.com 2024-10-15T12:54:55+02:00 sampling : add XTC sampler (#9742)
dcdd535302fc9702a4709be25f56540d65163a44 4c42f93b22146c83b763d8cbee5fafc512746649 Georgi Gerganov ggerganov@gmail.com 2024-10-15T12:48:44+03:00 GitHub noreply@github.com 2024-10-15T12:48:44+03:00 server : update preact (#9895)
4c42f93b22146c83b763d8cbee5fafc512746649 a89f75e1b7b90cb2d4d4c52ca53ef9e9b466aa45 Michał Tuszyński srgtuszy@gmail.com 2024-10-15T10:20:34+02:00 GitHub noreply@github.com 2024-10-15T11:20:34+03:00 readme : update bindings list (#9889)
a89f75e1b7b90cb2d4d4c52ca53ef9e9b466aa45 13dca2a54a394757d56fdd652b9f0df08f44ea22 VoidIsVoid 343750470@qq.com 2024-10-14T15:04:36+08:00 GitHub noreply@github.com 2024-10-14T10:04:36+03:00 server : handle "logprobs" field with false value (#9871)
13dca2a54a394757d56fdd652b9f0df08f44ea22 d4c19c0f5cdb1e512573e8c86c79e8d0238c73c4 agray3 agray3@users.noreply.github.com 2024-10-14T01:49:08+01:00 GitHub noreply@github.com 2024-10-14T02:49:08+02:00 Vectorize load instructions in dmmv f16 CUDA kernel (#9816)
d4c19c0f5cdb1e512573e8c86c79e8d0238c73c4 c7181bd294757dd80a7904e3dd0fea2d0be914e7 Georgi Gerganov ggerganov@gmail.com 2024-10-13T21:31:35+03:00 GitHub noreply@github.com 2024-10-13T21:31:35+03:00 server : accept extra_context for the infill endpoint (#9874)
c7181bd294757dd80a7904e3dd0fea2d0be914e7 92be9f12164f18ce845a5bab60cefa5f7fec6836 Georgi Gerganov ggerganov@gmail.com 2024-10-13T18:52:48+03:00 GitHub noreply@github.com 2024-10-13T18:52:48+03:00 server : reuse cached context chunks (#9866)
92be9f12164f18ce845a5bab60cefa5f7fec6836 edc265661cd707327297b6ec4d83423c43cb50a5 Georgi Gerganov ggerganov@gmail.com 2024-10-13T06:11:26+03:00 GitHub noreply@github.com 2024-10-12T20:11:26-07:00 flake.lock: Update (#9870)
edc265661cd707327297b6ec4d83423c43cb50a5 1bde94dd024b632f98428f4bf2ce483295130779 Georgi Gerganov ggerganov@gmail.com 2024-10-12T16:14:27+03:00 GitHub noreply@github.com 2024-10-12T16:14:27+03:00 server : add option to time limit the generation phase (#9865)
1bde94dd024b632f98428f4bf2ce483295130779 95c76e8e92ecc93f784b185eafae36a0e7ad2fa3 Georgi Gerganov ggerganov@gmail.com 2024-10-12T16:06:31+03:00 GitHub noreply@github.com 2024-10-12T16:06:31+03:00 server : remove self-extend features (#9860)
95c76e8e92ecc93f784b185eafae36a0e7ad2fa3 11ac9800aff532715a5bc7991062c68ba3472e6e Georgi Gerganov ggerganov@gmail.com 2024-10-12T14:51:54+03:00 GitHub noreply@github.com 2024-10-12T14:51:54+03:00 server : remove legacy system_prompt feature (#9857)
11ac9800aff532715a5bc7991062c68ba3472e6e 943d20b4111c746bcd9dbc7e4771de313b08b50c Georgi Gerganov ggerganov@gmail.com 2024-10-12T08:21:51+03:00 GitHub noreply@github.com 2024-10-12T08:21:51+03:00 llama : improve infill support and special token detection (#9798)
943d20b4111c746bcd9dbc7e4771de313b08b50c 96776405a17034dcfd53d3ddf5d142d34bdbb657 R0CKSTAR xiaodong.ye@mthreads.com 2024-10-12T13:09:53+08:00 GitHub noreply@github.com 2024-10-12T08:09:53+03:00 musa : update doc (#9856)
96776405a17034dcfd53d3ddf5d142d34bdbb657 7eee341bee09957139789c2d828995953f0fc7ff Diego Devesa slarengh@gmail.com 2024-10-11T15:34:45+02:00 GitHub noreply@github.com 2024-10-11T15:34:45+02:00 ggml : move more prints to the ggml log system (#9839)
7eee341bee09957139789c2d828995953f0fc7ff 0e9f760eb12546704ef8fa72577bc1a3ffe1bc04 Diego Devesa slarengh@gmail.com 2024-10-10T22:57:42+02:00 GitHub noreply@github.com 2024-10-10T22:57:42+02:00 common : use common_ prefix for common library functions (#9805)
0e9f760eb12546704ef8fa72577bc1a3ffe1bc04 cf8e0a3bb9c0e93e371773b282054cdbbb231038 Diego Devesa slarengh@gmail.com 2024-10-10T20:14:55+02:00 GitHub noreply@github.com 2024-10-10T20:14:55+02:00 rpc : add backend registry / device interfaces (#9812)
cf8e0a3bb9c0e93e371773b282054cdbbb231038 c7499c557cc1efafaf0a6bc12963c39826299703 R0CKSTAR xiaodong.ye@mthreads.com 2024-10-11T02:10:37+08:00 GitHub noreply@github.com 2024-10-10T20:10:37+02:00 musa: add docker image support (#9685)
c7499c557cc1efafaf0a6bc12963c39826299703 c81f3bbb051f8b736e117dfc78c99d7c4e0450f6 Diego Devesa slarengh@gmail.com 2024-10-10T19:50:49+02:00 GitHub noreply@github.com 2024-10-10T19:50:49+02:00 examples : do not use common library in simple example (#9803)
c81f3bbb051f8b736e117dfc78c99d7c4e0450f6 e7022064ab637ccb5f37867196f1802c4a453c91 Diego Devesa slarengh@gmail.com 2024-10-09T18:49:52+02:00 GitHub noreply@github.com 2024-10-09T18:49:52+02:00 cmake : do not build common library by default when standalone (#9804)
e7022064ab637ccb5f37867196f1802c4a453c91 3dc48fe75ad48f8856118520a267c96f74df8e90 Georgi Gerganov ggerganov@gmail.com 2024-10-09T17:00:18+03:00 GitHub noreply@github.com 2024-10-09T17:00:18+03:00 perplexity : fix integer overflow (#9783)
3dc48fe75ad48f8856118520a267c96f74df8e90 dca1d4b58a7f1acf1bd253be84e50d6367f492fd Georgi Gerganov ggerganov@gmail.com 2024-10-09T10:55:42+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-09T10:55:42+03:00 examples : remove llama.vim
dca1d4b58a7f1acf1bd253be84e50d6367f492fd 458367a90606448a9c0262b276947c9e536086e0 Diego Devesa slarengh@gmail.com 2024-10-08T14:21:43+02:00 GitHub noreply@github.com 2024-10-08T14:21:43+02:00 ggml : fix BLAS with unsupported types (#9775)
458367a90606448a9c0262b276947c9e536086e0 fa42aa6d8902cc4eaf31866b3b3b7b61b69da930 Xuan Son Nguyen thichthat@gmail.com 2024-10-08T13:27:04+02:00 GitHub noreply@github.com 2024-10-08T13:27:04+02:00 server : better security control for public deployments (#9776)
fa42aa6d8902cc4eaf31866b3b3b7b61b69da930 6374743747b14db4eb73ce82ae449a2978bc3b47 standby24x7 standby24x7@gmail.com 2024-10-08T15:19:53+09:00 GitHub noreply@github.com 2024-10-08T09:19:53+03:00 scripts : fix spelling typo in messages and comments (#9782)
6374743747b14db4eb73ce82ae449a2978bc3b47 f1af42fa8c925096407c61ff0a3d5d5d669cc535 Diego Devesa slarengh@gmail.com 2024-10-07T21:55:08+02:00 GitHub noreply@github.com 2024-10-07T21:55:08+02:00 ggml : add backend registry / device interfaces to BLAS backend (#9752)
f1af42fa8c925096407c61ff0a3d5d5d669cc535 6279dac039ddeb6d5ebd125a6274fd3c37a77ba8 Andrew Minh Nguyen 40281306+amqdn@users.noreply.github.com 2024-10-07T09:37:31-07:00 GitHub noreply@github.com 2024-10-07T09:37:31-07:00 Update building for Android (#9672)
6279dac039ddeb6d5ebd125a6274fd3c37a77ba8 d5ac8cf2f2e30459489e6721a17d15b92a0c42a6 Georgi Gerganov ggerganov@gmail.com 2024-10-07T19:35:42+03:00 GitHub noreply@github.com 2024-10-07T09:35:42-07:00 flake.lock: Update (#9753)
d5ac8cf2f2e30459489e6721a17d15b92a0c42a6 96b69121033d2b6b951d1b6b1b43f8b4f97dac99 Georgi Gerganov ggerganov@gmail.com 2024-10-07T18:27:51+03:00 GitHub noreply@github.com 2024-10-07T18:27:51+03:00 ggml : add metal backend registry / device (#9713)
96b69121033d2b6b951d1b6b1b43f8b4f97dac99 d5cb86844f26f600c48bf3643738ea68138f961d Paul Tsochantaris ptsochantaris@icloud.com 2024-10-07T13:26:31+01:00 GitHub noreply@github.com 2024-10-07T15:26:31+03:00 metal : single allocation of encode_async block (#9747)
d5cb86844f26f600c48bf3643738ea68138f961d f4b2dcdf4992ef11a854abc9b662624490e37b4c Georgi Gerganov ggerganov@gmail.com 2024-10-06T14:15:27+03:00 GitHub noreply@github.com 2024-10-06T14:15:27+03:00 contrib : simplify + minor edits [no ci]
f4b2dcdf4992ef11a854abc9b662624490e37b4c b6d6c5289f1c9c677657c380591201ddb210b649 Georgi Gerganov ggerganov@gmail.com 2024-10-06T13:49:41+03:00 GitHub noreply@github.com 2024-10-06T13:49:41+03:00 readme : fix typo [no ci]
b6d6c5289f1c9c677657c380591201ddb210b649 b0915d5b51cbaa982ce9bbb9ce302bb9abdca0eb Georgi Gerganov ggerganov@gmail.com 2024-10-06T12:53:28+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-06T12:53:28+03:00 sync : llama.cpp
b0915d5b51cbaa982ce9bbb9ce302bb9abdca0eb 8c475b97b8ba7d678d4c9904b1161bd8811a9b44 SRHMorris 69468379+SRHMorris@users.noreply.github.com 2024-10-06T08:34:20+01:00 Georgi Gerganov ggerganov@gmail.com 2024-10-06T12:52:11+03:00 vulkan : retry allocation with fallback flags (whisper/2451)
8c475b97b8ba7d678d4c9904b1161bd8811a9b44 58b16695e146628481c6b9b8a3b101c0c9bac00f Georgi Gerganov ggerganov@gmail.com 2024-10-05T15:55:04+03:00 GitHub noreply@github.com 2024-10-05T15:55:04+03:00 rerank : use [SEP] token instead of [BOS] (#9737)
58b16695e146628481c6b9b8a3b101c0c9bac00f 905f5485b279518d30b402565c23fb153f822c0d Georgi Gerganov ggerganov@gmail.com 2024-10-05T15:53:49+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-05T15:53:49+03:00 sync : ggml
905f5485b279518d30b402565c23fb153f822c0d 71967c2a6d30da9f61580d3e2d4cb00e0223b6fa Georgi Gerganov ggerganov@gmail.com 2024-10-05T14:33:54+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-05T15:53:00+03:00 metal : zero-init buffer contexts (whisper/0)
71967c2a6d30da9f61580d3e2d4cb00e0223b6fa 17880771ad7dca16cdc969062f2a56f779662835 Viet-Anh NGUYEN (Andrew) vietanh.dev@gmail.com 2024-10-05T01:29:35+07:00 GitHub noreply@github.com 2024-10-04T20:29:35+02:00 Add Llama Assistant (#9744)
17880771ad7dca16cdc969062f2a56f779662835 55951c018d7c107b6ef0a4c8561a6e68183d19d9 Georgi Gerganov ggerganov@gmail.com 2024-10-04T18:50:25+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-04T18:50:25+03:00 sync : ggml
55951c018d7c107b6ef0a4c8561a6e68183d19d9 ff565769f289c6adcc91ed1b8fdabaf9a0d4f6ee Daniel Bevenius daniel.bevenius@gmail.com 2024-10-04T15:46:18+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-04T18:50:05+03:00 ggml : fix typo in example usage ggml_gallocr_new (ggml/984)
ff565769f289c6adcc91ed1b8fdabaf9a0d4f6ee f3fdcfaa79afa12047def3a8793d4a566e0532d4 Diego Devesa slarengh@gmail.com 2024-10-04T08:41:40+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-04T18:50:04+03:00 ggml : fixes after sync (ggml/983)
f3fdcfaa79afa12047def3a8793d4a566e0532d4 133c7b46b3482f7c126c0c4ba14265f684138306 Xuan Son Nguyen thichthat@gmail.com 2024-10-04T11:47:19+02:00 GitHub noreply@github.com 2024-10-04T11:47:19+02:00 ci : fine-grant permission (#9710)
133c7b46b3482f7c126c0c4ba14265f684138306 d5ed2b929d85bbd7dbeecb690880f07d9d7a6077 Daniel Kleine 53251018+d-kleine@users.noreply.github.com 2024-10-04T10:54:44+02:00 GitHub noreply@github.com 2024-10-04T10:54:44+02:00 Fixed RNG seed docs (#9723)
d5ed2b929d85bbd7dbeecb690880f07d9d7a6077 1bb8a64ebfcbe599dacb4fc8069731b6cba0b5d6 Georgi Gerganov ggerganov@gmail.com 2024-10-03T21:18:19+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-03T21:18:19+03:00 metal : remove abort (skip) (ggml/0)
1bb8a64ebfcbe599dacb4fc8069731b6cba0b5d6 fabdc3bda396307565c4f3f4ecbc3a751a2eb6d3 Georgi Gerganov ggerganov@gmail.com 2024-10-03T21:17:49+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-03T21:17:49+03:00 sync : ggml
fabdc3bda396307565c4f3f4ecbc3a751a2eb6d3 eee39bdc96065b69242877fe8f1be05c885fc2aa Johannes Gäßler johannesg@5d6.de 2024-10-03T17:29:59+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-03T21:17:26+03:00 ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
eee39bdc96065b69242877fe8f1be05c885fc2aa 5d5ab1e5cca0d6d63701a1cb85dbe26cb57d2c4e Johannes Gäßler johannesg@5d6.de 2024-10-02T15:32:39+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-03T21:17:26+03:00 ggml: refactor cross entropy loss CPU impl. (ggml/976)
5d5ab1e5cca0d6d63701a1cb85dbe26cb57d2c4e a7ad553513a5d70b4ceacd36f64705cf3654dc97 Jack Mousseau jack@software.inc 2024-10-03T11:01:46-07:00 GitHub noreply@github.com 2024-10-03T21:01:46+03:00 metal : fix compute pass descriptor autorelease crash (#9718)
a7ad553513a5d70b4ceacd36f64705cf3654dc97 d6fe7abf04e8ec5240dead6e2773ed1b7e7495d3 Diego Devesa slarengh@gmail.com 2024-10-03T17:39:18+02:00 GitHub noreply@github.com 2024-10-03T17:39:18+02:00 ggml-backend : add device description to CPU backend (#9720)
d6fe7abf04e8ec5240dead6e2773ed1b7e7495d3 e3c355ba654d4164c1c09e5d0dcacecb8b214af8 bandoti 141645996+bandoti@users.noreply.github.com 2024-10-03T12:39:03-03:00 GitHub noreply@github.com 2024-10-03T17:39:03+02:00 ggml: unify backend logging mechanism (#9709)
e3c355ba654d4164c1c09e5d0dcacecb8b214af8 841713e1e487bdb82fd106a52ad998c5f87b59e9 compilade git@compilade.net 2024-10-03T10:22:15-04:00 GitHub noreply@github.com 2024-10-03T17:22:15+03:00 convert : handle tokenizer merges format from transformers 4.45 (#9696)
841713e1e487bdb82fd106a52ad998c5f87b59e9 5639971466ed74386a1811938022f0c333007b55 Radoslav Gerganov rgerganov@gmail.com 2024-10-03T13:00:52+03:00 GitHub noreply@github.com 2024-10-03T13:00:52+03:00 rpc : enable vulkan (#9714)
5639971466ed74386a1811938022f0c333007b55 c83ad6d01e7b89ec71080d97c7e5db7ac1f4fda6 Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-10-03T07:50:44+01:00 GitHub noreply@github.com 2024-10-03T07:50:44+01:00 Fixed dequant precision issues in Q4_1 and Q5_1 (#9711)
c83ad6d01e7b89ec71080d97c7e5db7ac1f4fda6 a39ab216aa624308fda7fa84439c6b61dc98b87a Diego Devesa slarengh@gmail.com 2024-10-03T01:49:47+02:00 GitHub noreply@github.com 2024-10-03T01:49:47+02:00 ggml-backend : add device and backend reg interfaces (#9707)
a39ab216aa624308fda7fa84439c6b61dc98b87a f536f4c4391bec74c432a924625c04e8c484d3ee Xuan Son Nguyen thichthat@gmail.com 2024-10-02T15:49:55+02:00 GitHub noreply@github.com 2024-10-02T15:49:55+02:00 llama : reduce compile time and binary size (#9712)
f536f4c4391bec74c432a924625c04e8c484d3ee 00b7317e636ffdc7dae59cb51dbd1cda357a3168 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-10-02T13:57:18+01:00 GitHub noreply@github.com 2024-10-02T13:57:18+01:00 [SYCL] Initial cmake support of SYCL for AMD GPUs (#9658)
00b7317e636ffdc7dae59cb51dbd1cda357a3168 76b37d1541a880b9645557c8715a343fd074cc5c Radoslav Gerganov rgerganov@gmail.com 2024-10-02T13:49:16+03:00 GitHub noreply@github.com 2024-10-02T13:49:16+03:00 vulkan : do not use tensor->extra (#9407)
76b37d1541a880b9645557c8715a343fd074cc5c 148844fe97fff4c1563a3111bf238ba4dd22ef56 Zhenwei Jin 109658203+kylo5aby@users.noreply.github.com 2024-10-02T15:21:57+08:00 GitHub noreply@github.com 2024-10-02T10:21:57+03:00 gguf-split : improve --split and --merge logic (#9619)
148844fe97fff4c1563a3111bf238ba4dd22ef56 3f1ae2e32cde00c39b96be6d01c2997c29bae555 Georgi Gerganov ggerganov@gmail.com 2024-10-02T10:14:44+03:00 GitHub noreply@github.com 2024-10-02T10:14:44+03:00 examples : remove benchmark (#9704)
3f1ae2e32cde00c39b96be6d01c2997c29bae555 f1b8c4271125c2bfb9cfebd72a8b9e9f99061a30 Paweł Wodnicki 151604+32bitmicro@users.noreply.github.com 2024-10-01T12:18:46-05:00 GitHub noreply@github.com 2024-10-01T19:18:46+02:00 Update README.md (#9591)
f1b8c4271125c2bfb9cfebd72a8b9e9f99061a30 e98c1c188ebbeb55d0cd6389ad03f0cbf995b451 Georgi Gerganov ggerganov@gmail.com 2024-10-01T16:09:42+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-01T16:09:42+03:00 sync : ggml
e98c1c188ebbeb55d0cd6389ad03f0cbf995b451 cb00020504416606601f8cb35f55ee07b710b4b7 Johannes Gäßler johannesg@5d6.de 2024-09-30T09:55:23+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-01T16:07:40+03:00 test: fix OPT_STEP_ADAMW for test-backend-ops (ggml/974)
cb00020504416606601f8cb35f55ee07b710b4b7 6c5322481a75f1be54e58a000c3f78484d07f948 Salvatore Mesoraca s.mesoraca16@gmail.com 2024-09-30T09:14:09+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-01T16:07:39+03:00 vulkan : mul_mat: fix UB with small warps (ggml/952)
6c5322481a75f1be54e58a000c3f78484d07f948 7254cdf7e8d6312840509ca8d766358c687c6266 Borislav Stanimirov b.stanimirov@abv.bg 2024-09-30T10:11:41+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-01T16:07:39+03:00 ggml : fix ggml_cast (ggml/973)
7254cdf7e8d6312840509ca8d766358c687c6266 cad341d88924788b940997c55e7bef000c99e784 Johannes Gäßler johannesg@5d6.de 2024-09-29T23:18:02+02:00 Georgi Gerganov ggerganov@gmail.com 2024-10-01T16:07:38+03:00 ggml: fix gradient allocation logic (ggml/966)
cad341d88924788b940997c55e7bef000c99e784 a90484c6d9db699bf739d0f33daf1c50cbdd45c9 Georgi Gerganov ggerganov@gmail.com 2024-10-01T16:00:25+03:00 GitHub noreply@github.com 2024-10-01T16:00:25+03:00 metal : reduce command encoding overhead (#9698)
a90484c6d9db699bf739d0f33daf1c50cbdd45c9 1927378bcce20ba72b6c89d5977b854a4bcaeb5d Georgi Gerganov ggerganov@gmail.com 2024-10-01T11:42:01+03:00 Georgi Gerganov ggerganov@gmail.com 2024-10-01T11:42:01+03:00 llama : print correct model type for Llama 3.2 1B and 3B
1927378bcce20ba72b6c89d5977b854a4bcaeb5d 6f1d9d71f4c568778a7637ff6582e6f6ba5fb9d3 compilade git@compilade.net 2024-10-01T02:31:36-04:00 GitHub noreply@github.com 2024-10-01T09:31:36+03:00 convert : refactor rope_freqs generation (#9396)
6f1d9d71f4c568778a7637ff6582e6f6ba5fb9d3 511636df0c90826b4dd1fc21ff260c19d69a3b5d serhii-nakon 57632032+serhii-nakon@users.noreply.github.com 2024-09-30T21:57:12+03:00 GitHub noreply@github.com 2024-09-30T20:57:12+02:00 Fix Docker ROCM builds, use AMDGPU_TARGETS instead of GPU_TARGETS (#9641)
511636df0c90826b4dd1fc21ff260c19d69a3b5d 08a43d05b6ba74de97610ae519450ad9996475e0 compilade git@compilade.net 2024-09-30T14:13:16-04:00 GitHub noreply@github.com 2024-09-30T14:13:16-04:00 ci : reduce severity of unused Pyright ignore comments (#9697)
08a43d05b6ba74de97610ae519450ad9996475e0 ace4f4be37abed4801fbd54a94cf38a7ae462416 vb vaibhavs10@gmail.com 2024-09-30T17:03:47+02:00 GitHub noreply@github.com 2024-09-30T18:03:47+03:00 py : update transfomers version (#9694)
ace4f4be37abed4801fbd54a94cf38a7ae462416 8277a817f18967581b02b2248989d773e8e99998 Georgi Gerganov ggerganov@gmail.com 2024-09-30T17:48:49+03:00 GitHub noreply@github.com 2024-09-30T07:48:49-07:00 flake.lock: Update (#9680)
8277a817f18967581b02b2248989d773e8e99998 c919d5db39c8a7fcb64737f008e4b105ee0acd20 Ruchira Hasaranga ruchira66@gmail.com 2024-09-30T13:53:42+05:30 GitHub noreply@github.com 2024-09-30T11:23:42+03:00 console : utf-8 fix for windows stdin (#9690)
c919d5db39c8a7fcb64737f008e4b105ee0acd20 d0b1d663e430354ab35853a6e1bce51cc8819376 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:18:23+03:00 GitHub noreply@github.com 2024-09-29T21:18:23+03:00 ggml : define missing HWCAP flags (#9684)
d0b1d663e430354ab35853a6e1bce51cc8819376 aaa40999251f5d18309b3fddc5a7d576f5fdb4e1 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:16:07+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:16:07+03:00 sync : ggml
aaa40999251f5d18309b3fddc5a7d576f5fdb4e1 641002fba8d2a0c0269027e23d2ef58e90546028 Johannes Gäßler johannesg@5d6.de 2024-09-29T19:56:17+02:00 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:15:37+03:00 CUDA: remove bad assert (ggml/972)
641002fba8d2a0c0269027e23d2ef58e90546028 0de8b203f1d31cf5ee0d2a3560a0ad78d44f4d4c Jeff Bolz jbolz@nvidia.com 2024-09-29T11:50:17-05:00 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:15:37+03:00 vulkan : multithread pipeline creation (ggml/963)
0de8b203f1d31cf5ee0d2a3560a0ad78d44f4d4c 544f409b4bd8fc98a3e87820f0ac934e00402de7 Jeff Bolz jbolz@nvidia.com 2024-09-27T02:58:01-05:00 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:15:37+03:00 vulkan : fix build for GGML_VULKAN_RUN_TESTS, add TFLOPS to log (ggml/961)
544f409b4bd8fc98a3e87820f0ac934e00402de7 6084bfb261b03f812de2255b05b6b5bb8d1c7171 Salvatore Mesoraca s.mesoraca16@gmail.com 2024-09-26T08:59:42+02:00 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:15:37+03:00 vulkan : argsort barriers must be under uniform control flow (ggml/951)
6084bfb261b03f812de2255b05b6b5bb8d1c7171 faac0bae265449fd988c57bf894018edc36fbe1e Georgi Gerganov ggerganov@gmail.com 2024-09-24T13:23:59+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-29T21:15:35+03:00 ggml : fix GGML_MAX_N_THREADS + improve formatting (ggml/969)
faac0bae265449fd988c57bf894018edc36fbe1e f99d3f8367174f7aba73c07fd87de687d4a0ece1 matiaslin 45382001+matiaslin@users.noreply.github.com 2024-09-29T05:25:00-07:00 GitHub noreply@github.com 2024-09-29T15:25:00+03:00 common : ensure llama_batch size does not exceed max size (#9668)
f99d3f8367174f7aba73c07fd87de687d4a0ece1 589b48d41efb0e95133b77c335f4fb9779af9bfb nopperl 54780682+nopperl@users.noreply.github.com 2024-09-29T12:02:06Z GitHub noreply@github.com 2024-09-29T15:02:06+03:00 py : add model class for Chameleon conversion (#9683)
589b48d41efb0e95133b77c335f4fb9779af9bfb f4d2b8846a6b34419ff9e9491aee6cd95e444bfc Georgi Gerganov ggerganov@gmail.com 2024-09-29T14:38:18+03:00 GitHub noreply@github.com 2024-09-29T14:38:18+03:00 contrib : add Resources section (#9675)
f4d2b8846a6b34419ff9e9491aee6cd95e444bfc 1b2f992cd2cff0b69e5abe78bb8888d51ed19d67 Georgi Gerganov ggerganov@gmail.com 2024-09-28T17:42:03+03:00 GitHub noreply@github.com 2024-09-28T17:42:03+03:00 llama : add reranking support (#9510)
1b2f992cd2cff0b69e5abe78bb8888d51ed19d67 739842703e32cd43443c45e0b4f6647cc4e6b3d6 slaren slarengh@gmail.com 2024-09-28T14:32:46+02:00 GitHub noreply@github.com 2024-09-28T14:32:46+02:00 test-backend-ops : use flops for some performance tests (#9657)
739842703e32cd43443c45e0b4f6647cc4e6b3d6 6102037bbb55521880ae78a6ee6c2a0c00c901df Georgi Gerganov ggerganov@gmail.com 2024-09-28T15:13:21+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-28T15:13:42+03:00 llama : add comment about thread-safety [no ci] (#9449)
6102037bbb55521880ae78a6ee6c2a0c00c901df 9a913110cf471a8287ac06c43cbe307d3cf6df99 Zhenwei Jin 109658203+kylo5aby@users.noreply.github.com 2024-09-28T20:10:58+08:00 GitHub noreply@github.com 2024-09-28T15:10:58+03:00 vocab : refactor tokenizer to reduce init overhead (#9449)
9a913110cf471a8287ac06c43cbe307d3cf6df99 43bcdd9703ec19af7d2a519640b5ed6f4aac3d53 nopperl 54780682+nopperl@users.noreply.github.com 2024-09-28T12:08:43Z GitHub noreply@github.com 2024-09-28T15:08:43+03:00 llama : add support for Chameleon (#8543)
43bcdd9703ec19af7d2a519640b5ed6f4aac3d53 6a0f7794847244fb3b99a983e03137d7e832b585 Aarni Koskela akx@iki.fi 2024-09-28T15:07:14+03:00 GitHub noreply@github.com 2024-09-28T15:07:14+03:00 readme : add tool (#9655)
6a0f7794847244fb3b99a983e03137d7e832b585 89f9944981010d195e411a9fbfbb19959412f710 Dan Johansson 164997844+eddnjjn@users.noreply.github.com 2024-09-28T14:06:16+02:00 GitHub noreply@github.com 2024-09-28T15:06:16+03:00 ggml : add run-time detection of neon, i8mm and sve (#9331)
89f9944981010d195e411a9fbfbb19959412f710 b5de3b74a595cbfefab7eeb5a567425c6a9690cf Markus Tavenrath mtavenrath@users.noreply.github.com 2024-09-28T12:05:05+02:00 GitHub noreply@github.com 2024-09-28T12:05:05+02:00 Enable use to the rebar feature to upload buffers to the device. (#9251)
b5de3b74a595cbfefab7eeb5a567425c6a9690cf 44f59b4301c51f071daa2e951301bb17c14acc9b Georgi Gerganov ggerganov@gmail.com 2024-09-27T20:57:51+03:00 GitHub noreply@github.com 2024-09-27T20:57:51+03:00 readme : update hot topics
44f59b4301c51f071daa2e951301bb17c14acc9b 95bc82fbc0df6d48cf66c857a4dda3d044f45ca2 Borislav Stanimirov b.stanimirov@abv.bg 2024-09-27T10:42:06+03:00 GitHub noreply@github.com 2024-09-27T10:42:06+03:00 cmake : add option for common library (#9661)
95bc82fbc0df6d48cf66c857a4dda3d044f45ca2 7691654c68aa5316108f881136c59f815ccb6809 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-09-26T17:38:31+08:00 GitHub noreply@github.com 2024-09-26T17:38:31+08:00 [SYCL] add missed dll file in package (#9577)
7691654c68aa5316108f881136c59f815ccb6809 ea9c32be71b91b42ecc538bd902e93cbb5fb36cb R0CKSTAR xiaodong.ye@mthreads.com 2024-09-26T09:27:40+08:00 GitHub noreply@github.com 2024-09-26T03:27:40+02:00 mtgpu: enable VMM (#9597)
ea9c32be71b91b42ecc538bd902e93cbb5fb36cb 1e436302188a704ac9ea044af03193648806f19c Xuan Son Nguyen thichthat@gmail.com 2024-09-25T17:26:01+02:00 GitHub noreply@github.com 2024-09-25T17:26:01+02:00 ci : fix docker build number and tag name (#9638)
1e436302188a704ac9ea044af03193648806f19c afbbfaa537a96f562c34df4542930fa951b40d9e Charles Xu 63788048+chaxu01@users.noreply.github.com 2024-09-25T15:12:20+02:00 GitHub noreply@github.com 2024-09-25T16:12:20+03:00 ggml : remove assert for AArch64 GEMV and GEMM Q4 kernels (#9217)
afbbfaa537a96f562c34df4542930fa951b40d9e 3d6bf6919f7b10726421779cd344f2da05421c68 Xuan Son Nguyen thichthat@gmail.com 2024-09-25T14:05:13+02:00 GitHub noreply@github.com 2024-09-25T14:05:13+02:00 server : add more env vars, improve gen-docs (#9635)
3d6bf6919f7b10726421779cd344f2da05421c68 904837e0cb2f5f01bf5d5901b7aa57a026860ae4 Gabe Goodhart ghart@us.ibm.com 2024-09-25T01:06:52-06:00 GitHub noreply@github.com 2024-09-25T10:06:52+03:00 llama : add IBM Granite MoE architecture (#9438)
904837e0cb2f5f01bf5d5901b7aa57a026860ae4 70392f1f81470607ba3afef04aa56c9f65587664 Dou Xinpeng 15529241576@163.com 2024-09-25T11:30:38+08:00 GitHub noreply@github.com 2024-09-25T11:30:38+08:00 cann: fix crash when llama-bench is running on multiple cann devices (#9627)
70392f1f81470607ba3afef04aa56c9f65587664 bb5f8199754b5f055cf436711d14c58e7be28e12 Eric Zhang 34133756+EZForever@users.noreply.github.com 2024-09-24T16:03:21+08:00 GitHub noreply@github.com 2024-09-24T11:03:21+03:00 ggml : add AVX512DQ requirement for AVX512 builds (#9622)
bb5f8199754b5f055cf436711d14c58e7be28e12 c038931615d2525d732943fa8661e29aa361b192 Georgi Gerganov ggerganov@gmail.com 2024-09-24T11:01:18+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-24T11:01:18+03:00 sync : ggml
c038931615d2525d732943fa8661e29aa361b192 31ac5834fe75c296476658a124c06c84772aa641 Georgi Gerganov ggerganov@gmail.com 2024-09-20T21:50:16+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-24T11:00:52+03:00 examples : adapt to ggml.h changes (ggml/0)
31ac5834fe75c296476658a124c06c84772aa641 cea1486ecf34a1c7e014a9e290eb458f5a11f876 Georgi Gerganov ggerganov@gmail.com 2024-09-24T10:16:06+03:00 GitHub noreply@github.com 2024-09-24T10:16:06+03:00 llama : keep track of all EOG tokens in the vocab (#9609)
cea1486ecf34a1c7e014a9e290eb458f5a11f876 0aa15011e315659640504731d1d05663837130fa Georgi Gerganov ggerganov@gmail.com 2024-09-24T10:15:35+03:00 GitHub noreply@github.com 2024-09-24T10:15:35+03:00 log : add CONT level for continuing previous log entry (#9610)
0aa15011e315659640504731d1d05663837130fa b0f27361f3539a81d983a8b045f3c61e682d9fc0 StrangeBytesDev 141275258+StrangeBytesDev@users.noreply.github.com 2024-09-23T23:04:39-07:00 GitHub noreply@github.com 2024-09-24T09:04:39+03:00 server : add newline after chat example (#9616)
b0f27361f3539a81d983a8b045f3c61e682d9fc0 c087b6f11d3385f4293b6841ebfb755063479490 Georgi Gerganov ggerganov@gmail.com 2024-09-24T09:03:17+03:00 GitHub noreply@github.com 2024-09-24T09:03:17+03:00 sampling : avoid expensive softmax during greedy sampling (#9605)
c087b6f11d3385f4293b6841ebfb755063479490 116efee0eef09d8c3c4c60b52fa01b56ddeb432c Max Krasnyansky quic_maxk@quicinc.com 2024-09-23T21:18:48-07:00 GitHub noreply@github.com 2024-09-23T21:18:48-07:00 threads: fix msvc build without openmp (#9615)
116efee0eef09d8c3c4c60b52fa01b56ddeb432c 0b3bf966f47bf2ba88e5d4e3ed429602008c7e63 Ivan nekotekina@gmail.com 2024-09-24T03:14:24+03:00 GitHub noreply@github.com 2024-09-24T02:14:24+02:00 cuda: add q8_0->f32 cpy operation (#9571)
0b3bf966f47bf2ba88e5d4e3ed429602008c7e63 f0c7b5edf82aa200656fd88c11ae3a805d7130bf Xuan Son Nguyen thichthat@gmail.com 2024-09-23T22:23:54+02:00 GitHub noreply@github.com 2024-09-23T22:23:54+02:00 server : add --no-context-shift option (#9607)
f0c7b5edf82aa200656fd88c11ae3a805d7130bf 1d48e98e4f3316bd2f6b187d288c7b6cb88d5cb3 Max Krasnyansky quic_maxk@quicinc.com 2024-09-23T11:42:43-07:00 GitHub noreply@github.com 2024-09-23T11:42:43-07:00 threads: improve ggml_barrier scaling with large number of threads (#9598)
1d48e98e4f3316bd2f6b187d288c7b6cb88d5cb3 f3979df762b75a2b1c0f622a2cd15d1bc60f037f Riceball LEE snowyu.lee@gmail.com 2024-09-23T23:58:17+08:00 GitHub noreply@github.com 2024-09-23T18:58:17+03:00 readme : add programmable prompt engine language CLI (#9599)
f3979df762b75a2b1c0f622a2cd15d1bc60f037f 1e7b9299c6ccb5bbc55d3db7cfa9b51f3ab09b59 Georgi Gerganov ggerganov@gmail.com 2024-09-23T18:43:40+03:00 GitHub noreply@github.com 2024-09-23T08:43:40-07:00 flake.lock: Update (#9586)
1e7b9299c6ccb5bbc55d3db7cfa9b51f3ab09b59 37f8c7b4c97784496cfd91040d55fa22f50b1d57 Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-09-23T19:36:38+05:30 GitHub noreply@github.com 2024-09-23T17:06:38+03:00 ggml : AVX512 gemm for Q4_0_8_8 (#9532)
37f8c7b4c97784496cfd91040d55fa22f50b1d57 bf9c1013ac40e5f1bd8e60b6d8bf16e0e8401445 Georgi Gerganov ggerganov@gmail.com 2024-09-23T11:28:02+03:00 GitHub noreply@github.com 2024-09-23T11:28:02+03:00 perplexity : remove extra new lines after chunks (#9596)
bf9c1013ac40e5f1bd8e60b6d8bf16e0e8401445 e62e9789cda3bf5573a747e55ec2a7ee32908f56 Georgi Gerganov ggerganov@gmail.com 2024-09-23T11:27:47+03:00 GitHub noreply@github.com 2024-09-23T11:27:47+03:00 metal : use F32 prec for K*Q in vec FA (#9595)
e62e9789cda3bf5573a747e55ec2a7ee32908f56 c35e586ea57221844442c65a1172498c54971cb0 Akarshan Biswas akarshanbiswas@fedoraproject.org 2024-09-23T08:58:06+05:30 GitHub noreply@github.com 2024-09-23T11:28:06+08:00 Revert "[SYCL] fallback mmvq (#9088)" (#9579)
c35e586ea57221844442c65a1172498c54971cb0 912c331d3dba3a079815844208dc36164baa8cc7 R0CKSTAR xiaodong.ye@mthreads.com 2024-09-22T22:55:49+08:00 GitHub noreply@github.com 2024-09-22T16:55:49+02:00 musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (#9526)
912c331d3dba3a079815844208dc36164baa8cc7 a5b57b08ce1998f7046df75324e86b9e2561c7af Molly Sophia mollysophia379@gmail.com 2024-09-22T21:26:50+08:00 GitHub noreply@github.com 2024-09-22T15:26:50+02:00 Fix merge error in #9454 (#9589)
a5b57b08ce1998f7046df75324e86b9e2561c7af ecd5d6b65be08927e62de1587d5fd22778cdc250 Johannes Gäßler johannesg@5d6.de 2024-09-22T09:34:52+02:00 GitHub noreply@github.com 2024-09-22T09:34:52+02:00 CUDA: enable Gemma FA for HIP/Pascal (#9581)
ecd5d6b65be08927e62de1587d5fd22778cdc250 2a63caaa69fad6c2afddc47bae052cf2afb01529 Shankar gshankar.87@gmail.com 2024-09-21T19:30:34-07:00 GitHub noreply@github.com 2024-09-22T04:30:34+02:00 llama: remove redundant loop when constructing ubatch (#9574)
2a63caaa69fad6c2afddc47bae052cf2afb01529 d09770cae71b416c032ec143dda530f7413c4038 Molly Sophia mollysophia379@gmail.com 2024-09-22T10:29:12+08:00 GitHub noreply@github.com 2024-09-22T04:29:12+02:00 RWKV v6: RWKV_WKV op CUDA implementation (#9454)
d09770cae71b416c032ec143dda530f7413c4038 41f477879fd5ccc31211634292e8e293ec700e85 slaren slarengh@gmail.com 2024-09-21T14:24:23+02:00 GitHub noreply@github.com 2024-09-21T14:24:23+02:00 ggml-alloc : fix list of allocated tensors with GGML_ALLOCATOR_DEBUG (#9573)
41f477879fd5ccc31211634292e8e293ec700e85 e948a7da7af7f2dbfdcd695b3ba903ab12575f78 agray3 agray3@users.noreply.github.com 2024-09-21T01:41:07+01:00 GitHub noreply@github.com 2024-09-21T02:41:07+02:00 Update CUDA graph on scale change plus clear nodes/params (#9550)
e948a7da7af7f2dbfdcd695b3ba903ab12575f78 63351143b2ea5efe9f8b9c61f553af8a51f1deff Huang Qi huangqi3@xiaomi.com 2024-09-21T08:39:41+08:00 GitHub noreply@github.com 2024-09-21T02:39:41+02:00 CI: Provide prebuilt windows binary for hip (#9467)
63351143b2ea5efe9f8b9c61f553af8a51f1deff d13edb17ed1ce3b961016cbdb616b1c8d161c026 slaren slarengh@gmail.com 2024-09-20T20:55:36+02:00 GitHub noreply@github.com 2024-09-20T20:55:36+02:00 quantize : improve type name parsing (#9570)
d13edb17ed1ce3b961016cbdb616b1c8d161c026 27609c49b94766a136764ce7919c8a082bf71548 Georgi Gerganov ggerganov@gmail.com 2024-09-20T20:12:52+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-20T21:15:05+03:00 ggml : fix builds (#0)
27609c49b94766a136764ce7919c8a082bf71548 43015353262de835215103475055b74045cb9f49 Georgi Gerganov ggerganov@gmail.com 2024-09-20T19:13:02+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-20T21:15:05+03:00 ggml : fix trailing whitespace (#0)
43015353262de835215103475055b74045cb9f49 424c5d00a9b97dd5559635872db9b57f87c23b02 Georgi Gerganov ggerganov@gmail.com 2024-09-20T19:06:59+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-20T21:15:05+03:00 sync : ggml
424c5d00a9b97dd5559635872db9b57f87c23b02 a6809c6a2e8163cba296d4cc0c5cd4bbc8073638 Johannes Gäßler johannesg@5d6.de 2024-09-20T19:04:44+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-20T21:15:05+03:00 ggml/examples: add backend support for numerical optimization (ggml/949)
a6809c6a2e8163cba296d4cc0c5cd4bbc8073638 5cb12f68395a5ec00b357e408883ce124a11dcdb Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:10:43+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-20T21:15:05+03:00 examples : add null threadpool args where needed (ggml/0)
5cb12f68395a5ec00b357e408883ce124a11dcdb d39e26741f9f02340651dbc640c9776e1a1128ef Johannes Gäßler johannesg@5d6.de 2024-09-20T18:35:35+02:00 GitHub noreply@github.com 2024-09-20T18:35:35+02:00 CUDA: fix sum.cu compilation for CUDA < 11.7 (#9562)
d39e26741f9f02340651dbc640c9776e1a1128ef 722ec1eb51ed53be2ab1ef31c6a1da8261803c71 Georgi Gerganov ggerganov@gmail.com 2024-09-20T11:46:56+03:00 GitHub noreply@github.com 2024-09-20T11:46:56+03:00 examples : flush log upon ctrl+c (#9559)
722ec1eb51ed53be2ab1ef31c6a1da8261803c71 6026da52d6942b253df835070619775d849d0258 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-09-20T08:38:10+02:00 GitHub noreply@github.com 2024-09-20T09:38:10+03:00 perplexity : do not escape input data by default (#9548)
6026da52d6942b253df835070619775d849d0258 eca0fab44eb449ed34e17a9b651ae7398b494117 Georgi Gerganov ggerganov@gmail.com 2024-09-19T12:44:53+03:00 GitHub noreply@github.com 2024-09-19T12:44:53+03:00 server : clean-up completed tasks from waiting list (#9531)
eca0fab44eb449ed34e17a9b651ae7398b494117 64c6af3195c3cd4aa3328a1282d29cd2635c34c9 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-09-19T09:58:14+02:00 GitHub noreply@github.com 2024-09-19T10:58:14+03:00 imatrix : disable prompt escape by default (#9543)
64c6af3195c3cd4aa3328a1282d29cd2635c34c9 0d2f22e45c3c3b6f8222acb6284d0c8c93443ba1 slaren slarengh@gmail.com 2024-09-18T19:13:08+02:00 GitHub noreply@github.com 2024-09-18T10:13:08-07:00 ggml : fix n_threads_cur initialization with one thread (#9538)
0d2f22e45c3c3b6f8222acb6284d0c8c93443ba1 6443ddd98576a9da904ef9f07df4e4398bb6a01a Georgi Gerganov ggerganov@gmail.com 2024-09-18T18:34:32+03:00 GitHub noreply@github.com 2024-09-18T18:34:32+03:00 scripts : verify py deps at the start of compare (#9520)
6443ddd98576a9da904ef9f07df4e4398bb6a01a 8a308354f6520df6bea851b435bd8054ee5617b4 Daniel Bevenius daniel.bevenius@gmail.com 2024-09-18T13:42:36+02:00 GitHub noreply@github.com 2024-09-18T14:42:36+03:00 llama : use reserve/emplace_back in sampler_sample (#9534)
8a308354f6520df6bea851b435bd8054ee5617b4 f799155ab8279cfa668086ce584aa52ecc05f5e5 Vinesh Janarthanan 36610342+VJHack@users.noreply.github.com 2024-09-18T01:50:34-05:00 GitHub noreply@github.com 2024-09-18T09:50:34+03:00 server : match OAI structured output response (#9527)
f799155ab8279cfa668086ce584aa52ecc05f5e5 faf67b3de4688f47c3b1019c89df255df2fd59b4 Eric Zhang 34133756+EZForever@users.noreply.github.com 2024-09-18T14:28:20+08:00 GitHub noreply@github.com 2024-09-18T09:28:20+03:00 server : fix OpenSSL build (remove obsolete `LOG_INFO`) (#9529)
faf67b3de4688f47c3b1019c89df255df2fd59b4 7be099fa817e9c53ffb4c3ed7d063e1cffcd675a Neo Zhang Jianyu jianyu.zhang@intel.com 2024-09-18T08:30:31+08:00 GitHub noreply@github.com 2024-09-18T08:30:31+08:00 [SYCL]set context default value to avoid memory issue, update guide (#9476)
7be099fa817e9c53ffb4c3ed7d063e1cffcd675a 8b836ae731bbb2c5640bc47df5b0a78ffcb129cb Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-17T22:41:38+02:00 GitHub noreply@github.com 2024-09-17T22:41:38+02:00 llama-bench: correct argument parsing error message (#9524)
8b836ae731bbb2c5640bc47df5b0a78ffcb129cb 8344ef58f8cdb8ebd6faf4463ca32ae91c374c81 Bert Wagner github@bertwagner.com 2024-09-17T09:35:38-04:00 GitHub noreply@github.com 2024-09-17T16:35:38+03:00 arg : add env variable for parallel (#9513)
8344ef58f8cdb8ebd6faf4463ca32ae91c374c81 0226613853133c081b55bb892a41bb5eacc0bc94 Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-17T12:18:22+02:00 GitHub noreply@github.com 2024-09-17T13:18:22+03:00 llama : fix n_vocab init for 'no_vocab' case (#9511)
0226613853133c081b55bb892a41bb5eacc0bc94 503147a9f9d195d6a14e7c998df23b6eb61f2bae Max Krasnyansky quic_maxk@quicinc.com 2024-09-17T01:19:46-07:00 GitHub noreply@github.com 2024-09-17T11:19:46+03:00 threadpool : skip polling for unused threads (#9461)
503147a9f9d195d6a14e7c998df23b6eb61f2bae 0d2ec438330271d201c2e9224aca23d0d5c908bf Yuri Khrustalev ykhrustalev@users.noreply.github.com 2024-09-17T02:51:15-04:00 GitHub noreply@github.com 2024-09-17T09:51:15+03:00 unicode : add <algorithm> (#9508)
0d2ec438330271d201c2e9224aca23d0d5c908bf 37f3a3810e545be6ac835c726f97956c1403ea02 Gabe Goodhart ghart@us.ibm.com 2024-09-17T00:44:58-06:00 GitHub noreply@github.com 2024-09-17T09:44:58+03:00 llama : support IBM Granite architecture (#9412)
37f3a3810e545be6ac835c726f97956c1403ea02 23e0d70bacaaca1429d365a44aa9e7434f17823b Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-17T08:23:30+02:00 GitHub noreply@github.com 2024-09-17T09:23:30+03:00 llama : add llama_n_head() (#9512)
23e0d70bacaaca1429d365a44aa9e7434f17823b acb2c32c336ce60d765bb189563cc216e57e9fc2 slaren slarengh@gmail.com 2024-09-16T16:22:07+02:00 GitHub noreply@github.com 2024-09-16T16:22:07+02:00 ggml : move common CPU backend impl to new header (#9509)
acb2c32c336ce60d765bb189563cc216e57e9fc2 a6a3a5c531c73aef85750a847d21e7d4671e723d Daniel Bevenius daniel.bevenius@gmail.com 2024-09-16T13:07:13+02:00 GitHub noreply@github.com 2024-09-16T14:07:13+03:00 llama : rename n_embed to n_embd in rwkv6_time_mix (#9504)
a6a3a5c531c73aef85750a847d21e7d4671e723d d54c21df7e2669c6cd7492713479d1aeb5846883 Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-16T13:06:50+02:00 GitHub noreply@github.com 2024-09-16T14:06:50+03:00 ggml : link MATH_LIBRARY not by its full path (#9339)
d54c21df7e2669c6cd7492713479d1aeb5846883 19514d632e5274bc0c27c2269e8f2ad88b526d62 compilade git@compilade.net 2024-09-16T03:30:22-04:00 GitHub noreply@github.com 2024-09-16T10:30:22+03:00 convert : identify missing model files (#9397)
19514d632e5274bc0c27c2269e8f2ad88b526d62 5c3d0f1824714e9a97fc9b06e046eefcb6ecc721 Georgi Gerganov ggerganov@gmail.com 2024-09-16T10:27:50+03:00 GitHub noreply@github.com 2024-09-16T10:27:50+03:00 cmake : do not hide GGML options + rename option (#9465)
5c3d0f1824714e9a97fc9b06e046eefcb6ecc721 0aadac10c7dd704f8285ddf5a63d6f764cb340aa Eve 139727413+netrunnereve@users.noreply.github.com 2024-09-16T06:48:24Z GitHub noreply@github.com 2024-09-16T09:48:24+03:00 ggml : IQ4_NL sgemm + Q4_0 AVX optimization (#9422)
0aadac10c7dd704f8285ddf5a63d6f764cb340aa 95ca85168b5b089b80a811b59528ce0a2f1bd1dd Shane A shanea@allenai.org 2024-09-15T23:47:37-07:00 GitHub noreply@github.com 2024-09-16T09:47:37+03:00 llama : support OLMoE (#9462)
95ca85168b5b089b80a811b59528ce0a2f1bd1dd 441b72b91f818fe69497e5816f87969e90c73c43 CarryFun 76023481+CarryFun@users.noreply.github.com 2024-09-16T14:45:20+08:00 GitHub noreply@github.com 2024-09-16T09:45:20+03:00 llama : support MiniCPM3 (#9322)
441b72b91f818fe69497e5816f87969e90c73c43 c4965a64f72ac9434c21cf0e1c3421d13e889155 Vinesh Janarthanan 36610342+VJHack@users.noreply.github.com 2024-09-16T01:20:01-05:00 GitHub noreply@github.com 2024-09-16T09:20:01+03:00 main : option to disable context shift (#9484)
c4965a64f72ac9434c21cf0e1c3421d13e889155 90a2fff0e7f80c6fea3fc6cf9a7b482744f3f164 Georgi Gerganov ggerganov@gmail.com 2024-09-16T09:05:56+03:00 GitHub noreply@github.com 2024-09-16T09:05:56+03:00 metal : handle zero-sized allocs (#9466)
90a2fff0e7f80c6fea3fc6cf9a7b482744f3f164 6262d13e0b2da91f230129a93a996609a2f5a2f2 Georgi Gerganov ggerganov@gmail.com 2024-09-16T05:14:23+03:00 GitHub noreply@github.com 2024-09-15T19:14:23-07:00 flake.lock: Update (#9488)
6262d13e0b2da91f230129a93a996609a2f5a2f2 e6deac31f7e62db43b6afbc3be814f764fd5a187 Georgi Gerganov ggerganov@gmail.com 2024-09-15T20:46:12+03:00 GitHub noreply@github.com 2024-09-15T20:46:12+03:00 common : reimplement logging (#9418)
e6deac31f7e62db43b6afbc3be814f764fd5a187 6988da94a261444859f78595899212eeedc5ff9d slaren slarengh@gmail.com 2024-09-15T19:02:27+02:00 GitHub noreply@github.com 2024-09-15T19:02:27+02:00 gguf-split : add basic checks (#9499)
6988da94a261444859f78595899212eeedc5ff9d 3c7989fd29a2db2b75e28fd708cc441febe99a82 Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-15T18:55:52+02:00 GitHub noreply@github.com 2024-09-15T19:55:52+03:00 cmake : correct order of sycl flags (#9497)
3c7989fd29a2db2b75e28fd708cc441febe99a82 d6b37c881f056bd32b681dcd7658a37ea6ec3a1e Csaba Kecskemeti csaba.kecskemeti@gmail.com 2024-09-15T00:48:25-07:00 GitHub noreply@github.com 2024-09-15T10:48:25+03:00 py : add "LLaMAForCausalLM" conversion support (#9485)
d6b37c881f056bd32b681dcd7658a37ea6ec3a1e 7596487bebd58eade3cd0133d42a9008aaaf9d09 OSecret 135510162+OLSecret@users.noreply.github.com 2024-09-15T10:36:53+03:00 GitHub noreply@github.com 2024-09-15T10:36:53+03:00 readme : update tools list (#9475)
7596487bebd58eade3cd0133d42a9008aaaf9d09 822b6322dea704110797a5671fc80ae39ee6ac97 Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-15T09:06:38+02:00 GitHub noreply@github.com 2024-09-15T10:06:38+03:00 cmake : try to fix sycl+intel build (#9487)
822b6322dea704110797a5671fc80ae39ee6ac97 dcdcee3a744f39714503ee2b19c49b7c7b6209c9 Yuri Khrustalev ykhrustalev@users.noreply.github.com 2024-09-14T05:54:37-04:00 GitHub noreply@github.com 2024-09-14T12:54:37+03:00 ggml : ggml_type_name return "NONE" for invalid values (#9458)
dcdcee3a744f39714503ee2b19c49b7c7b6209c9 1f4111e540bacec8d00ca9fd96417bf4c1339394 VoidIsVoid 343750470@qq.com 2024-09-14T17:36:44+08:00 GitHub noreply@github.com 2024-09-14T11:36:44+02:00 server: add data: [DONE] to /chat/completions stream response (#9459)
1f4111e540bacec8d00ca9fd96417bf4c1339394 befaf1197fa447f61714de041828852a270659d2 Georgi Gerganov ggerganov@gmail.com 2024-09-14T10:55:05+03:00 GitHub noreply@github.com 2024-09-14T10:55:05+03:00 cmake : use list(APPEND ...) instead of set() + dedup linker (#9463)
befaf1197fa447f61714de041828852a270659d2 feff4aa8461da7c432d144c11da4802e41fef3cf Daniel Bevenius daniel.bevenius@gmail.com 2024-09-14T09:50:12+02:00 GitHub noreply@github.com 2024-09-14T10:50:12+03:00 llama : make cell_id const in inp_s_mask block (#9470)
feff4aa8461da7c432d144c11da4802e41fef3cf 0abc6a2c25272d5cf01384dda8ee8bfec4ba8745 Xuan Son Nguyen thichthat@gmail.com 2024-09-13T14:23:11+02:00 GitHub noreply@github.com 2024-09-13T14:23:11+02:00 server : add loading html page while model is loading (#9468)
0abc6a2c25272d5cf01384dda8ee8bfec4ba8745 bd35cb0ae357185c173345f10dc89a4ff925fc25 Georgi Gerganov ggerganov@gmail.com 2024-09-13T09:53:38+03:00 GitHub noreply@github.com 2024-09-13T09:53:38+03:00 llama : llama_perf + option to disable timings during decode (#9355)
bd35cb0ae357185c173345f10dc89a4ff925fc25 78203641fee3b1f82abaff0c7f667e1b4a286390 Gilad S. 7817232+giladgd@users.noreply.github.com 2024-09-13T04:54:49+03:00 GitHub noreply@github.com 2024-09-13T03:54:49+02:00 feat: remove a sampler from a chain (#9445)
78203641fee3b1f82abaff0c7f667e1b4a286390 e6b7801bd189d102d901d3e72035611a25456ef1 Mathijs Henquet mathijs.henquet@gmail.com 2024-09-12T22:30:11+02:00 GitHub noreply@github.com 2024-09-12T22:30:11+02:00 server : Add option to return token pieces in /tokenize endpoint (#9108)
e6b7801bd189d102d901d3e72035611a25456ef1 e665744317c77fc3483fc5224fe6d586b5166b33 Dou Xinpeng 81913537+Dou-Git@users.noreply.github.com 2024-09-12T19:46:43+08:00 GitHub noreply@github.com 2024-09-12T19:46:43+08:00 cann: Add host buffer type for Ascend NPU (#9406)
e665744317c77fc3483fc5224fe6d586b5166b33 d4c3c10fad1bd6adec72d2f1f236761a8d6a07f8 fengerhu1 2748250768@qq.com 2024-09-12T19:34:22+08:00 GitHub noreply@github.com 2024-09-12T14:34:22+03:00 llava : fix the script error in MobileVLM README (#9054)
d4c3c10fad1bd6adec72d2f1f236761a8d6a07f8 2a825116b6f7f3a9b1726e5e0c3eb22f7768bd33 Xuan Son Nguyen thichthat@gmail.com 2024-09-12T13:33:57+02:00 GitHub noreply@github.com 2024-09-12T14:33:57+03:00 lora : raise error if lm_head is ignored (#9103)
2a825116b6f7f3a9b1726e5e0c3eb22f7768bd33 4dc4f5f14ae522494649d82ad06b031cf9501038 Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-12T13:30:01+02:00 GitHub noreply@github.com 2024-09-12T14:30:01+03:00 cmake : fix for builds without `GGML_CDEF_PUBLIC` (#9338)
4dc4f5f14ae522494649d82ad06b031cf9501038 c837981bba7cf6839b69d32b25552ce685936b14 Huang Qi huangqi3@xiaomi.com 2024-09-12T19:28:43+08:00 GitHub noreply@github.com 2024-09-12T14:28:43+03:00 ci : update HIP SDK to 24.Q3 (ROCm 6.1) (#9329)
c837981bba7cf6839b69d32b25552ce685936b14 3c26a1644dacfa6b5d58af550210524efd7b93fc daminho 37615795+daminho@users.noreply.github.com 2024-09-12T20:28:20+09:00 GitHub noreply@github.com 2024-09-12T14:28:20+03:00 py : add Phi-1.5/Phi-2 tokenizer (#9361)
3c26a1644dacfa6b5d58af550210524efd7b93fc ff76e18516dbe269b35ba1bb500524ed5e39225c Trivikram Kamat 16024985+trivikr@users.noreply.github.com 2024-09-12T04:27:45-07:00 GitHub noreply@github.com 2024-09-12T14:27:45+03:00 ci : bump actions/checkout to v4 (#9377)
ff76e18516dbe269b35ba1bb500524ed5e39225c 39f852f44039b058fdd0611ee127c6efa7ba4a04 Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-12T13:27:14+02:00 GitHub noreply@github.com 2024-09-12T14:27:14+03:00 cmake : fixed the order of linking libraries for llama-quantize (#9450)
39f852f44039b058fdd0611ee127c6efa7ba4a04 2b00fa799773cc75d53b841c03d21d7468a1e3a1 Molly Sophia mollysophia379@gmail.com 2024-09-12T19:25:16+08:00 GitHub noreply@github.com 2024-09-12T14:25:16+03:00 py : add special tokens in hf_converter for RWKV v6 (#9428)
2b00fa799773cc75d53b841c03d21d7468a1e3a1 d6a04f872dea8ade92527bb1488d4b0b90cc49f0 Ahmad Tameem 113388789+Tameem-10xE@users.noreply.github.com 2024-09-12T16:24:31+05:00 GitHub noreply@github.com 2024-09-12T14:24:31+03:00 riscv : modify Makefile and add a RISCV_VECT to print log info (#9442)
d6a04f872dea8ade92527bb1488d4b0b90cc49f0 c9c8575a1a8a170329afca4c4df4c005806efb1d Georgi Gerganov ggerganov@gmail.com 2024-09-12T14:23:49+03:00 GitHub noreply@github.com 2024-09-12T14:23:49+03:00 ggml : hide ggml_object, ggml_cgraph, ggml_hash_set (#9408)
c9c8575a1a8a170329afca4c4df4c005806efb1d df4b7945aeccae2a71348e5a9c1eab5241e3e0ef Neo Zhang Jianyu jianyu.zhang@intel.com 2024-09-12T17:44:17+08:00 GitHub noreply@github.com 2024-09-12T17:44:17+08:00 enhance run script to be easy to change the parameters (#9448)
df4b7945aeccae2a71348e5a9c1eab5241e3e0ef 449ccfb6f5f1bbd70e04f75a330d9d7c1af82187 Xinpeng Dou 81913537+Dou-Git@users.noreply.github.com 2024-09-12T09:02:35+08:00 GitHub noreply@github.com 2024-09-12T09:02:35+08:00 cann: Fix error when running a non-exist op (#9424)
449ccfb6f5f1bbd70e04f75a330d9d7c1af82187 1b28061400eb9832603c9f1dfbec4d339a8490a2 Faisal Zaghloul quic_fzaghlou@quicinc.com 2024-09-11T20:29:53-04:00 GitHub noreply@github.com 2024-09-12T02:29:53+02:00 Add Jais to list of supported models (#9439)
1b28061400eb9832603c9f1dfbec4d339a8490a2 8db003a19d7055b5bd248ce2afff9324e5b8da95 slaren slarengh@gmail.com 2024-09-11T17:52:13+02:00 GitHub noreply@github.com 2024-09-11T17:52:13+02:00 llama : skip token bounds check when evaluating embeddings (#9437)
8db003a19d7055b5bd248ce2afff9324e5b8da95 0996c5597f680effacc046832bb807c14900e22d Pavel Zloi github.com@drteam.rocks 2024-09-11T15:29:51+03:00 GitHub noreply@github.com 2024-09-11T15:29:51+03:00 py : support converting local models (#7547)
0996c5597f680effacc046832bb807c14900e22d 5bb2c5dbd26b246d334f0087b3cbd800f2e65c54 Xuan Son Nguyen thichthat@gmail.com 2024-09-11T12:59:13+02:00 GitHub noreply@github.com 2024-09-11T12:59:13+02:00 llava : correct args for minicpmv-cli (#9429)
5bb2c5dbd26b246d334f0087b3cbd800f2e65c54 67155ab7f5e47c01b62aa989eab30f517bf6dc67 Xuan Son Nguyen thichthat@gmail.com 2024-09-11T12:02:09+02:00 GitHub noreply@github.com 2024-09-11T13:02:09+03:00 files : remove accidentally added `lora_test` submodule (#9430)
67155ab7f5e47c01b62aa989eab30f517bf6dc67 5af118efdaf1098798a06b24fd8a557760e99631 Farbod Bijary 110523279+farbodbj@users.noreply.github.com 2024-09-11T12:52:37+03:30 GitHub noreply@github.com 2024-09-11T11:22:37+02:00 feat: Implements retrying logic for downloading models using --model-url flag (#9255)
5af118efdaf1098798a06b24fd8a557760e99631 d2b496bff4f353a6429f8e833448f071bd237ba7 Johannes Gäßler johannesg@5d6.de 2024-09-11T10:22:40+02:00 GitHub noreply@github.com 2024-09-11T10:22:40+02:00 CUDA: fix --split-mode row race condition (#9413)
d2b496bff4f353a6429f8e833448f071bd237ba7 b34e02348064c2f0cef1f89b44d9bee4eb15b9e7 Georgi Gerganov ggerganov@gmail.com 2024-09-11T10:03:54+03:00 GitHub noreply@github.com 2024-09-11T10:03:54+03:00 batched-bench : remove unused code (#9305)
b34e02348064c2f0cef1f89b44d9bee4eb15b9e7 51b603863627c4074e77b7e556e18ece86bdf9a3 R0CKSTAR xiaodong.ye@mthreads.com 2024-09-11T09:46:55+08:00 GitHub noreply@github.com 2024-09-11T03:46:55+02:00 musa: remove Clang builtins mapping (#9421)
51b603863627c4074e77b7e556e18ece86bdf9a3 cb9c933eb2a0d2b514556bdcb934b56dfe5d6771 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-09-11T01:53:42+01:00 GitHub noreply@github.com 2024-09-11T08:53:42+08:00 sycl : update support conditions (#9394)
cb9c933eb2a0d2b514556bdcb934b56dfe5d6771 6cd4e034442f71718563e600070c2b6fc389e100 Georgi Gerganov ggerganov@gmail.com 2024-09-11T01:46:59+03:00 GitHub noreply@github.com 2024-09-10T15:46:59-07:00 flake.lock: Update (#9360)
6cd4e034442f71718563e600070c2b6fc389e100 8d300bd35fbe23b35a4e1ece0cf0fe8f43331029 Xuan Son Nguyen thichthat@gmail.com 2024-09-10T22:41:29+02:00 GitHub noreply@github.com 2024-09-10T22:41:29+02:00 arg : bring back missing ifdef (#9411)
8d300bd35fbe23b35a4e1ece0cf0fe8f43331029 49006c67b4c6cc2e7c75a875b4d6e161ebae287c matteo matteogeniaccio@yahoo.it 2024-09-10T22:40:59+02:00 GitHub noreply@github.com 2024-09-10T22:40:59+02:00 enable --special arg for llama-server (#9419)
49006c67b4c6cc2e7c75a875b4d6e161ebae287c 00ba2ff78100e187ae17987bacd1c916211718b2 slaren slarengh@gmail.com 2024-09-10T18:04:25+02:00 GitHub noreply@github.com 2024-09-10T18:04:25+02:00 llama : move random seed generation to the samplers (#9398)
00ba2ff78100e187ae17987bacd1c916211718b2 83008b7cfe90ad89d0c0ed2c2424fd75edc25ac1 Georgi Gerganov ggerganov@gmail.com 2024-09-10T10:17:03+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-10T10:17:43+03:00 metal : fix compile warning with GGML_METAL_NDEBUG (#0)
83008b7cfe90ad89d0c0ed2c2424fd75edc25ac1 0b4ac75772b744bb0a0d674927587621d1057884 Daniel Bevenius daniel.bevenius@gmail.com 2024-09-10T09:03:21+02:00 GitHub noreply@github.com 2024-09-10T10:03:21+03:00 llama : update llm_build_copy_mask_state comment [no ci] (#9385)
0b4ac75772b744bb0a0d674927587621d1057884 fb3f2498156b3140e2050ec9c7bf61372f63ff56 Molly Sophia mollysophia379@gmail.com 2024-09-10T15:02:30+08:00 GitHub noreply@github.com 2024-09-10T10:02:30+03:00 RWKV v6: Add time_mix_decay_w1/w2 in quant exclusion list (#9387)
fb3f2498156b3140e2050ec9c7bf61372f63ff56 bfe76d4a17228bfd1565761f203123bc4914771b slaren slarengh@gmail.com 2024-09-10T08:23:33+02:00 GitHub noreply@github.com 2024-09-10T09:23:33+03:00 make : do not run llama-gen-docs when building (#9399)
bfe76d4a17228bfd1565761f203123bc4914771b 293bebe0773c907c0c866213856eeba41b035df1 Xuan Son Nguyen thichthat@gmail.com 2024-09-09T23:36:09+02:00 GitHub noreply@github.com 2024-09-09T23:36:09+02:00 common : move arg parser code to `arg.cpp` (#9388)
293bebe0773c907c0c866213856eeba41b035df1 5fac4d57643b1de8e9ab746f14d2fc4e319ae0c2 Radoslav Gerganov rgerganov@gmail.com 2024-09-09T18:40:10+03:00 GitHub noreply@github.com 2024-09-09T18:40:10+03:00 rpc : fix segfault with nkvo (#9389)
5fac4d57643b1de8e9ab746f14d2fc4e319ae0c2 5fb5e24811cb01d48b482c15a974bfbd9f433e1d Prashant Vithule 119530321+Vithulep@users.noreply.github.com 2024-09-09T21:07:18+05:30 GitHub noreply@github.com 2024-09-09T18:37:18+03:00 ggml : vector length agnostic SVE support (#9290)
5fb5e24811cb01d48b482c15a974bfbd9f433e1d 38ca6f644bd48301e9caa80f9913c22e70a8fd1b slaren slarengh@gmail.com 2024-09-09T17:10:46+02:00 GitHub noreply@github.com 2024-09-09T17:10:46+02:00 llama : minor sampling refactor (2) (#9386)
38ca6f644bd48301e9caa80f9913c22e70a8fd1b 8e6e2fbe1458ac91387266241262294a964d6b95 Georgi Gerganov ggerganov@gmail.com 2024-09-09T15:51:37+03:00 GitHub noreply@github.com 2024-09-09T15:51:37+03:00 readme : update hot topics
8e6e2fbe1458ac91387266241262294a964d6b95 5ed087573e1f326cfa70e29c1895d074a7a1a00c Johannes Gäßler johannesg@5d6.de 2024-09-09T14:22:53+02:00 GitHub noreply@github.com 2024-09-09T14:22:53+02:00 CUDA: fix variable name conflict for Windows build (#9382)
5ed087573e1f326cfa70e29c1895d074a7a1a00c 54f376d0b92c6ff6feb1fa2ef8ed2022348100ba Antonis Makropoulos benuix@gmail.com 2024-09-09T14:21:38+03:00 GitHub noreply@github.com 2024-09-09T14:21:38+03:00 readme : add LLMUnity to UI projects (#9381)
54f376d0b92c6ff6feb1fa2ef8ed2022348100ba b2e89a327457179a34eae4d7de0d412ed945679c Radoslav Gerganov rgerganov@gmail.com 2024-09-09T11:04:39+03:00 GitHub noreply@github.com 2024-09-09T11:04:39+03:00 rpc : update README [no ci] (#9320)
b2e89a327457179a34eae4d7de0d412ed945679c daa9623ab051a8162ae750b150b9522571b55f21 Dan Johansson 164997844+eddnjjn@users.noreply.github.com 2024-09-09T09:02:45+02:00 GitHub noreply@github.com 2024-09-09T10:02:45+03:00 Arm AArch64: Documentation updates (#9321)
daa9623ab051a8162ae750b150b9522571b55f21 e079bffb6678e1b5ded21c719600bedd7175e726 Markus Tavenrath mtavenrath@users.noreply.github.com 2024-09-08T21:43:48+02:00 GitHub noreply@github.com 2024-09-08T21:43:48+02:00 Overlap cmdbuffer creation and cmdbuffer execution in Vulkan backend by submitting smaller cmdbuffers early. (#9118)
e079bffb6678e1b5ded21c719600bedd7175e726 3f7ccfd649abc83d059b5462221ac14de4ede6b7 Georgi Gerganov ggerganov@gmail.com 2024-09-08T22:01:02+03:00 GitHub noreply@github.com 2024-09-08T22:01:02+03:00 cuda : fix FA Q src index (1 -> 0) (#9374)
3f7ccfd649abc83d059b5462221ac14de4ede6b7 a249843d89bd6373772eede26d7f2aa708b26600 Xuan Son Nguyen thichthat@gmail.com 2024-09-08T18:08:55+02:00 GitHub noreply@github.com 2024-09-08T18:08:55+02:00 common : bring back missing args, add env var duplication check (#9375)
a249843d89bd6373772eede26d7f2aa708b26600 19f4a7b296efda7c13a6b21d428b2286b5d1aa06 slaren slarengh@gmail.com 2024-09-08T16:44:42+02:00 GitHub noreply@github.com 2024-09-08T16:44:42+02:00 common : restore --n-gpu-layers (#9371)
19f4a7b296efda7c13a6b21d428b2286b5d1aa06 2a358fb0c4b6e917ac852aa17444cc94dd28a2a6 slaren slarengh@gmail.com 2024-09-08T15:52:07+02:00 GitHub noreply@github.com 2024-09-08T15:52:07+02:00 llama : refactor samplers internal implementation (#9370)
2a358fb0c4b6e917ac852aa17444cc94dd28a2a6 eae597182cb61bbde0b26e7cec5999d28b9327fe Neo Zhang Jianyu jianyu.zhang@intel.com 2024-09-08T19:05:29+08:00 GitHub noreply@github.com 2024-09-08T19:05:29+08:00 [SYCL] add check malloc result on device (#9346)
eae597182cb61bbde0b26e7cec5999d28b9327fe 00b02bb249406ec0123757a67a5b714c3f33a699 slaren slarengh@gmail.com 2024-09-08T12:41:51+02:00 GitHub noreply@github.com 2024-09-08T12:41:51+02:00 llama : sanitize tokens in the upper bound (#9359)
00b02bb249406ec0123757a67a5b714c3f33a699 a8768614558262b6762fc0feeddcc6f7ce4bc5fc Xuan Son Nguyen thichthat@gmail.com 2024-09-08T12:12:17+02:00 GitHub noreply@github.com 2024-09-08T12:12:17+02:00 imatrix : fix arg parser for imatrix (#9366)
a8768614558262b6762fc0feeddcc6f7ce4bc5fc 385decbd632f70db9f1fbd93dbb2b908853e135c Georgi Gerganov ggerganov@gmail.com 2024-09-08T09:57:57+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 metal : update support condition for im2col + fix warning (#0)
385decbd632f70db9f1fbd93dbb2b908853e135c 60a3107ccd791d858e12a87e6024ce918d3bf595 Georgi Gerganov ggerganov@gmail.com 2024-09-08T09:38:56+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 sync : ggml
60a3107ccd791d858e12a87e6024ce918d3bf595 406c1a32a18807b9b5711aa2fa1859527106bc1d Georgi Gerganov ggerganov@gmail.com 2024-09-08T09:38:42+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 scripts : option to increase git patch context
406c1a32a18807b9b5711aa2fa1859527106bc1d 9cb9260861e464e40d38764cfb021856786c7202 Salvatore Mesoraca s.mesoraca16@gmail.com 2024-09-06T14:34:25+02:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 vulkan: add dryrun support to sin and cos ops (ggml/947)
9cb9260861e464e40d38764cfb021856786c7202 202084d31d4247764fc6d6d40d2e2bda0c89a73a Salvatore Mesoraca s.mesoraca16@gmail.com 2024-09-06T14:34:07+02:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 vulkan: correctly report support for OP_CONT (ggml/946)
202084d31d4247764fc6d6d40d2e2bda0c89a73a dbbebcab339c7d63d3806e8f32574bb9aad9a694 Johannes Gäßler johannesg@5d6.de 2024-09-03T17:21:46+02:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 tests: add gradient tests for all backends (ggml/932)
dbbebcab339c7d63d3806e8f32574bb9aad9a694 ba1cf846edeb94fc567a9ab0f7ef705282e6d7d3 Johannes Gäßler johannesg@5d6.de 2024-08-31T14:35:42+02:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 ggml: fix ggml_graph_cpy undefined behavior (ggml/943)
ba1cf846edeb94fc567a9ab0f7ef705282e6d7d3 d2d3200b385970cb2a4658fd9342a3b1212bdb46 Georgi Gerganov ggerganov@gmail.com 2024-08-28T18:45:01+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 cann : fix doxy (ggml/0)
d2d3200b385970cb2a4658fd9342a3b1212bdb46 51d964a4efdb0e8d43f5b85a775951185f6b641e Mengqing Cao cmq0113@163.com 2024-08-09T20:21:56+08:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 cann : add Ascend NPU support (whisper/2336)
51d964a4efdb0e8d43f5b85a775951185f6b641e efe6a83e3046a94cda38c8be5a7801eadc21d78d Georgi Gerganov ggerganov@gmail.com 2024-08-28T17:08:03+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 cuda : mark BF16 CONT as unsupported
efe6a83e3046a94cda38c8be5a7801eadc21d78d fbb7fcffbcfc50009c275e75982b1603a6cb6b80 Salvatore Mesoraca s.mesoraca16@gmail.com 2024-08-28T10:23:02+02:00 Georgi Gerganov ggerganov@gmail.com 2024-09-08T11:05:55+03:00 ggml : fix cont with transposed tensors when one dimension is 1 (ggml/934)
fbb7fcffbcfc50009c275e75982b1603a6cb6b80 a5b5d9a1014248f385939e6739e9db9de7147e55 Kevin Gibbons bakkot@gmail.com 2024-09-07T22:51:00-07:00 GitHub noreply@github.com 2024-09-08T08:51:00+03:00 llama : set attrs of mislabelled EOT/EOM tokens (#9348)
a5b5d9a1014248f385939e6739e9db9de7147e55 f12295b8a9336962bf02a359beb1be0d322b4be7 Georgi Gerganov ggerganov@gmail.com 2024-09-08T00:33:50+03:00 GitHub noreply@github.com 2024-09-08T00:33:50+03:00 llama.android : fix build (#9350)
f12295b8a9336962bf02a359beb1be0d322b4be7 faf69d4237c9ae4d7f572b4674d1002463e8acd3 Georgi Gerganov ggerganov@gmail.com 2024-09-08T00:33:33+03:00 GitHub noreply@github.com 2024-09-08T00:33:33+03:00 llama : fix empty ring buffer push (#9358)
faf69d4237c9ae4d7f572b4674d1002463e8acd3 e536426ded3fb4a8cd13626e53508cd92928d6c2 Georgi Gerganov ggerganov@gmail.com 2024-09-08T00:33:13+03:00 GitHub noreply@github.com 2024-09-08T00:33:13+03:00 llama : sanitize invalid tokens (#9357)
e536426ded3fb4a8cd13626e53508cd92928d6c2 1b9ae5189cd279c6b45e36d43e4f9ccae628d02f Eve 139727413+netrunnereve@users.noreply.github.com 2024-09-07T19:02:26Z GitHub noreply@github.com 2024-09-07T22:02:26+03:00 llamafile : disable sgemm for batch-size 1 (#9330)
1b9ae5189cd279c6b45e36d43e4f9ccae628d02f e32d0816edfd247784f6780b0bb9ae0bceef5e47 Xuan Son Nguyen thichthat@gmail.com 2024-09-07T20:43:51+02:00 GitHub noreply@github.com 2024-09-07T20:43:51+02:00 common : refactor arg parser (#9308)
e32d0816edfd247784f6780b0bb9ae0bceef5e47 df270ef74596da8f1178f08991f4c51f18c9ee82 slaren slarengh@gmail.com 2024-09-07T20:23:07+02:00 GitHub noreply@github.com 2024-09-07T20:23:07+02:00 ggml : always check bounds on get_rows operations (#9354)
df270ef74596da8f1178f08991f4c51f18c9ee82 947538acb8617756a092042ff7e58db18dde05ec Georgi Gerganov ggerganov@gmail.com 2024-09-07T15:16:19+03:00 GitHub noreply@github.com 2024-09-07T15:16:19+03:00 llama : refactor sampling v2 (#9294)
947538acb8617756a092042ff7e58db18dde05ec 6c89eb0b4749184f4d19e96ada5dcb8847129b9c Xuan Son Nguyen thichthat@gmail.com 2024-09-07T12:01:34+02:00 GitHub noreply@github.com 2024-09-07T12:01:34+02:00 ggml : fix missing `cpu_set_t` on emscripten (#9336)
6c89eb0b4749184f4d19e96ada5dcb8847129b9c 9b2c24c0993487d3b34a873980e292da571481f3 slaren slarengh@gmail.com 2024-09-07T09:48:54+02:00 GitHub noreply@github.com 2024-09-07T10:48:54+03:00 ci : disable rocm image creation (#9340)
9b2c24c0993487d3b34a873980e292da571481f3 134bc38ecf3e2c5460581badce289a1ffa680453 Xuan Son Nguyen thichthat@gmail.com 2024-09-06T23:21:29+02:00 GitHub noreply@github.com 2024-09-06T23:21:29+02:00 server : simplify state machine for slot (#9283)
134bc38ecf3e2c5460581badce289a1ffa680453 815b1fb20a53e439882171757825bacb1350de04 Aarni Koskela akx@iki.fi 2024-09-07T00:03:01+03:00 GitHub noreply@github.com 2024-09-06T23:03:01+02:00 llama-bench : log benchmark progress (#9287)
815b1fb20a53e439882171757825bacb1350de04 409dc4f8bb5185786087f52259ee4626be93f54d Aarni Koskela akx@iki.fi 2024-09-06T18:59:58+03:00 GitHub noreply@github.com 2024-09-06T17:59:58+02:00 batched-bench : add `--output-format jsonl` option (#9293)
409dc4f8bb5185786087f52259ee4626be93f54d 4a1411b4f17b6569dc0a067e5914dcc0f738b370 Changyeon Kim cyzero.kim@samsung.com 2024-09-06T21:54:50+09:00 GitHub noreply@github.com 2024-09-06T15:54:50+03:00 ggml : fix build break for the vulkan-debug (#9265)
4a1411b4f17b6569dc0a067e5914dcc0f738b370 8ebe8ddebd68526757c631cd019de009697c63c2 Xuan Son Nguyen thichthat@gmail.com 2024-09-06T14:06:04+02:00 GitHub noreply@github.com 2024-09-06T14:06:04+02:00 server : fix missing lock (#9334)
8ebe8ddebd68526757c631cd019de009697c63c2 9bc6db28d011d47a5f318dc4aebbe7927fac4629 Markus Tavenrath mtavenrath@users.noreply.github.com 2024-09-06T08:56:17+02:00 GitHub noreply@github.com 2024-09-06T08:56:17+02:00 Improve Vulkan shader build system (#9239)
9bc6db28d011d47a5f318dc4aebbe7927fac4629 32b2ec88bc44b086f3807c739daf28a1613abde1 compilade git@compilade.net 2024-09-05T21:48:47-04:00 GitHub noreply@github.com 2024-09-05T21:48:47-04:00 ggml-quants : ternary packing for TriLMs and BitNet b1.58 (#8151)
32b2ec88bc44b086f3807c739daf28a1613abde1 1031771faaba28d07b4ec6a812cb21532efc8c31 awatuna 23447591+awatuna@users.noreply.github.com 2024-09-06T06:34:36+08:00 GitHub noreply@github.com 2024-09-06T00:34:36+02:00 Update build.yml (#9184)
1031771faaba28d07b4ec6a812cb21532efc8c31 4db04784f96757d74f74c8c110c2a00d55e33514 Michael Podvitskiy podvitskiymichael@gmail.com 2024-09-06T00:14:12+02:00 GitHub noreply@github.com 2024-09-06T00:14:12+02:00 CMake fix: host for msvc compiler can only be x86 or x64 (#8624)
4db04784f96757d74f74c8c110c2a00d55e33514 bdf314f38a2c90e18285f7d7067e8d736a14000a slaren slarengh@gmail.com 2024-09-05T11:13:11+02:00 GitHub noreply@github.com 2024-09-05T11:13:11+02:00 cuda : fix defrag with quantized KV (#9319)
bdf314f38a2c90e18285f7d7067e8d736a14000a 581c305186a0ff93f360346c57e21fe16e967bb7 slaren slarengh@gmail.com 2024-09-05T02:19:39+02:00 GitHub noreply@github.com 2024-09-05T02:19:39+02:00 llama-bench : fix NUL terminators in CPU name (#9313)
581c305186a0ff93f360346c57e21fe16e967bb7 5910ea942772ab6cbc21d0ad2d1208750ba39e1d Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-09-04T22:21:22+05:30 GitHub noreply@github.com 2024-09-04T19:51:22+03:00 ggml : AVX2 support for Q4_0_8_8 (#8713)
5910ea942772ab6cbc21d0ad2d1208750ba39e1d c8671ae28214b29920b86c66a5d36fd6dd0db870 Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-09-04T16:26:33+01:00 GitHub noreply@github.com 2024-09-04T16:26:33+01:00 [SYCL] Fix DMMV dequantization (#9279)
c8671ae28214b29920b86c66a5d36fd6dd0db870 82e3b03c11826d20a24ab66d60f4de58f48ddcdb 杨朱 · Kiki baofa.fan@daocloud.io 2024-09-04T19:45:28+08:00 GitHub noreply@github.com 2024-09-04T13:45:28+02:00 Fix broken links in docker.md (#9306)
82e3b03c11826d20a24ab66d60f4de58f48ddcdb 9379d3cc1718e9ec6ab5ffcb60a06bbe048a61ee Radoslav Gerganov rgerganov@gmail.com 2024-09-04T11:08:32+03:00 GitHub noreply@github.com 2024-09-04T11:08:32+03:00 rpc : make RPC servers come first in the device list (#9296)
9379d3cc1718e9ec6ab5ffcb60a06bbe048a61ee 7605ae7daf31c02211bcfec2f46635ef6ec4b98a Pascal Patry ppatry@mtacitlabs.com 2024-09-04T02:45:40-04:00 GitHub noreply@github.com 2024-09-04T09:45:40+03:00 readme : rename result_format to response_format (#9300)
7605ae7daf31c02211bcfec2f46635ef6ec4b98a 8962422b1c6f9b8b15f5aeaea42600bcc2d44177 Georgi Gerganov ggerganov@gmail.com 2024-09-04T02:36:43+03:00 GitHub noreply@github.com 2024-09-03T16:36:43-07:00 flake.lock: Update (#9261)
8962422b1c6f9b8b15f5aeaea42600bcc2d44177 b69a480af4db8ffd6cbb2efe7ed8132bc7d39073 Aarni Koskela akx@iki.fi 2024-09-03T20:58:54+03:00 GitHub noreply@github.com 2024-09-03T19:58:54+02:00 llama-bench : add JSONL (NDJSON) output mode (#9288)
b69a480af4db8ffd6cbb2efe7ed8132bc7d39073 48baa61eccdca9205daf8d620ba28055c2347b64 Georgi Gerganov ggerganov@gmail.com 2024-09-03T10:00:36+03:00 GitHub noreply@github.com 2024-09-03T10:00:36+03:00 readme : refactor API section + remove old hot topics
48baa61eccdca9205daf8d620ba28055c2347b64 f1485161e58d562099dd050f8ac3a9ea9f4cd765 Xuan Son Nguyen thichthat@gmail.com 2024-09-02T22:08:38+02:00 GitHub noreply@github.com 2024-09-02T22:08:38+02:00 server : test script : add timeout for all requests (#9282)
f1485161e58d562099dd050f8ac3a9ea9f4cd765 048de848ee4baad4531fcd6239438f5d55be365c Zhenwei Jin 109658203+kylo5aby@users.noreply.github.com 2024-09-03T01:53:23+08:00 GitHub noreply@github.com 2024-09-02T13:53:23-04:00 src: make tail invalid when kv cell is intersection for mamba (#9249)
048de848ee4baad4531fcd6239438f5d55be365c f771d064a95d212ddadea452ad0cc1e42b2c3db3 slaren slarengh@gmail.com 2024-09-02T18:11:13+02:00 GitHub noreply@github.com 2024-09-02T18:11:13+02:00 docker : fix missing binaries in full-cuda image (#9278)
f771d064a95d212ddadea452ad0cc1e42b2c3db3 6e7d133a5f9409dd257fad90d7f320721b07a1b2 yuri@FreeBSD yurivict@users.noreply.github.com 2024-09-02T08:25:30-07:00 GitHub noreply@github.com 2024-09-02T18:25:30+03:00 ggml : add pthread includes on FreeBSD (#9258)
6e7d133a5f9409dd257fad90d7f320721b07a1b2 b60074f1c26d8354053a952844ef3f7ebefd8803 Xuan Son Nguyen thichthat@gmail.com 2024-09-02T17:11:51+02:00 GitHub noreply@github.com 2024-09-02T17:11:51+02:00 server : refactor multitask handling (#9274)
b60074f1c26d8354053a952844ef3f7ebefd8803 9c1ba557335c3cab46807e6478eb7d17a63361f1 Guoliang Hua 32868157+nbcsm@users.noreply.github.com 2024-09-02T20:36:43+08:00 GitHub noreply@github.com 2024-09-02T15:36:43+03:00 llama-cli : remove duplicated log message (#9275)
9c1ba557335c3cab46807e6478eb7d17a63361f1 c6d4cb46559b359d2682cf2a002e7fe01bb7a767 Tushar ditsuke@protonmail.com 2024-09-02T16:51:01+05:30 GitHub noreply@github.com 2024-09-02T14:21:01+03:00 build(nix): Package gguf-py (#5664)
c6d4cb46559b359d2682cf2a002e7fe01bb7a767 8f1d81a0b6f50b9bad72db0b6fcd299ad9ecd48c Georgi Gerganov ggerganov@gmail.com 2024-09-02T11:52:04+03:00 Georgi Gerganov ggerganov@gmail.com 2024-09-02T11:52:37+03:00 llama : minor style
8f1d81a0b6f50b9bad72db0b6fcd299ad9ecd48c a47667cff41f5a198eb791974e0afcc1cddd3229 Molly Sophia mollysophia379@gmail.com 2024-09-01T22:38:17+08:00 GitHub noreply@github.com 2024-09-01T17:38:17+03:00 llama : support RWKV v6 models (#8980)
a47667cff41f5a198eb791974e0afcc1cddd3229 ea5d7478b1edcfc83f8ea8f9f0934585cc0b92e3 Echo Nolan echo@echonolan.net 2024-08-22T17:19:14-04:00 Someone else@someonex.net 2024-08-31T08:44:21Z nix: fix CUDA build - replace deprecated autoAddOpenGLRunpathHook
ea5d7478b1edcfc83f8ea8f9f0934585cc0b92e3 49271efbaf3f7a4ae52c4ca299b6d4e82598a97d Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-08-31T13:50:35+05:30 GitHub noreply@github.com 2024-08-31T11:20:35+03:00 sgemm : improved Q4_0 and Q8_0 performance via 4xN and Mx4 gemm (#8908)
49271efbaf3f7a4ae52c4ca299b6d4e82598a97d 0ab30f8d82fc7156b750c194d64a887e80cbfb82 Daniel Bevenius daniel.bevenius@gmail.com 2024-08-31T09:50:22+02:00 GitHub noreply@github.com 2024-08-31T10:50:22+03:00 llama : fix typo in xcda_array_view comment [no ci] (#9132)
0ab30f8d82fc7156b750c194d64a887e80cbfb82 cddae4884c853b1a7ab420458236d666e2e34423 Sutou Kouhei kou@cozmixng.org 2024-08-31T03:08:10+09:00 GitHub noreply@github.com 2024-08-30T20:08:10+02:00 llama : fix llama_split_mode enum values in main_gpu document (#9057)
cddae4884c853b1a7ab420458236d666e2e34423 7ea8d80d53364f2cdcd7bbc0f031106f529b1fe3 蕭澧邦 45505768+shou692199@users.noreply.github.com 2024-08-30T20:10:01+08:00 GitHub noreply@github.com 2024-08-30T22:10:01+10:00 Correct typo run_llama2.sh > run-llama2.sh (#9149)
7ea8d80d53364f2cdcd7bbc0f031106f529b1fe3 42c76d1358021ccdbe8ba89109c143dd7ae166df tc-mb 157115220+tc-mb@users.noreply.github.com 2024-08-30T13:21:57+08:00 GitHub noreply@github.com 2024-08-30T07:21:57+02:00 llava : the function "clip" should be int (#9237)
42c76d1358021ccdbe8ba89109c143dd7ae166df 9f7d4bcf5c27d37b0c7da82eeaf9c1499510554b Faisal Zaghloul quic_fzaghlou@quicinc.com 2024-08-29T19:20:53-04:00 GitHub noreply@github.com 2024-08-30T01:20:53+02:00 Threadpool: take 2 (#8672)
9f7d4bcf5c27d37b0c7da82eeaf9c1499510554b 1d1ccce67613674c75c9c7e3fa4c1e24e428ba48 Jan Boon jan.boon@kaetemi.be 2024-08-27T18:28:06+08:00 Jan Boon kaetemi@gmail.com 2024-08-30T07:15:26+08:00 server : fix crash when error handler dumps invalid utf-8 json (#9195)
1d1ccce67613674c75c9c7e3fa4c1e24e428ba48 9fe94ccac92693d4ae1bc283ff0574e8b3f4e765 Georgi Gerganov ggerganov@gmail.com 2024-08-29T07:28:14+03:00 GitHub noreply@github.com 2024-08-28T21:28:14-07:00 flake.lock: Update (#9162)
9fe94ccac92693d4ae1bc283ff0574e8b3f4e765 66b039a5011522b6a61495eea2a9862601e169f7 slaren slarengh@gmail.com 2024-08-28T17:28:00+02:00 GitHub noreply@github.com 2024-08-28T17:28:00+02:00 docker : build images only once (#9225)
66b039a5011522b6a61495eea2a9862601e169f7 20f1789dfb4e535d64ba2f523c64929e7891f428 slaren slarengh@gmail.com 2024-08-28T13:20:36+02:00 GitHub noreply@github.com 2024-08-28T13:20:36+02:00 docker : update CUDA images (#9213)
20f1789dfb4e535d64ba2f523c64929e7891f428 231cff5f6f1c050bcb448a8ac5857533b4c05dc7 Georgi Gerganov ggerganov@gmail.com 2024-08-27T22:10:58+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-27T22:41:27+03:00 vulkan : fix build (#0)
231cff5f6f1c050bcb448a8ac5857533b4c05dc7 3246fe84d78c8ccccd4291132809236ef477e9ea Georgi Gerganov ggerganov@gmail.com 2024-08-27T22:01:45+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-27T22:41:27+03:00 sync : ggml
3246fe84d78c8ccccd4291132809236ef477e9ea 78eb487bb0038eae95506d3d832b94c979185b09 Xie Yanbo xieyanbo@gmail.com 2024-08-27T20:33:08+08:00 GitHub noreply@github.com 2024-08-27T14:33:08+02:00 Fix minicpm example directory (#9111)
78eb487bb0038eae95506d3d832b94c979185b09 a77feb5d71831c61e455541e8a655b9f0337ea8c compilade git@compilade.net 2024-08-27T06:09:23-04:00 GitHub noreply@github.com 2024-08-27T13:09:23+03:00 llama : fix qs.n_attention_wv for DeepSeek-V2 (#9156)
a77feb5d71831c61e455541e8a655b9f0337ea8c 2e59d61c1b321431c597c1f12249273427d5640d Xuan Son Nguyen thichthat@gmail.com 2024-08-27T11:07:01+02:00 GitHub noreply@github.com 2024-08-27T11:07:01+02:00 server : add some missing env variables (#9116)
2e59d61c1b321431c597c1f12249273427d5640d 75e1dbbaaba5d762223370f3dab9db24a7f53465 CausalLM 148736309+CausalLM@users.noreply.github.com 2024-08-27T14:58:22+08:00 GitHub noreply@github.com 2024-08-27T09:58:22+03:00 llama : fix ChatGLM4 wrong shape (#9194)
75e1dbbaaba5d762223370f3dab9db24a7f53465 ad76569f8e78ab6ca921bda25cef25a157361719 Carsten Kragelund Jørgensen carsten@kragelund.me 2024-08-27T08:53:40+02:00 GitHub noreply@github.com 2024-08-27T09:53:40+03:00 llama : fix llama3.1 rope_freqs not respecting custom head_dim (#9141)
ad76569f8e78ab6ca921bda25cef25a157361719 7d787ed96c32be18603c158ab0276992cf0dc346 arch-btw 57669023+arch-btw@users.noreply.github.com 2024-08-26T22:58:50-07:00 GitHub noreply@github.com 2024-08-27T08:58:50+03:00 common : Update stb_image.h to latest version (#9161)
7d787ed96c32be18603c158ab0276992cf0dc346 06658ad7c37f440502de2b9486ce43c47b4ec710 slaren slarengh@gmail.com 2024-08-26T19:44:43+02:00 GitHub noreply@github.com 2024-08-26T19:44:43+02:00 ggml : do not crash when quantizing q4_x_x with an imatrix (#9192)
06658ad7c37f440502de2b9486ce43c47b4ec710 fc18425b6a8ad03847383ce2b69d52edfd49b0ff Georgi Gerganov ggerganov@gmail.com 2024-08-26T18:31:02+03:00 GitHub noreply@github.com 2024-08-26T18:31:02+03:00 metal : separate scale and mask from QKT in FA kernel (#9189)
fc18425b6a8ad03847383ce2b69d52edfd49b0ff 879275ac984235373dd44ed780d5e3a3883cb558 Georgi Gerganov ggerganov@gmail.com 2024-08-26T17:55:36+03:00 GitHub noreply@github.com 2024-08-26T17:55:36+03:00 ggml : add SSM Metal kernels (#8546)
879275ac984235373dd44ed780d5e3a3883cb558 7a3df798fc43e1b366146b1ad99137a9e95c87dd Georgi Gerganov ggerganov@gmail.com 2024-08-26T16:30:25+03:00 GitHub noreply@github.com 2024-08-26T16:30:25+03:00 tests : fix compile warnings for unreachable code (#9185)
7a3df798fc43e1b366146b1ad99137a9e95c87dd e5edb210cd361689da41f032f1b36c45ff1bf9be Georgi Gerganov ggerganov@gmail.com 2024-08-26T12:19:39+03:00 GitHub noreply@github.com 2024-08-26T12:19:39+03:00 ci : add VULKAN support to ggml-ci (#9055)
e5edb210cd361689da41f032f1b36c45ff1bf9be 0c41e03ceba1aafaf469476a56347419d5785376 Georgi Gerganov ggerganov@gmail.com 2024-08-26T12:16:57+03:00 GitHub noreply@github.com 2024-08-26T12:16:57+03:00 server : update deps (#9183)
0c41e03ceba1aafaf469476a56347419d5785376 f12ceaca0c7b59def30b7a832a6904df7ed3f4f7 slaren slarengh@gmail.com 2024-08-26T11:08:59+02:00 GitHub noreply@github.com 2024-08-26T11:08:59+02:00 metal : gemma2 flash attention support (#9159)
f12ceaca0c7b59def30b7a832a6904df7ed3f4f7 436787f170329b3f549e6c2c46593d2af8482e7c slaren slarengh@gmail.com 2024-08-26T11:03:30+02:00 GitHub noreply@github.com 2024-08-26T11:03:30+02:00 ggml-ci : try to improve build time (#9160)
436787f170329b3f549e6c2c46593d2af8482e7c 93bc3839f980ff14be86efe408b4cd7e89b26835 Justine Tunney jtunney@mozilla.com 2024-08-25T23:09:53-07:00 GitHub noreply@github.com 2024-08-26T09:09:53+03:00 llama : fix time complexity of string replacement (#9163)
93bc3839f980ff14be86efe408b4cd7e89b26835 f91fc5639be1e272194303ea26e1d4c226ec981b Herman Semenov GermanAizek@yandex.ru 2024-08-25T22:54:37Z GitHub noreply@github.com 2024-08-26T00:54:37+02:00 common: fixed not working find argument --n-gpu-layers-draft (#9175)
f91fc5639be1e272194303ea26e1d4c226ec981b e11bd856d538e44d24d8cad4b0381fba0984d162 Johannes Gäßler johannesg@5d6.de 2024-08-25T22:11:48+02:00 GitHub noreply@github.com 2024-08-25T22:11:48+02:00 CUDA: fix Gemma 2 numerical issues for FA (#9166)
e11bd856d538e44d24d8cad4b0381fba0984d162 8f824ffe8ee1feadd14428f1dda1283fa3b933be Johannes Gäßler johannesg@5d6.de 2024-08-24T21:34:59+02:00 GitHub noreply@github.com 2024-08-24T21:34:59+02:00 CPU/CUDA: Gemma 2 FlashAttention support (#8542)
8f824ffe8ee1feadd14428f1dda1283fa3b933be 3ba780e2a8f0ffe13f571b27f0bbf2ca5a199efc João Dinis Ferreira hello@joaof.eu 2024-08-24T07:22:45+01:00 GitHub noreply@github.com 2024-08-24T09:22:45+03:00 quantize : fix typo in usage help of `quantize.cpp` (#9145)
3ba780e2a8f0ffe13f571b27f0bbf2ca5a199efc a07c32ea54850c989f0ef6989da5b955b77b7172 Xuan Son Nguyen thichthat@gmail.com 2024-08-23T12:58:53+02:00 GitHub noreply@github.com 2024-08-23T12:58:53+02:00 lora : fix llama conversion script with ROPE_FREQS (#9117)
a07c32ea54850c989f0ef6989da5b955b77b7172 11b84eb4578864827afcf956db5b571003f18180 piDack 104877312+piDack@users.noreply.github.com 2024-08-23T15:27:17+08:00 GitHub noreply@github.com 2024-08-23T10:27:17+03:00 llama : use F32 precision in GLM4 attention and no FA (#9130)
11b84eb4578864827afcf956db5b571003f18180 1731d4238f9e4f925a750810e7f5480827c66dcf Akarshan Biswas akarshan.biswas@gmail.com 2024-08-22T19:39:47+05:30 GitHub noreply@github.com 2024-08-22T22:09:47+08:00 [SYCL] Add a space to supress a cmake warning (#9133)
1731d4238f9e4f925a750810e7f5480827c66dcf a1631e53f6763e17da522ba219b030d8932900bd luoyu-intel yu.luo@intel.com 2024-08-22T12:50:10+08:00 GitHub noreply@github.com 2024-08-22T12:50:10+08:00 [SYCL] Add oneDNN primitive support (#9091)
a1631e53f6763e17da522ba219b030d8932900bd fc54ef0d1c138133a01933296d50a36a1ab64735 compilade git@compilade.net 2024-08-21T17:58:11-04:00 GitHub noreply@github.com 2024-08-21T17:58:11-04:00 llama : simplify Mamba with advanced batch splits (#8526)
fc54ef0d1c138133a01933296d50a36a1ab64735 b40eb84895bf723c7b327a1e3bf6e0e2c41877f8 Xuan Son Nguyen thichthat@gmail.com 2024-08-21T11:04:34+02:00 GitHub noreply@github.com 2024-08-21T11:04:34+02:00 server : support reading arguments from environment variables (#9105)
b40eb84895bf723c7b327a1e3bf6e0e2c41877f8 f63f603c879c2232eaeded8c0aeba4244471d720 Younes Belkada 49240599+younesbelkada@users.noreply.github.com 2024-08-21T12:06:36+04:00 GitHub noreply@github.com 2024-08-21T11:06:36+03:00 llama : support for `falcon-mamba` architecture (#9074)
f63f603c879c2232eaeded8c0aeba4244471d720 8455340b874aa90d5f70104c99ed2778d9e31c36 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-08-21T09:45:49+02:00 GitHub noreply@github.com 2024-08-21T09:45:49+02:00 llava : zero-initialize clip_ctx structure fields with aggregate initialization 908)
8455340b874aa90d5f70104c99ed2778d9e31c36 2f3c1466ff46a2413b0e363a5005c46538186ee6 Daniel Bevenius daniel.bevenius@gmail.com 2024-08-21T09:32:58+02:00 GitHub noreply@github.com 2024-08-21T10:32:58+03:00 llama : std::move llm_bigram_bpe from work_queue (#9062)
2f3c1466ff46a2413b0e363a5005c46538186ee6 50addec9a532a6518146ab837a85504850627316 Changyeon Kim cyzero.kim@samsung.com 2024-08-21T04:00:00+09:00 GitHub noreply@github.com 2024-08-20T21:00:00+02:00 llava: Add ACC OP for GPU acceleration to the Vulkan backend in the LLAVA CLIP model. (#8984)
50addec9a532a6518146ab837a85504850627316 4f8d19ff17faa601f456ee1db7d8b8a15fa3f90b Meng, Hengyu hengyu.meng@intel.com 2024-08-20T23:50:17+08:00 GitHub noreply@github.com 2024-08-20T23:50:17+08:00 [SYCL] fallback mmvq (#9088)
4f8d19ff17faa601f456ee1db7d8b8a15fa3f90b 90db8146d56d83a605f2c475eca39bcda29cf16d zhentaoyu zhentao.yu@intel.com 2024-08-20T23:06:51+08:00 GitHub noreply@github.com 2024-08-20T23:06:51+08:00 [SYCL] Fix SYCL `im2col` and `convert` Overflow with Large Dims (#9052)
90db8146d56d83a605f2c475eca39bcda29cf16d cfac111e2b3953cdb6b0126e67a2487687646971 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-08-20T11:09:55+02:00 GitHub noreply@github.com 2024-08-20T12:09:55+03:00 tests : add missing comma in grammar integration tests (#9099)
cfac111e2b3953cdb6b0126e67a2487687646971 1b6ff90ff8301d9fe2027be2bb9fea26177d775e wangshuai09 391746016@qq.com 2024-08-19T16:46:38+08:00 GitHub noreply@github.com 2024-08-19T16:46:38+08:00 cann: add doc for cann backend (#8867)
1b6ff90ff8301d9fe2027be2bb9fea26177d775e 18eaf29f4c5f7c8e89412f3a9a0392a9e4e01d75 Radoslav Gerganov rgerganov@gmail.com 2024-08-19T10:11:45+03:00 GitHub noreply@github.com 2024-08-19T10:11:45+03:00 rpc : print error message when failed to connect endpoint (#9042)
18eaf29f4c5f7c8e89412f3a9a0392a9e4e01d75 554b049068de24201d19dde2fa83e35389d4585d Radoslav Gerganov rgerganov@gmail.com 2024-08-19T10:10:21+03:00 GitHub noreply@github.com 2024-08-19T10:10:21+03:00 rpc : prevent crashes on invalid input (#9040)
554b049068de24201d19dde2fa83e35389d4585d 2339a0be1c8e31fcf4531427183b94f2ef019e56 Georgi Gerganov ggerganov@gmail.com 2024-08-18T17:43:32+03:00 GitHub noreply@github.com 2024-08-18T07:43:32-07:00 flake.lock: Update (#9068)
2339a0be1c8e31fcf4531427183b94f2ef019e56 2fb9267887d24a431892ce4dccc75c7095b0d54d ltoniazzi 61414566+ltoniazzi@users.noreply.github.com 2024-08-18T10:58:04+01:00 GitHub noreply@github.com 2024-08-18T11:58:04+02:00 tests : add integration test for lora adapters (#8957)
2fb9267887d24a431892ce4dccc75c7095b0d54d 8b3befc0e2ed8fb18b903735831496b8b0c80949 Yoshi Suhara ysuhara@nvidia.com 2024-08-17T06:34:21-07:00 GitHub noreply@github.com 2024-08-17T15:34:21+02:00 Fix incorrect use of ctx_split for bias tensors (#9063)
8b3befc0e2ed8fb18b903735831496b8b0c80949 d565bb2fd5a2a58b9924a7a34e77a87c78c52137 Xuan Son Nguyen thichthat@gmail.com 2024-08-16T17:19:05+02:00 GitHub noreply@github.com 2024-08-16T17:19:05+02:00 server : refactor middleware and /health endpoint (#9056)
d565bb2fd5a2a58b9924a7a34e77a87c78c52137 ee2984bdaf10c14d440ad873a049bcc09b786d9b tc-mb 157115220+tc-mb@users.noreply.github.com 2024-08-16T21:34:41+08:00 GitHub noreply@github.com 2024-08-16T16:34:41+03:00 llava : support MiniCPM-V-2.6 (#8967)
ee2984bdaf10c14d440ad873a049bcc09b786d9b c8ddce85606d9fb6e30745b6e4fe103eecadc73f Farbod Bijary 110523279+farbodbj@users.noreply.github.com 2024-08-16T14:06:30+03:30 GitHub noreply@github.com 2024-08-16T13:36:30+03:00 py : fix wrong input type for raw_dtype in ggml to gguf scripts (#8928)
c8ddce85606d9fb6e30745b6e4fe103eecadc73f 23fd4535445e3e859ded2f02d6142b6e93b43890 Aisuko urakiny@gmail.com 2024-08-16T19:08:59+10:00 GitHub noreply@github.com 2024-08-16T11:08:59+02:00 Fix inference example lacks required parameters (#9035)
23fd4535445e3e859ded2f02d6142b6e93b43890 c679e0cb5c378bfb63643c44a453345ba8b90126 compilade git@compilade.net 2024-08-16T02:36:11-04:00 GitHub noreply@github.com 2024-08-16T09:36:11+03:00 gguf-py : bump version from 0.9.1 to 0.10.0 (#9051)
c679e0cb5c378bfb63643c44a453345ba8b90126 fb487bb5671b37267f35ff43fe8849ddccd925cd Minsoo Cheong icycle0409@snu.ac.kr 2024-08-16T15:35:18+09:00 GitHub noreply@github.com 2024-08-16T09:35:18+03:00 llama : add EXAONE model support (#9025)
fb487bb5671b37267f35ff43fe8849ddccd925cd 2a24c8caa6d10a7263ca317fa7cb64f0edc72aae Liu Jia jia3.liu@intel.com 2024-08-16T14:23:12+08:00 GitHub noreply@github.com 2024-08-16T09:23:12+03:00 common : add support for cpu_get_num_physical_cores() on Windows (#8771)
2a24c8caa6d10a7263ca317fa7cb64f0edc72aae e3f6fd56b1ab3c426596217d786910d641ae6ce0 Yoshi Suhara y.suhara@gmail.com 2024-08-15T19:23:33-07:00 GitHub noreply@github.com 2024-08-16T04:23:33+02:00 Add Nemotron/Minitron GGUF Conversion & Inference Support (#8922)
e3f6fd56b1ab3c426596217d786910d641ae6ce0 4b9afbbe9037f8a2d659097c0c7d9fce32c6494c Nico Bosshard nico@bosshome.ch 2024-08-16T04:22:55+02:00 GitHub noreply@github.com 2024-08-16T04:22:55+02:00 ggml : dynamic ggml_sched_max_splits based on graph_size (#9047)
4b9afbbe9037f8a2d659097c0c7d9fce32c6494c 37501d9c79ed5897db4b73ea7502211d25b3f763 gtygo gtydoit@gmail.com 2024-08-15T15:40:12+08:00 GitHub noreply@github.com 2024-08-15T10:40:12+03:00 retrieval : fix memory leak in retrieval query handling (#8955)
37501d9c79ed5897db4b73ea7502211d25b3f763 4af8420afba39de4968adf2695ce12fd42422a13 Riceball LEE snowyu.lee@gmail.com 2024-08-15T15:28:05+08:00 GitHub noreply@github.com 2024-08-15T10:28:05+03:00 server : fix duplicated n_predict key in the generation_settings (#8994)
4af8420afba39de4968adf2695ce12fd42422a13 6bda7ce6c3a9284fcbb70c1ace4107db8eb63e5c Zhenwei Jin 109658203+kylo5aby@users.noreply.github.com 2024-08-15T15:23:23+08:00 GitHub noreply@github.com 2024-08-15T10:23:23+03:00 common : remove duplicate function llama_should_add_bos_token (#8778)
6bda7ce6c3a9284fcbb70c1ace4107db8eb63e5c d5492f0525fa533817a67e93a4bde9d71d81cf58 Esko Toivonen eskot98@gmail.com 2024-08-15T10:17:12+03:00 GitHub noreply@github.com 2024-08-15T10:17:12+03:00 llama : add pre-tokenizer regexes for BLOOM and gpt3-finnish (#8850)
d5492f0525fa533817a67e93a4bde9d71d81cf58 234b30676a97ce227b604c38beb9dcaca406dea9 Georgi Gerganov ggerganov@gmail.com 2024-08-15T10:11:11+03:00 GitHub noreply@github.com 2024-08-15T10:11:11+03:00 ci : disable bench workflow (#9010)
234b30676a97ce227b604c38beb9dcaca406dea9 5fd89a70ead34d1a17015ddecad05aaa2490ca46 Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2024-08-15T08:21:57+02:00 GitHub noreply@github.com 2024-08-15T09:21:57+03:00 server : init stop and error fields of the result struct (#9026)
5fd89a70ead34d1a17015ddecad05aaa2490ca46 98a532d474c73d3494a5353024cb6a4fbbabbb35 0cc4m picard12@live.de 2024-08-14T18:32:53+02:00 GitHub noreply@github.com 2024-08-14T18:32:53+02:00 Vulkan Optimizations and Fixes (#8959)
98a532d474c73d3494a5353024cb6a4fbbabbb35 43bdd3ce188cd247bda7c1bd1ad01fa64e566690 compilade git@compilade.net 2024-08-14T02:51:02-04:00 GitHub noreply@github.com 2024-08-14T09:51:02+03:00 server : fix segfault on long system prompt (#8987)
43bdd3ce188cd247bda7c1bd1ad01fa64e566690 06943a69f678fb32829ff06d9c18367b17d4b361 Georgi Gerganov ggerganov@gmail.com 2024-08-14T09:14:49+03:00 GitHub noreply@github.com 2024-08-14T09:14:49+03:00 cmake : remove unused option GGML_CURL (#9011)
06943a69f678fb32829ff06d9c18367b17d4b361 828d6ff7d796f48b2c345f6be2805a3c531a089c Daniel Bevenius daniel.bevenius@gmail.com 2024-08-13T21:13:15+02:00 GitHub noreply@github.com 2024-08-13T21:13:15+02:00 ggml : move rope type enum to ggml.h (#8949)
828d6ff7d796f48b2c345f6be2805a3c531a089c fc4ca27b25464a11b3b86c9dbb5b6ed6065965c2 Xuan Son Nguyen thichthat@gmail.com 2024-08-13T11:41:14+02:00 GitHub noreply@github.com 2024-08-13T11:41:14+02:00 export-lora : throw error if lora is quantized (#9002)
fc4ca27b25464a11b3b86c9dbb5b6ed6065965c2 1f67436c5ee6f4c99e71a8518bdfc214c27ce934 Diogo Teles Sant'Anna diogoteles@google.com 2024-08-12T13:28:23-03:00 GitHub noreply@github.com 2024-08-12T19:28:23+03:00 ci : fix github workflow vulnerable to script injection (#9008)
1f67436c5ee6f4c99e71a8518bdfc214c27ce934 0fd93cdef5e583aa980b3c0d693c0d207f0787a7 Radoslav Gerganov rgerganov@gmail.com 2024-08-12T19:17:03+03:00 GitHub noreply@github.com 2024-08-12T19:17:03+03:00 ci : enable RPC in all of the released builds (#9006)
0fd93cdef5e583aa980b3c0d693c0d207f0787a7 84eb2f4fad28ceadd415a4e775320c983f4d9a7d Nico Bosshard nico@bosshome.ch 2024-08-12T17:13:59+02:00 GitHub noreply@github.com 2024-08-12T17:13:59+02:00 llama : model-based max number of graph nodes calculation (#8970)
84eb2f4fad28ceadd415a4e775320c983f4d9a7d 1262e7ed13ac197c944f15e1ddb083cb4f36cf65 Frank Mai thxcode0824@gmail.com 2024-08-12T20:45:50+08:00 GitHub noreply@github.com 2024-08-12T14:45:50+02:00 docs: introduce gpustack and gguf-parser (#8873)
1262e7ed13ac197c944f15e1ddb083cb4f36cf65 df5478fbea7e652cfad4ee7974ac3b624fd6c7f6 DavidKorczynski david@adalogics.com 2024-08-12T13:36:41+01:00 GitHub noreply@github.com 2024-08-12T15:36:41+03:00 grammar-parser : fix possible null-deref (#9004)
df5478fbea7e652cfad4ee7974ac3b624fd6c7f6 2589292cde038ba876c041bcd7b3f0c81f3f11fe DavidKorczynski david@adalogics.com 2024-08-12T13:21:41+01:00 GitHub noreply@github.com 2024-08-12T14:21:41+02:00 ggml: fix div-by-zero (#9003)
2589292cde038ba876c041bcd7b3f0c81f3f11fe d3ae0ee8d75033921a076131d4d0fa1c6ec579a7 Liu Jia jia3.liu@intel.com 2024-08-12T17:46:03+08:00 GitHub noreply@github.com 2024-08-12T11:46:03+02:00 Fix a spelling mistake (#9001)
d3ae0ee8d75033921a076131d4d0fa1c6ec579a7 5ef07e25ac39e62297a67208c5bcced50835a2dd Georgi Gerganov ggerganov@gmail.com 2024-08-12T11:02:01+03:00 GitHub noreply@github.com 2024-08-12T11:02:01+03:00 py : fix requirements check '==' -> '~=' (#8982)
5ef07e25ac39e62297a67208c5bcced50835a2dd 4134999e01f31256b15342b41c4de9e2477c4a6c Georgi Gerganov ggerganov@gmail.com 2024-08-12T10:21:50+03:00 GitHub noreply@github.com 2024-08-12T10:21:50+03:00 server : handle models with missing EOS token (#8997)
4134999e01f31256b15342b41c4de9e2477c4a6c 8cd1bcfd3fc9f2b5cbafd7fb7581b3278acec25f compilade git@compilade.net 2024-08-11T14:45:41-04:00 GitHub noreply@github.com 2024-08-11T14:45:41-04:00 gguf-py : Numpy dequantization for most types (#8939)
8cd1bcfd3fc9f2b5cbafd7fb7581b3278acec25f a21c6fd45032a20180e026773582d21294c85619 Georgi Gerganov ggerganov@gmail.com 2024-08-11T16:58:58+03:00 GitHub noreply@github.com 2024-08-11T06:58:58-07:00 flake.lock: Update (#8979)
a21c6fd45032a20180e026773582d21294c85619 33309f661a93c9c0ab65a79e5e7e30fa6162992e Neo Zhang zhang.jianyu@outlook.com 2024-08-11T16:37:43+08:00 GitHub noreply@github.com 2024-08-11T14:07:43+05:30 update guide (#8909)
33309f661a93c9c0ab65a79e5e7e30fa6162992e 7c5bfd57f83fd3630934cfa70892aa4022d3faf7 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-08-11T10:35:26+02:00 GitHub noreply@github.com 2024-08-11T10:35:26+02:00 llama : check all graph nodes when searching for result_embd_pooled (#8956)
7c5bfd57f83fd3630934cfa70892aa4022d3faf7 6e02327e8b7837358e0406bf90a4632e18e27846 Markus Tavenrath mtavenrath@users.noreply.github.com 2024-08-11T10:09:09+02:00 GitHub noreply@github.com 2024-08-11T10:09:09+02:00 Optimize Vulkan backend for better CPU performance and less GPU synchronization overhead. (#8943)
6e02327e8b7837358e0406bf90a4632e18e27846 7eb23840ed0f388e10c4bbc4d65802fdfb977b40 slaren slarengh@gmail.com 2024-08-10T15:42:10+02:00 GitHub noreply@github.com 2024-08-10T15:42:10+02:00 metal : fix uninitialized abort_callback (#8968)
7eb23840ed0f388e10c4bbc4d65802fdfb977b40 7c3f55c10051c634546247387c5c359c9d499360 Xuan Son Nguyen thichthat@gmail.com 2024-08-10T13:04:40+02:00 GitHub noreply@github.com 2024-08-10T13:04:40+02:00 llama : default n_swa for phi-3 (#8931)
7c3f55c10051c634546247387c5c359c9d499360 911b437f228e75aa3d235acec21bfddd23ecce2f fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-08-10T11:43:26+02:00 GitHub noreply@github.com 2024-08-10T11:43:26+02:00 Add support for encoder-only T5 models (#8900)
911b437f228e75aa3d235acec21bfddd23ecce2f b72942fac998672a79a1ae3c03b340f7e629980b Matteo Mortari matteo.mortari@gmail.com 2024-08-10T07:58:49+02:00 GitHub noreply@github.com 2024-08-10T08:58:49+03:00 gguf-py : fix double call to add_architecture() (#8952)
b72942fac998672a79a1ae3c03b340f7e629980b 6afd1a99dc9792096d4567ab9fa1ad530c81c6cd Georgi Gerganov ggerganov@gmail.com 2024-08-09T23:03:21+03:00 GitHub noreply@github.com 2024-08-09T23:03:21+03:00 Merge commit from fork
6afd1a99dc9792096d4567ab9fa1ad530c81c6cd 272e3bd95e620d285b2fb9faaa7b6b1b8edbbf3a fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-08-09T18:53:09+02:00 GitHub noreply@github.com 2024-08-09T18:53:09+02:00 llama : add support for lora adapters in T5 model (#8938)
272e3bd95e620d285b2fb9faaa7b6b1b8edbbf3a 45a55b91aa87958a75d691be64b070266d4fbb94 Georgi Gerganov ggerganov@gmail.com 2024-08-09T18:24:30+03:00 GitHub noreply@github.com 2024-08-09T18:24:30+03:00 make : fix llava obj file race (#8946)
45a55b91aa87958a75d691be64b070266d4fbb94 3071c0a5f218f107dabd13b73f6090af683ef5ec Georgi Gerganov ggerganov@gmail.com 2024-08-09T18:23:52+03:00 GitHub noreply@github.com 2024-08-09T18:23:52+03:00 llama : better replace_all (cont) (#8926)
3071c0a5f218f107dabd13b73f6090af683ef5ec 4305b57c80eff4f0df5f6acb60b292f03b8f0dd0 tc-mb 157115220+tc-mb@users.noreply.github.com 2024-08-09T18:33:53+08:00 GitHub noreply@github.com 2024-08-09T13:33:53+03:00 llava : support MiniCPM-V-2.5 (#7599)
4305b57c80eff4f0df5f6acb60b292f03b8f0dd0 70c0ea35609a2ab87a358e25f4ffad1aad408992 Georgi Gerganov ggerganov@gmail.com 2024-08-09T10:03:48+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-09T10:03:48+03:00 sync : ggml
70c0ea35609a2ab87a358e25f4ffad1aad408992 5b2c04f4925e152c362b824f4b41eb2a081fa623 Matt Stephenson mstephenson6@users.noreply.github.com 2024-07-16T03:21:09-04:00 Georgi Gerganov ggerganov@gmail.com 2024-08-09T10:03:44+03:00 whisper : use vulkan as gpu backend when available (whisper/2302)
5b2c04f4925e152c362b824f4b41eb2a081fa623 6f6496bb0999d1bce5daff0cfc55ceb0dd13c888 Daniel Bevenius daniel.bevenius@gmail.com 2024-08-09T08:33:30+02:00 GitHub noreply@github.com 2024-08-09T09:33:30+03:00 embedding : add --pooling option to README.md [no ci] (#8934)
6f6496bb0999d1bce5daff0cfc55ceb0dd13c888 daef3ab233fc02e4c53afd9b07366379876f00d1 Daniel Bevenius daniel.bevenius@gmail.com 2024-08-09T08:32:23+02:00 GitHub noreply@github.com 2024-08-09T09:32:23+03:00 llama : fix typo in llama_tensor_get_type comment [no ci] (#8937)
daef3ab233fc02e4c53afd9b07366379876f00d1 345a686d8271a24db20d31789c0d4b9ed51dcb0c Mathieu Geli mathieu.geli@gmail.com 2024-08-09T08:32:02+02:00 GitHub noreply@github.com 2024-08-09T09:32:02+03:00 server : add one level list nesting for embeddings (#8936)
345a686d8271a24db20d31789c0d4b9ed51dcb0c 3a14e00366399040a139c67dd5951177a8cb5695 compilade git@compilade.net 2024-08-08T23:54:00-04:00 GitHub noreply@github.com 2024-08-08T23:54:00-04:00 llama : reduce useless copies when saving session (#8916)
3a14e00366399040a139c67dd5951177a8cb5695 afd27f01fe832ece3d07ef03b7d34a9e80c4a895 compilade git@compilade.net 2024-08-08T13:33:09-04:00 GitHub noreply@github.com 2024-08-08T13:33:09-04:00 gguf-py : simplify support for quant types (#8838)
afd27f01fe832ece3d07ef03b7d34a9e80c4a895 366d486c163ea883442313cff1a3b154ab93e168 Georgi Gerganov ggerganov@gmail.com 2024-08-08T14:56:52+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-08T14:56:52+03:00 scripts : sync cann files (#0)
366d486c163ea883442313cff1a3b154ab93e168 e44a561ab090b11d3a6fe539b768404663e4c3d2 Georgi Gerganov ggerganov@gmail.com 2024-08-08T14:40:12+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-08T14:40:12+03:00 scripts : fix sync filenames (#0)
e44a561ab090b11d3a6fe539b768404663e4c3d2 f93d49ab1e2d1d698af8f17ead6d635405d4b289 Georgi Gerganov ggerganov@gmail.com 2024-08-08T13:19:47+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-08T13:19:47+03:00 sync : ggml
f93d49ab1e2d1d698af8f17ead6d635405d4b289 5b33ea1ee72fadfa4f96d86dc614631739758cce Borislav Stanimirov b.stanimirov@abv.bg 2024-08-07T10:00:56+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-08T13:19:31+03:00 ggml : ignore more msvc warnings (ggml/906)
5b33ea1ee72fadfa4f96d86dc614631739758cce 85fca8deb6273b956bc9184bc9d70a07e1d50d07 Georgi Gerganov ggerganov@gmail.com 2024-08-07T09:57:00+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-08T13:19:31+03:00 metal : fix struct name (ggml/912)
85fca8deb6273b956bc9184bc9d70a07e1d50d07 ebd541a5705b6f7a4ce67824d1c2d4fc790f1770 Conrad Kramer conrad@conradkramer.com 2024-08-07T02:55:49-04:00 Georgi Gerganov ggerganov@gmail.com 2024-08-08T13:19:30+03:00 metal : add abort callback (ggml/905)
ebd541a5705b6f7a4ce67824d1c2d4fc790f1770 15fa07a5c564d3ed7e7eb64b73272cedb27e73ec Pablo Duboue pablo.duboue@gmail.com 2024-08-08T04:44:51-04:00 GitHub noreply@github.com 2024-08-08T11:44:51+03:00 make : clean llamafile objects (#8923)
15fa07a5c564d3ed7e7eb64b73272cedb27e73ec be55695eff44784a141a863f273661a6bce63dfc slaren slarengh@gmail.com 2024-08-07T18:24:05+02:00 GitHub noreply@github.com 2024-08-07T18:24:05+02:00 make : use C compiler to build metal embed object (#8899)
be55695eff44784a141a863f273661a6bce63dfc 0478174d5959b66096ae6609fcb0df14cab66b51 slaren slarengh@gmail.com 2024-08-07T13:29:02+02:00 GitHub noreply@github.com 2024-08-07T13:29:02+02:00 ggml-backend : fix async copy from CPU (#8897)
0478174d5959b66096ae6609fcb0df14cab66b51 a8dbc6f753f296108121d50f78f67463403dd6fc Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-08-07T11:25:36+01:00 GitHub noreply@github.com 2024-08-07T11:25:36+01:00 [SYCL] Updated SYCL device filtering (#8901)
a8dbc6f753f296108121d50f78f67463403dd6fc 506122d854c5d05b4a3d45a294f14bd4c02d9868 Johannes Gäßler johannesg@5d6.de 2024-08-07T09:07:52+02:00 GitHub noreply@github.com 2024-08-07T09:07:52+02:00 CUDA/HIP: fix tests/test-backend-ops (#8896)
506122d854c5d05b4a3d45a294f14bd4c02d9868 725e3d94379d5b619c027347308bccf2e0ead89f Zhenwei Jin 109658203+kylo5aby@users.noreply.github.com 2024-08-07T09:01:06+08:00 GitHub noreply@github.com 2024-08-07T03:01:06+02:00 llama-bench : add support for getting cpu info on Windows (#8824)
725e3d94379d5b619c027347308bccf2e0ead89f 31958546c3e4695a8a24bb7ba3b79a8f76d05afe Daniel Bevenius daniel.bevenius@gmail.com 2024-08-07T01:43:00+02:00 GitHub noreply@github.com 2024-08-07T01:43:00+02:00 quantize : update usage comment in quantize.cpp (#8889)
31958546c3e4695a8a24bb7ba3b79a8f76d05afe 1e6f6554aa11fa10160a5fda689e736c3c34169f Nexes the Old 124105151+Nexesenex@users.noreply.github.com 2024-08-07T01:41:54+02:00 GitHub noreply@github.com 2024-08-07T01:41:54+02:00 typo correction (#8891)
1e6f6554aa11fa10160a5fda689e736c3c34169f 641f5dd2a6422941faa9f32ab5cb50fc1b24c1f5 Xuan Son Nguyen thichthat@gmail.com 2024-08-06T17:33:39+02:00 GitHub noreply@github.com 2024-08-06T17:33:39+02:00 server : add lora hotswap endpoint (WIP) (#8857)
641f5dd2a6422941faa9f32ab5cb50fc1b24c1f5 5f4dcb1e60bbfe936b45778dad177a5b9a09b066 Johannes Gäßler johannesg@5d6.de 2024-08-06T17:13:55+02:00 GitHub noreply@github.com 2024-08-06T17:13:55+02:00 CUDA: fix padding logic for FP16/FP32 (#8884)
5f4dcb1e60bbfe936b45778dad177a5b9a09b066 db20f50cf4710c46e3a996919a26858cae8c80ed Daniel Bevenius daniel.bevenius@gmail.com 2024-08-06T16:44:35+02:00 GitHub noreply@github.com 2024-08-06T16:44:35+02:00 simple : update name of executable to llama-simple (#8885)
db20f50cf4710c46e3a996919a26858cae8c80ed efda90c93a62274ec0b0bfa80c4eee4bdb6966d0 Jaeden Amero jaeden@patater.com 2024-08-06T17:21:47+04:00 GitHub noreply@github.com 2024-08-06T15:21:47+02:00 cmake : Link vulkan-shaders-gen with pthreads (#8835)
efda90c93a62274ec0b0bfa80c4eee4bdb6966d0 0bf16de07b0692e7df26b9a633e232bbd66e0360 MaggotHATE clay1326@gmail.com 2024-08-06T16:32:03+05:00 GitHub noreply@github.com 2024-08-06T13:32:03+02:00 [Vulkan] Fix compilation of `vulkan-shaders-gen` on w64devkit after `e31a4f6` (#8880)
0bf16de07b0692e7df26b9a633e232bbd66e0360 2d5dd7bb3fa382806cd3e0bfc7a1d92349bc0ccf Georgi Gerganov ggerganov@gmail.com 2024-08-06T11:48:01+03:00 GitHub noreply@github.com 2024-08-06T11:48:01+03:00 contributing : add note about write access
2d5dd7bb3fa382806cd3e0bfc7a1d92349bc0ccf cdd1889de62a7140c8c016405ec917464bde8bd3 Molly Sophia mollysophia379@gmail.com 2024-08-06T15:26:46+08:00 GitHub noreply@github.com 2024-08-06T10:26:46+03:00 ggml : add epsilon as a parameter for group_norm (#8818)
cdd1889de62a7140c8c016405ec917464bde8bd3 c21a896405de4cdf4207eb8130555ceaac0ab110 Douglas Hanley thesecretaryofwar@gmail.com 2024-08-06T02:20:54-05:00 GitHub noreply@github.com 2024-08-06T10:20:54+03:00 convert : add support for XLMRoberta embedding models (#8658)
c21a896405de4cdf4207eb8130555ceaac0ab110 d4ff847153e9cf7220d1b39aa21172069e6e8cea Mengqing Cao cmq0113@163.com 2024-08-06T12:42:42+08:00 GitHub noreply@github.com 2024-08-06T12:42:42+08:00 [CANN]: Fix ggml_backend_cann_buffer_get_tensor (#8871)
d4ff847153e9cf7220d1b39aa21172069e6e8cea 0a4ce786814b123096d18aadca89cd352b9e590b Neo Zhang zhang.jianyu@outlook.com 2024-08-06T09:09:12+08:00 GitHub noreply@github.com 2024-08-06T09:09:12+08:00 [SYCL] correct cmd name (#8877)
0a4ce786814b123096d18aadca89cd352b9e590b bc0f887e159c0d78c28121e2c8b5c58094170875 Liu Jia 109258120+Septa2112@users.noreply.github.com 2024-08-06T00:14:10+08:00 GitHub noreply@github.com 2024-08-05T18:14:10+02:00 common : Changed tuple to struct (TODO fix) (#8823)
bc0f887e159c0d78c28121e2c8b5c58094170875 b42978e7e4d56eaaa93588414e804d9fbbc3cae2 wangshuai09 391746016@qq.com 2024-08-05T21:10:37+08:00 GitHub noreply@github.com 2024-08-05T21:10:37+08:00 cann: fix buffer_num and runtime speed slowly error (#8865)
b42978e7e4d56eaaa93588414e804d9fbbc3cae2 b9dfc25ca385a83bde9e9456c4d4fae15377bc7b Eric Curtin ericcurtin17@gmail.com 2024-08-05T13:45:01+01:00 GitHub noreply@github.com 2024-08-05T15:45:01+03:00 readme : add ramalama to the availables UI (#8811)
b9dfc25ca385a83bde9e9456c4d4fae15377bc7b 1ef14b30075da594cb24f0ab858a14bf1d8d1797 Justine Tunney jtunney@mozilla.com 2024-08-05T05:43:40-07:00 GitHub noreply@github.com 2024-08-05T15:43:40+03:00 ggml : fix overflows in elu function (#8866)
1ef14b30075da594cb24f0ab858a14bf1d8d1797 d3f0c7166adfa952237e0f437a5344362d8256d4 Brian mofosyne@gmail.com 2024-08-05T21:15:28+10:00 GitHub noreply@github.com 2024-08-05T21:15:28+10:00 py: Add more authorship metadata from model card (#8810)
d3f0c7166adfa952237e0f437a5344362d8256d4 e31a4f679779220312c165b0f5994c680a610e38 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-08-05T09:38:01+02:00 GitHub noreply@github.com 2024-08-05T09:38:01+02:00 Stop the generation when <|eom_id|> token is encountered - needed for Llama 3.1 tool call support (#8858)
e31a4f679779220312c165b0f5994c680a610e38 400ae6f65f0b55babd48d1e3ec7fd663a97fc8d0 stduhpf stephduh@live.fr 2024-08-05T08:18:27+02:00 GitHub noreply@github.com 2024-08-05T08:18:27+02:00 cmake: fix paths for vulkan shaders compilation on Windows (#8573)
400ae6f65f0b55babd48d1e3ec7fd663a97fc8d0 f1ea5146d741a0c9be6d8fbfab9323fea6c4a3f0 BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2024-08-05T01:54:10-04:00 GitHub noreply@github.com 2024-08-05T08:54:10+03:00 readme : update model list (#8851)
f1ea5146d741a0c9be6d8fbfab9323fea6c4a3f0 064cdc265fb63590c7c8f04a609d36ef200d55a7 Georgi Gerganov ggerganov@gmail.com 2024-08-05T08:53:39+03:00 GitHub noreply@github.com 2024-08-05T08:53:39+03:00 llama : better replace_all (#8852)
064cdc265fb63590c7c8f04a609d36ef200d55a7 5587e57a76630651752031223cc7024cb32cf308 0cc4m picard12@live.de 2024-08-05T07:52:55+02:00 GitHub noreply@github.com 2024-08-05T08:52:55+03:00 vulkan : fix Qantized Mat-Vec Mul on AMD GPUs for ncols < 64 (#8855)
5587e57a76630651752031223cc7024cb32cf308 a3738b2fa7c60ef2c4592435d1aa7fb8f1f69c3e Georgi Gerganov ggerganov@gmail.com 2024-08-04T19:13:25+03:00 Georgi Gerganov ggerganov@gmail.com 2024-08-05T08:50:57+03:00 sync : ggml
a3738b2fa7c60ef2c4592435d1aa7fb8f1f69c3e 655858ace0cf2720e56eb01f84ad05e0c94ada3c 0cc4m picard12@live.de 2024-08-04T17:28:08+02:00 Georgi Gerganov ggerganov@gmail.com 2024-08-05T08:50:57+03:00 vulkan : implement Stable Diffusion operators (ggml/904)
655858ace0cf2720e56eb01f84ad05e0c94ada3c c02b0a8a4dee489b29073f25a27ed6e5628e86e1 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-29T15:06:06+02:00 Georgi Gerganov ggerganov@gmail.com 2024-08-05T08:50:57+03:00 ggml : move c parameter comment to ggml_rope_ext (ggml/901)
c02b0a8a4dee489b29073f25a27ed6e5628e86e1 0d6fb52be0c1b7e77eb855f3adc4952771c8ce4c wangshuai09 391746016@qq.com 2024-08-05T12:22:30+08:00 GitHub noreply@github.com 2024-08-05T12:22:30+08:00 cann: support q4_0 model (#8822)
0d6fb52be0c1b7e77eb855f3adc4952771c8ce4c 978ba3d83d17b10fdf9807006048432b5b3769fc Brandon Squizzato 35474886+bsquizz@users.noreply.github.com 2024-08-04T14:17:16-04:00 GitHub noreply@github.com 2024-08-04T20:17:16+02:00 Install curl in runtime layer (#8693)
978ba3d83d17b10fdf9807006048432b5b3769fc ecf6b7f23e664afd7ff856ec39034240ce438daa ardfork 134447697+ardfork@users.noreply.github.com 2024-08-04T18:16:23Z GitHub noreply@github.com 2024-08-04T20:16:23+02:00 Server: Don't ignore llama.cpp params (#8754)
ecf6b7f23e664afd7ff856ec39034240ce438daa 01aae2b4975b57a265ce8194928fd87f2d71027e Brian Cunnie brian.cunnie@gmail.com 2024-08-04T03:55:03-07:00 GitHub noreply@github.com 2024-08-04T13:55:03+03:00 batched-bench : handle empty `-npl` (#8839)
01aae2b4975b57a265ce8194928fd87f2d71027e 4b77ea95f56a4c49bc995f08eac62a6416875ccc Daniel Bevenius daniel.bevenius@gmail.com 2024-08-03T15:07:47+02:00 Georgi Gerganov ggerganov@gmail.com 2024-08-04T13:24:59+03:00 baby-llama : remove duplicate vector include
4b77ea95f56a4c49bc995f08eac62a6416875ccc 76614f352e94d25659306d9e97321f204e5de0d3 Georgi Gerganov ggerganov@gmail.com 2024-08-04T05:53:20+03:00 GitHub noreply@github.com 2024-08-03T19:53:20-07:00 flake.lock: Update (#8847)
76614f352e94d25659306d9e97321f204e5de0d3 b72c20b85c1029d135022d39e9a20d4807c11893 jdomke 28772296+jdomke@users.noreply.github.com 2024-08-04T01:34:41+09:00 GitHub noreply@github.com 2024-08-03T18:34:41+02:00 ggml : reading the runtime sve config of the cpu (#8709)
b72c20b85c1029d135022d39e9a20d4807c11893 e09a800f9a9b19c73aa78e03b4c4be8ed988f3e6 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-08-02T21:11:39+02:00 GitHub noreply@github.com 2024-08-02T15:11:39-04:00 Fix conversion of unnormalized BF16->BF16 weights (#7843)
e09a800f9a9b19c73aa78e03b4c4be8ed988f3e6 0fbbd884589d585c3b43cae8c16938ffffb863b9 Mengqing Cao cmq0113@163.com 2024-08-02T16:50:53+08:00 GitHub noreply@github.com 2024-08-02T16:50:53+08:00 cann: Fix ggml_cann_im2col for 1D im2col (#8819)
0fbbd884589d585c3b43cae8c16938ffffb863b9 afbb4c1322a747d2a7b4bf67c868148f8afcc6c8 Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-08-02T01:55:17+01:00 GitHub noreply@github.com 2024-08-02T08:55:17+08:00 [SYCL] Fixing wrong VDR iq4nl value (#8812)
afbb4c1322a747d2a7b4bf67c868148f8afcc6c8 b7a08fd5e0e7c898c68d1743066ea495202d9608 matteo matteogeniaccio@yahoo.it 2024-08-01T23:28:28+02:00 GitHub noreply@github.com 2024-08-01T23:28:28+02:00 ggml-cuda: Adding support for unified memory (#8035)
b7a08fd5e0e7c898c68d1743066ea495202d9608 7a11eb3a260915aee16101808f291a244e2facc7 Alex O'Connell 35843486+acon96@users.noreply.github.com 2024-08-01T12:53:46-04:00 GitHub noreply@github.com 2024-08-01T18:53:46+02:00 Build: Only include execinfo.h on linux systems that support it (#8783)
7a11eb3a260915aee16101808f291a244e2facc7 c8a0090922bad576623de4aae227717085249262 slaren slarengh@gmail.com 2024-08-01T15:26:22+02:00 GitHub noreply@github.com 2024-08-01T15:26:22+02:00 cuda : fix dmmv cols requirement to 2*GGML_CUDA_DMMV_X (#8800)
c8a0090922bad576623de4aae227717085249262 afbbcf3c04e3c6420cad3d72571478cd62ac176c wangshuai09 391746016@qq.com 2024-08-01T10:39:05+08:00 GitHub noreply@github.com 2024-08-01T10:39:05+08:00 cann: support q8_0 for Ascend backend (#8805)
afbbcf3c04e3c6420cad3d72571478cd62ac176c ed9d2854c9de4ae1f448334294e61167b04bec2a Igor Okulist okigan@gmail.com 2024-07-31T18:59:09-05:00 GitHub noreply@github.com 2024-07-31T19:59:09-04:00 server : update llama-server embedding flag documentation (#8779)
ed9d2854c9de4ae1f448334294e61167b04bec2a 398ede5efeb07b9adf9fbda7ea63f630d476a792 Clint Herron hanclinto@gmail.com 2024-07-31T15:51:06-04:00 GitHub noreply@github.com 2024-07-31T15:51:06-04:00 Build: Fix potential race condition (#8781)
398ede5efeb07b9adf9fbda7ea63f630d476a792 44d28ddd5caaa5e9de573bdaaa5b5b2448a29ace pculliton phillipculliton@gmail.com 2024-07-31T11:12:10-04:00 GitHub noreply@github.com 2024-07-31T17:12:10+02:00 Adding Gemma 2 2B configs (#8784)
44d28ddd5caaa5e9de573bdaaa5b5b2448a29ace 268c5660062270a2c19a36fc655168aa287aaec2 Borislav Stanimirov b.stanimirov@abv.bg 2024-07-31T16:40:08+03:00 GitHub noreply@github.com 2024-07-31T15:40:08+02:00 cmake : fix use of external ggml (#8787)
268c5660062270a2c19a36fc655168aa287aaec2 7e72aa74fd676a093eb9970e761085ec22734c71 Someone sergei.kozlukov@aalto.fi 2024-07-30T23:35:30+03:00 GitHub noreply@github.com 2024-07-30T13:35:30-07:00 nix: cuda: rely on propagatedBuildInputs (#8772)
7e72aa74fd676a093eb9970e761085ec22734c71 7c27a19b2eb91bb0f43c7f7aec0386cec2dddc33 Brian mofosyne@gmail.com 2024-07-31T00:57:03+10:00 GitHub noreply@github.com 2024-07-31T00:57:03+10:00 py: add_array() will not add to kv store if value is an empty array (#8774)
7c27a19b2eb91bb0f43c7f7aec0386cec2dddc33 140074bb8647df41840d6f32f4409fa8959bcf9f l3utterfly gc.pthzfoldr@gmail.com 2024-07-30T23:40:18+09:00 GitHub noreply@github.com 2024-07-30T16:40:18+02:00 added android implementation of ggml_print_backtrace_symbols (#8751)
140074bb8647df41840d6f32f4409fa8959bcf9f 6e2b6000e5fe808954a7dcef8225b5b7f2c1b9e9 Georgi Gerganov ggerganov@gmail.com 2024-07-30T15:58:57+03:00 GitHub noreply@github.com 2024-07-30T05:58:57-07:00 flake.lock: Update (#8729)
6e2b6000e5fe808954a7dcef8225b5b7f2c1b9e9 c887d8b01726b11ea03dbcaa9d44fa74422d0076 wangshuai09 391746016@qq.com 2024-07-30T18:37:35+08:00 GitHub noreply@github.com 2024-07-30T12:37:35+02:00 cann: update cmake (#8765)
c887d8b01726b11ea03dbcaa9d44fa74422d0076 75af08c475e285888f66556d0f459c533b7deb95 zhentaoyu zhentao.yu@intel.com 2024-07-30T14:56:51+08:00 GitHub noreply@github.com 2024-07-30T14:56:51+08:00 [SYCL] Add `TIMESTEP_EMBEDDING` OP (#8707)
75af08c475e285888f66556d0f459c533b7deb95 439b3fc75a8deb42899ac47a8f52aae75e0339fe CarterLi999 664681047@qq.com 2024-07-30T00:38:34+08:00 GitHub noreply@github.com 2024-07-29T18:38:34+02:00 ggml: bugfix: fix the inactive elements is agnostic for risc-v vector (#8748)
439b3fc75a8deb42899ac47a8f52aae75e0339fe 0832de723695ab400316a6c49b9f712380e3a731 R0CKSTAR xiaodong.ye@mthreads.com 2024-07-29T20:56:12+08:00 GitHub noreply@github.com 2024-07-29T14:56:12+02:00 cuda : organize vendor-specific headers into vendors directory (#8746)
0832de723695ab400316a6c49b9f712380e3a731 6eeaeba126ff701f3e8f79f246805b7023709972 Meng, Hengyu hengyu.meng@intel.com 2024-07-29T10:50:27+08:00 GitHub noreply@github.com 2024-07-29T10:50:27+08:00 [SYCL] add conv support (#8688)
6eeaeba126ff701f3e8f79f246805b7023709972 4730faca618ff9cee0780580145e3cbe86f24876 Johannes Gäßler johannesg@5d6.de 2024-07-28T22:32:44+02:00 GitHub noreply@github.com 2024-07-28T22:32:44+02:00 cmake: use 1 more thread for non-ggml in CI (#8740)
4730faca618ff9cee0780580145e3cbe86f24876 4c676c85e59ef8f771f3a129e6eb217552139231 Austin 77757836+teleprint-me@users.noreply.github.com 2024-07-28T03:52:42-04:00 GitHub noreply@github.com 2024-07-28T09:52:42+02:00 chore : Fix vulkan related compiler warnings, add help text, improve CLI options (#8477)
4c676c85e59ef8f771f3a129e6eb217552139231 e54c35e4fb5777c76316a50671640e6e144c9538 compilade git@compilade.net 2024-07-28T00:42:05-04:00 GitHub noreply@github.com 2024-07-28T00:42:05-04:00 llama : refactor session file management (#8699)
e54c35e4fb5777c76316a50671640e6e144c9538 5e2727fe0321c38d1664d26173c654fa1801dc5f R0CKSTAR yeahdongcn@gmail.com 2024-07-28T07:41:25+08:00 GitHub noreply@github.com 2024-07-28T01:41:25+02:00 feat: Support Moore Threads GPU (#8383)
5e2727fe0321c38d1664d26173c654fa1801dc5f 56f20aa25d5f97248a204b473c99f4040900f0e5 Georgi Gerganov ggerganov@gmail.com 2024-07-27T18:08:31+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T18:08:47+03:00 scripts : sync vulkan-shaders (#0)
56f20aa25d5f97248a204b473c99f4040900f0e5 345c8c0c87a97c1595f9c8b14833d531c8c7d8df Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:19:35+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T18:07:33+03:00 scripts : sync ggml-aarch64 sources
345c8c0c87a97c1595f9c8b14833d531c8c7d8df ae7985cd7beca3b849328d169a8d592469cd021f Georgi Gerganov ggerganov@gmail.com 2024-07-27T15:57:09+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:43:44+03:00 ggml : add missing semicolon (#0)
ae7985cd7beca3b849328d169a8d592469cd021f a05ca9369716a8319014cd1fc365980d43f8aae9 Georgi Gerganov ggerganov@gmail.com 2024-07-27T15:53:48+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:43:44+03:00 sync : ggml
a05ca9369716a8319014cd1fc365980d43f8aae9 9f77d899b7b0d56496f679e54b797da6199fed8e Mahesh Madhav 67384846+heshpdx@users.noreply.github.com 2024-07-25T00:54:08-07:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:43:44+03:00 ggml : loop tiling optimizations for scalar path (ggml/898)
9f77d899b7b0d56496f679e54b797da6199fed8e 203b7f1531303a060730ec1d1e01920e70302398 Ivan Filipov 159561759+vanaka11@users.noreply.github.com 2024-07-22T14:32:02+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:43:44+03:00 ggml: add support for float16 input tensors in pooling operations (ggml/895)
203b7f1531303a060730ec1d1e01920e70302398 d2b851bfa131478665315bc5c7c707506c14d703 Tony Wasserka 4840017+neobrain@users.noreply.github.com 2024-07-20T20:49:44+02:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:43:44+03:00 vulkan : initialize vk_buffer_struct members to VK_NULL_HANDLE (ggml/893)
d2b851bfa131478665315bc5c7c707506c14d703 c12b6e8ee7d905e0f299caf311689189fb1b4ac5 Borislav Stanimirov b.stanimirov@abv.bg 2024-07-12T17:24:20+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:43:44+03:00 cmake : only enable GGML_NATIVE and x86 flags if not crosscompiling (ggml/885)
c12b6e8ee7d905e0f299caf311689189fb1b4ac5 b5e95468b1676e1e5c9d80d1eeeb26f542a38f42 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-08T12:03:42+02:00 Georgi Gerganov ggerganov@gmail.com 2024-07-27T17:43:44+03:00 ggml : remove unnecessary UNUSED macro call (ggml/880)
b5e95468b1676e1e5c9d80d1eeeb26f542a38f42 92090eca212650727e38b335c1d4accfbcc9b79c Jeffrey Morgan jmorganca@gmail.com 2024-07-27T05:03:45-07:00 GitHub noreply@github.com 2024-07-27T15:03:45+03:00 llama : add support for llama 3.1 rope scaling factors (#8676)
92090eca212650727e38b335c1d4accfbcc9b79c 9d03d085dd6cb275c078690bb64073b9b043e95f Georgi Gerganov ggerganov@gmail.com 2024-07-27T14:59:29+03:00 GitHub noreply@github.com 2024-07-27T14:59:29+03:00 llama : add function for model-based max number of graph nodes (#8622)
9d03d085dd6cb275c078690bb64073b9b043e95f bfb4c74981f0a40d757b450b596a9fe4ca983d26 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-27T12:45:02+02:00 GitHub noreply@github.com 2024-07-27T13:45:02+03:00 common : add --no-warmup option for main/llama-cli (#8712)
bfb4c74981f0a40d757b450b596a9fe4ca983d26 2b1f616b208a4a21c4ee7a7eb85d822ff1d787af wangshuai09 391746016@qq.com 2024-07-27T16:36:44+08:00 GitHub noreply@github.com 2024-07-27T16:36:44+08:00 cann: Fix Multi-NPU execution error (#8710)
2b1f616b208a4a21c4ee7a7eb85d822ff1d787af 01245f5b1629075543bc4478418c7d72a0b4b3c7 slaren slarengh@gmail.com 2024-07-27T04:41:55+02:00 GitHub noreply@github.com 2024-07-27T04:41:55+02:00 ggml : reduce hash table reset cost (#8698)
01245f5b1629075543bc4478418c7d72a0b4b3c7 01aec4a6310ab0160483196db0e726d78d4c94b6 Judd foldl@users.noreply.github.com 2024-07-26T16:38:12+08:00 GitHub noreply@github.com 2024-07-26T11:38:12+03:00 llama : fix order of parameters (#8706)
01aec4a6310ab0160483196db0e726d78d4c94b6 41cd47caab88c442edc50e90c8d8d0ac3e82768d Yaiko elyaiko@hotmail.com 2024-07-25T18:10:16-04:00 GitHub noreply@github.com 2024-07-26T00:10:16+02:00 server : add Speech Recognition & Synthesis to UI (#8679)
41cd47caab88c442edc50e90c8d8d0ac3e82768d 49ce0ab6d45402e8bb622bf86f86529f2b0ba552 Xuan Son Nguyen thichthat@gmail.com 2024-07-25T23:49:39+02:00 GitHub noreply@github.com 2024-07-25T23:49:39+02:00 examples : export-lora : fix issue with quantized base models (#8687)
49ce0ab6d45402e8bb622bf86f86529f2b0ba552 4226a8d10e3904db3a1297919fe6c7f06beba6c0 DavidKorczynski david@adalogics.com 2024-07-25T22:23:05+01:00 GitHub noreply@github.com 2024-07-25T23:23:05+02:00 ggml: handle ggml_init failure to fix NULL pointer deref (#8692)
4226a8d10e3904db3a1297919fe6c7f06beba6c0 bf5a81df375f1c71e41462e1f48d57db359c9e80 Georgi Gerganov ggerganov@gmail.com 2024-07-25T19:57:31+03:00 GitHub noreply@github.com 2024-07-25T19:57:31+03:00 llama : fix build + fix fabs compile warnings (#8683)
bf5a81df375f1c71e41462e1f48d57db359c9e80 88954f7fbd31aeb8c75140edee03e7a8ad5e2d9c Andreas (Andi) Kunar andreask@msn.com 2024-07-25T18:01:00+02:00 GitHub noreply@github.com 2024-07-25T19:01:00+03:00 ggml : fix build on Windows with Snapdragon X (#8531)
88954f7fbd31aeb8c75140edee03e7a8ad5e2d9c ed67bcb24f2d6ac0072cae72620b2bd971741b98 Georgi Gerganov ggerganov@gmail.com 2024-07-25T18:57:44+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-25T18:58:04+03:00 tests : fix printfs (#8068)
ed67bcb24f2d6ac0072cae72620b2bd971741b98 eddcb5238b2e09a37798b87cde1244017a194bcc Chen Xi xi2.chen@intel.com 2024-07-25T11:45:18Z GitHub noreply@github.com 2024-07-25T19:45:18+08:00 [SYCL] fix multi-gpu issue on sycl (#8554)
eddcb5238b2e09a37798b87cde1244017a194bcc be6d7c079173d941b4f784500f9148f46cec2724 Georgi Gerganov ggerganov@gmail.com 2024-07-25T12:37:42+03:00 GitHub noreply@github.com 2024-07-25T12:37:42+03:00 ggml : add and use ggml_cpu_has_llamafile() (#8664)
be6d7c079173d941b4f784500f9148f46cec2724 4b0eff3df58d8d86e47348fb73d54da3194d416d Xuan Son Nguyen thichthat@gmail.com 2024-07-25T10:39:04+02:00 GitHub noreply@github.com 2024-07-25T10:39:04+02:00 examples : remove `finetune` and `train-text-from-scratch` (#8669)
4b0eff3df58d8d86e47348fb73d54da3194d416d 8a4bad50a8ed24ed1e9df003521468dcc37320e8 Ujjawal Panchal 31011628+Ujjawal-K-Panchal@users.noreply.github.com 2024-07-25T13:43:27+05:30 GitHub noreply@github.com 2024-07-25T11:13:27+03:00 docs : Quantum -> Quantized (#8666)
8a4bad50a8ed24ed1e9df003521468dcc37320e8 68504f0970db5a3602d176953690f503059906b1 Fan Shupei dymarkfan@outlook.com 2024-07-25T15:21:09+08:00 GitHub noreply@github.com 2024-07-25T10:21:09+03:00 llama: use sliding window for phi3 (#8627)
68504f0970db5a3602d176953690f503059906b1 f19bf99c015d3d745143e8bb4f056e0ea015ad40 MorganRO8 47795945+MorganRO8@users.noreply.github.com 2024-07-24T12:48:00-04:00 GitHub noreply@github.com 2024-07-24T19:48:00+03:00 readme : update games list (#8673)
f19bf99c015d3d745143e8bb4f056e0ea015ad40 3a7ac5300a7e8ebbe4a3eb5aff9dba11ed76ea61 Joe Todd joe.todd@codeplay.com 2024-07-24T14:36:00+01:00 GitHub noreply@github.com 2024-07-24T14:36:00+01:00 Build Llama SYCL Intel with static libs (#8668)
3a7ac5300a7e8ebbe4a3eb5aff9dba11ed76ea61 96952e7181929c6001b2bc69a33f240de731cc3a Thorsten Sommer SommerEngineering@users.noreply.github.com 2024-07-24T14:52:30+02:00 GitHub noreply@github.com 2024-07-24T15:52:30+03:00 readme : update UI list [no ci] (#8505)
96952e7181929c6001b2bc69a33f240de731cc3a 79167d9e49aef9caa98e13ee7ca067ec9f88b4b5 Xuan Son Nguyen thichthat@gmail.com 2024-07-24T13:48:46+02:00 GitHub noreply@github.com 2024-07-24T13:48:46+02:00 llama : fix `llama_chat_format_single` for mistral (#8657)
79167d9e49aef9caa98e13ee7ca067ec9f88b4b5 b115105f05e3372bc75b2a486c1930c365fd2846 Joe Todd joe.todd@codeplay.com 2024-07-24T11:55:26+01:00 GitHub noreply@github.com 2024-07-24T11:55:26+01:00 Re-add erroneously removed -fsycl from GGML_EXTRA_LIBS (#8667)
b115105f05e3372bc75b2a486c1930c365fd2846 de280085e7917dbb7f5753de5842ff4455f82a81 Xuan Son Nguyen thichthat@gmail.com 2024-07-24T11:25:19+02:00 GitHub noreply@github.com 2024-07-24T11:25:19+02:00 add llama_lora_adapter_clear (#8653)
de280085e7917dbb7f5753de5842ff4455f82a81 b841d0740855c5af1344a81f261139a45a2b39ee Xuan Son Nguyen thichthat@gmail.com 2024-07-23T23:48:37+02:00 GitHub noreply@github.com 2024-07-23T23:48:37+02:00 examples : Fix `llama-export-lora` example (#8607)
b841d0740855c5af1344a81f261139a45a2b39ee 64cf50a0ed62d41e4f6c13e08a9b6b0816f46c6e Vali Malinoiu 0x4139@gmail.com 2024-07-23T17:37:42+03:00 GitHub noreply@github.com 2024-07-23T17:37:42+03:00 server : fix URL.parse in the UI (#8646)
64cf50a0ed62d41e4f6c13e08a9b6b0816f46c6e 938943cdbf4dd79005a394d732bc226f9e34e0ff Joe Todd joe.todd@codeplay.com 2024-07-23T14:58:37+01:00 GitHub noreply@github.com 2024-07-23T14:58:37+01:00 sycl : Add support for non-release DPC++ & oneMKL (#8644)
938943cdbf4dd79005a394d732bc226f9e34e0ff 751fcfc6c33ea5f43cadd4d976f8fb176871df5e Georgi Gerganov ggerganov@gmail.com 2024-07-23T13:10:17+03:00 GitHub noreply@github.com 2024-07-23T13:10:17+03:00 llama : move vocab, grammar and sampling into separate files (#8508)
751fcfc6c33ea5f43cadd4d976f8fb176871df5e 46e47417aa4f18c08738afd4d9a3e838e97ca03f 0cc4m picard12@live.de 2024-07-23T10:56:49+02:00 GitHub noreply@github.com 2024-07-23T10:56:49+02:00 Vulkan IQ4_NL Support (#8613)
46e47417aa4f18c08738afd4d9a3e838e97ca03f e7e6487ba06634edf58dfdf9673bad9df41b445a Jeroen Mostert jeroen.mostert@cm.com 2024-07-23T10:50:40+02:00 GitHub noreply@github.com 2024-07-23T10:50:40+02:00 Allow all RDNA2 archs to use sdot4 intrinsic (#8629)
e7e6487ba06634edf58dfdf9673bad9df41b445a 063d99ad11f1295046610ce5b97e105849a4b573 Georgi Gerganov ggerganov@gmail.com 2024-07-23T11:28:38+03:00 GitHub noreply@github.com 2024-07-23T11:28:38+03:00 contrib : clarify PR squashing + module names (#8630)
063d99ad11f1295046610ce5b97e105849a4b573 081fe431aa8fb6307145c4feb3eed4f48cab19f8 luoyu-intel yu.luo@intel.com 2024-07-23T07:43:28Z GitHub noreply@github.com 2024-07-23T15:43:28+08:00 [SYCL] fix scratch size of softmax (#8642)
081fe431aa8fb6307145c4feb3eed4f48cab19f8 d94c6e0ccbd29ee1ba4f44e9caa8682ad94df9fa Keke Han hankeke303@163.com 2024-07-23T00:43:43+08:00 GitHub noreply@github.com 2024-07-22T19:43:43+03:00 llama : fix codeshell support (#8599)
d94c6e0ccbd29ee1ba4f44e9caa8682ad94df9fa 566daa5a5b38018b2727950bbd280239adb981b6 Jason Stillerman jason.t.stillerman@gmail.com 2024-07-22T10:43:01-04:00 GitHub noreply@github.com 2024-07-22T17:43:01+03:00 llama : add support for SmolLm pre-tokenizer (#8609)
566daa5a5b38018b2727950bbd280239adb981b6 6f11a83e4e7700fdf353ed4a29599cb662c792f6 Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2024-07-22T15:44:53+02:00 GitHub noreply@github.com 2024-07-22T23:44:53+10:00 *.py: Stylistic adjustments for python (#8233)
6f11a83e4e7700fdf353ed4a29599cb662c792f6 e093dd238282a980d248db0786063f8174400f70 Georgi Gerganov ggerganov@gmail.com 2024-07-22T13:33:22+03:00 GitHub noreply@github.com 2024-07-22T13:33:22+03:00 llama : allow overrides for tokenizer flags (#8614)
e093dd238282a980d248db0786063f8174400f70 50e05353e88d50b644688caa91f5955e8bdb9eb9 Georgi Gerganov ggerganov@gmail.com 2024-07-22T13:32:49+03:00 GitHub noreply@github.com 2024-07-22T13:32:49+03:00 tests : re-enable tokenizer tests (#8611)
50e05353e88d50b644688caa91f5955e8bdb9eb9 628154492a0b2dcc954a3af99e5c267097568eae Douglas Hanley thesecretaryofwar@gmail.com 2024-07-22T03:06:17-05:00 GitHub noreply@github.com 2024-07-22T11:06:17+03:00 llama : add Mistral Nemo inference support (#8604)
628154492a0b2dcc954a3af99e5c267097568eae 04bab6b7da0ed2b0a57174a57784d602aabb6c10 Jan Boon jan.boon@kaetemi.be 2024-07-22T16:02:09+08:00 GitHub noreply@github.com 2024-07-22T11:02:09+03:00 server : update doc to clarify n_keep when there is bos token (#8619)
04bab6b7da0ed2b0a57174a57784d602aabb6c10 b7c11d36e605b35206901d0e21905f1b99508e33 Mark Zhuang zhuangqiubin@gmail.com 2024-07-22T15:56:45+08:00 GitHub noreply@github.com 2024-07-22T10:56:45+03:00 ggml: fix compile error for RISC-V (#8623)
b7c11d36e605b35206901d0e21905f1b99508e33 45f2c19cc57286eead7b232ce8028273a817aa4d devojony 61173062+devojony@users.noreply.github.com 2024-07-22T14:54:42+08:00 GitHub noreply@github.com 2024-07-22T09:54:42+03:00 examples: fix android example cannot be generated continuously (#8621)
45f2c19cc57286eead7b232ce8028273a817aa4d 22f281aa16f44d8f6ec2c180a0685ff27e04e714 Georgi Gerganov ggerganov@gmail.com 2024-07-21T16:45:10+03:00 GitHub noreply@github.com 2024-07-21T06:45:10-07:00 flake.lock: Update (#8610)
22f281aa16f44d8f6ec2c180a0685ff27e04e714 328884f4219c0228673cf1870ac63987fb4f9fd0 M-A maruel@gmail.com 2024-07-20T22:09:17-04:00 GitHub noreply@github.com 2024-07-20T22:09:17-04:00 examples : Rewrite pydantic_models_to_grammar_examples.py (#8493)
328884f4219c0228673cf1870ac63987fb4f9fd0 c69c63039cd75f8f33f253ab7485d82a8b4cd403 compilade git@compilade.net 2024-07-20T21:58:49-04:00 GitHub noreply@github.com 2024-07-20T21:58:49-04:00 gguf-py : fix some metadata name extraction edge cases (#8591)
c69c63039cd75f8f33f253ab7485d82a8b4cd403 69c487f4ed57bb4d4514a1b7ff12608d5a8e7ef0 compilade git@compilade.net 2024-07-20T21:53:01-04:00 GitHub noreply@github.com 2024-07-20T21:53:01-04:00 convert_hf : fix Gemma v1 conversion (#8597)
69c487f4ed57bb4d4514a1b7ff12608d5a8e7ef0 07283b1a90e1320aae4762c7e03c879043910252 Johannes Gäßler johannesg@5d6.de 2024-07-20T22:25:26+02:00 GitHub noreply@github.com 2024-07-20T22:25:26+02:00 CUDA: MMQ code deduplication + iquant support (#8495)
07283b1a90e1320aae4762c7e03c879043910252 940362224d20e35f13aa5fd34a0d937ae57bdf7d Georgi Gerganov ggerganov@gmail.com 2024-07-20T17:15:42+03:00 GitHub noreply@github.com 2024-07-20T17:15:42+03:00 gguf : handle null name during init (#8587)
940362224d20e35f13aa5fd34a0d937ae57bdf7d 69b9945b44c3057ec17cb556994cd36060455d44 Michael Coppola m18coppola@gmail.com 2024-07-20T09:43:51-04:00 GitHub noreply@github.com 2024-07-20T16:43:51+03:00 llama : add support for Tekken pre-tokenizer (#8579)
69b9945b44c3057ec17cb556994cd36060455d44 c3776cacabce2ee35f172fb72be7a519752125fa Huifeng Ou 79071290+ho2103@users.noreply.github.com 2024-07-20T09:09:37-04:00 GitHub noreply@github.com 2024-07-20T16:09:37+03:00 llama.swiftui: fix end of generation bug (#8268)
c3776cacabce2ee35f172fb72be7a519752125fa 87e397d00bdcedd5cbf6dfda06a7b0f302462728 Brian mofosyne@gmail.com 2024-07-20T17:35:25+10:00 GitHub noreply@github.com 2024-07-20T17:35:25+10:00 gguf_dump.py: fix markddown kv array print (#8588)
87e397d00bdcedd5cbf6dfda06a7b0f302462728 57b1d4f9eb6f2c139b31ea79626d954b261e1051 slaren slarengh@gmail.com 2024-07-19T17:17:27+02:00 GitHub noreply@github.com 2024-07-19T17:17:27+02:00 ggml : fix quant dot product with odd number of blocks (#8549)
57b1d4f9eb6f2c139b31ea79626d954b261e1051 d19754553029296c46d40b2f7bd4e2c2f531a8c5 Brian mofosyne@gmail.com 2024-07-20T00:04:38+10:00 GitHub noreply@github.com 2024-07-20T00:04:38+10:00 convert-*.py: remove add_name from ChatGLMModel class (#8590)
d19754553029296c46d40b2f7bd4e2c2f531a8c5 be0cfb41752551a4680ee7dfd29f2a05b50db442 Georgi Gerganov ggerganov@gmail.com 2024-07-19T16:50:47+03:00 GitHub noreply@github.com 2024-07-19T16:50:47+03:00 llama : bump max layers from 256 to 512 (#8530)
be0cfb41752551a4680ee7dfd29f2a05b50db442 b57eb9ca4fb79f3163c6a69e154ff76157a4f716 Georgi Gerganov ggerganov@gmail.com 2024-07-19T14:34:55+03:00 GitHub noreply@github.com 2024-07-19T14:34:55+03:00 readme : fix server badge
b57eb9ca4fb79f3163c6a69e154ff76157a4f716 f299aa98ecc19cbc574e9d698e03999e89de3d3d Clint Herron hanclinto@gmail.com 2024-07-19T07:05:45-04:00 GitHub noreply@github.com 2024-07-19T14:05:45+03:00 ggml : add friendlier error message to fopen errors (#8575)
f299aa98ecc19cbc574e9d698e03999e89de3d3d 3d0e4367d99087892e355ddbeebd232a0b2f40de Frank Mai thxcode0824@gmail.com 2024-07-19T17:44:41+08:00 GitHub noreply@github.com 2024-07-19T11:44:41+02:00 fix: typo of chatglm4 chat tmpl (#8586)
3d0e4367d99087892e355ddbeebd232a0b2f40de a15ef8f8a08e12f9c5162c221e67779e71182073 Brian mofosyne@gmail.com 2024-07-19T17:51:51+10:00 GitHub noreply@github.com 2024-07-19T17:51:51+10:00 convert-*.py: add general.name kv override (#8571)
a15ef8f8a08e12f9c5162c221e67779e71182073 705b7ecf60e667ced57c15d67aa86865e3cc7aa7 Johannes Gäßler johannesg@5d6.de 2024-07-18T23:48:47+02:00 GitHub noreply@github.com 2024-07-18T23:48:47+02:00 CUDA: fix partial offloading for ne0 % 256 != 0 (#8572)
705b7ecf60e667ced57c15d67aa86865e3cc7aa7 0d2c7321e9678e91b760ebe57f0d063856bb018b 65a 10104049+65a@users.noreply.github.com 2024-07-18T07:47:12-07:00 GitHub noreply@github.com 2024-07-18T17:47:12+03:00 cmake : install all ggml public headers (#8480)
0d2c7321e9678e91b760ebe57f0d063856bb018b 672a6f101839a150180fc28891ebed379c8f2353 Eric Zhang 34133756+EZForever@users.noreply.github.com 2024-07-18T18:43:49+08:00 GitHub noreply@github.com 2024-07-18T12:43:49+02:00 server: use relative routes for static files in new UI (#8552)
672a6f101839a150180fc28891ebed379c8f2353 3807c3de04cde853418033c95e96642876545f3e Brian mofosyne@gmail.com 2024-07-18T20:40:15+10:00 GitHub noreply@github.com 2024-07-18T20:40:15+10:00 convert-*.py: GGUF Naming Convention Refactor and Metadata Override Refactor (#7499)
3807c3de04cde853418033c95e96642876545f3e e02b597be3702174e7b47b44cd03e1da1553284b RunningLeon mnsheng@yeah.net 2024-07-18T16:06:22+08:00 GitHub noreply@github.com 2024-07-18T11:06:22+03:00 server : respect `--special` cli arg (#8553)
e02b597be3702174e7b47b44cd03e1da1553284b b3283448ce9a5098226afe1d8648ccc578511fe4 Johannes Gäßler johannesg@5d6.de 2024-07-17T23:35:44+02:00 GitHub noreply@github.com 2024-07-17T23:35:44+02:00 lookup: fibonacci hashing, fix crashes (#8548)
b3283448ce9a5098226afe1d8648ccc578511fe4 30f80ca0bcee58669ada7a94244eeccc8c4807cc Al Mochkin 14274697+amochkin@users.noreply.github.com 2024-07-17T20:21:55+02:00 GitHub noreply@github.com 2024-07-17T20:21:55+02:00 build : Fix docker build warnings (#8535) (#8537)
30f80ca0bcee58669ada7a94244eeccc8c4807cc 1bdd8ae19f50bc6f108fa247e90688e5c60559fc Brian mofosyne@gmail.com 2024-07-18T00:57:06+10:00 GitHub noreply@github.com 2024-07-17T17:57:06+03:00 CONTRIBUTING.md : remove mention of noci (#8541)
1bdd8ae19f50bc6f108fa247e90688e5c60559fc da3913d8f9475b0d4bcbeb4936c724af4eade092 hipudding huafengchun@gmail.com 2024-07-17T19:23:50+08:00 GitHub noreply@github.com 2024-07-17T14:23:50+03:00 [CANN] Add Ascend NPU backend (#6035)
da3913d8f9475b0d4bcbeb4936c724af4eade092 d65a8361fed8be97389776fb94217e937ec6b03c Masaya, Kato 62578291+msy-kato@users.noreply.github.com 2024-07-17T16:34:28+09:00 GitHub noreply@github.com 2024-07-17T10:34:28+03:00 batched: fix n_predict parameter (#8527)
d65a8361fed8be97389776fb94217e937ec6b03c 5e116e8dd51775f8f1c090570be148d5d7eea6c3 Georgi Gerganov ggerganov@gmail.com 2024-07-17T10:32:59+03:00 GitHub noreply@github.com 2024-07-17T10:32:59+03:00 llama : disable context-shift for DeepSeek v2 (#8501)
5e116e8dd51775f8f1c090570be148d5d7eea6c3 1666f92dcda14e002cbb08e1028f9fff341d73ad Johannes Gäßler johannesg@5d6.de 2024-07-16T21:20:59+02:00 GitHub noreply@github.com 2024-07-16T21:20:59+02:00 make/cmake: add missing force MMQ/cuBLAS for HIP (#8515)
1666f92dcda14e002cbb08e1028f9fff341d73ad 37b12f92ab696d70f9a65d7447ce721b094fb32e Brian mofosyne@gmail.com 2024-07-16T17:14:16+10:00 GitHub noreply@github.com 2024-07-16T10:14:16+03:00 gguf-hash : update clib.json to point to original xxhash repo (#8491)
37b12f92ab696d70f9a65d7447ce721b094fb32e 0efec57787cb8d8d76b17cd3765e6521e22c1f19 Steve Bonds sbonds@gmail.com 2024-07-16T00:04:45-07:00 GitHub noreply@github.com 2024-07-16T10:04:45+03:00 export-lora : handle help argument (#8497)
0efec57787cb8d8d76b17cd3765e6521e22c1f19 7acfd4e8d55082c1b597dfc3ffe04fb5d530c6dc Georgi Gerganov ggerganov@gmail.com 2024-07-16T10:00:30+03:00 GitHub noreply@github.com 2024-07-16T10:00:30+03:00 llama : valign + remove unused ftype (#8502)
7acfd4e8d55082c1b597dfc3ffe04fb5d530c6dc 97bdd26eee11fe109dec00de75690ceef61c03f2 compilade git@compilade.net 2024-07-15T23:13:10-04:00 GitHub noreply@github.com 2024-07-15T23:13:10-04:00 convert_hf : faster lazy safetensors (#8482)
97bdd26eee11fe109dec00de75690ceef61c03f2 4db8f60fe79a391e82b0464013ada123baced96a Xuan Son Nguyen thichthat@gmail.com 2024-07-15T20:50:47+02:00 GitHub noreply@github.com 2024-07-15T20:50:47+02:00 Refactor lora adapter support (#8332)
4db8f60fe79a391e82b0464013ada123baced96a 8fac431b0692e88cdc55250f29f8d4386be82c5d Xuan Son Nguyen thichthat@gmail.com 2024-07-15T19:23:10+02:00 GitHub noreply@github.com 2024-07-15T19:23:10+02:00 fix ci (#8494)
8fac431b0692e88cdc55250f29f8d4386be82c5d f17f39ff9cb71fa7b0eda0c649bb0ce9b2d8a6b8 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-15T14:48:17+02:00 GitHub noreply@github.com 2024-07-15T15:48:17+03:00 ggml : suppress unknown pragma 'GCC' on windows (#8460)
f17f39ff9cb71fa7b0eda0c649bb0ce9b2d8a6b8 9104bc20edf47f74dc49379b7d61d0c6e85a4882 M-A maruel@gmail.com 2024-07-15T08:04:56-04:00 GitHub noreply@github.com 2024-07-15T15:04:56+03:00 server: update README.md with llama-server --help output [no ci] (#8472)
9104bc20edf47f74dc49379b7d61d0c6e85a4882 fc690b018e459012cd82c37f1343e9bb658987d1 Georgi Gerganov ggerganov@gmail.com 2024-07-15T14:54:58+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-15T14:54:58+03:00 common : add --no-cont-batching arg (#6358)
fc690b018e459012cd82c37f1343e9bb658987d1 16bdfa42acb09175e88cf97e9d9e4e48f616d120 NikolaiLyssogor 59844691+NikolaiLyssogor@users.noreply.github.com 2024-07-15T04:46:39-07:00 GitHub noreply@github.com 2024-07-15T14:46:39+03:00 docs: fix links in development docs [no ci] (#8481)
16bdfa42acb09175e88cf97e9d9e4e48f616d120 3dfda05956befb350745c5c2f7134d06adfe8724 Meng, Hengyu hengyu.meng@intel.com 2024-07-15T19:32:15+08:00 GitHub noreply@github.com 2024-07-15T19:32:15+08:00 [SYCL] add concat through dim 1/2 (#8483)
3dfda05956befb350745c5c2f7134d06adfe8724 bda62d7999caa8c222b6c354ac1e7c7442508539 Georgi Gerganov ggerganov@gmail.com 2024-07-15T14:10:39+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-15T14:10:39+03:00 llama : de-duplicate deepseek2 norm
bda62d7999caa8c222b6c354ac1e7c7442508539 090fca7a073efe9ad3dd2b9ac12491a4f20ac957 0cc4m picard12@live.de 2024-07-15T09:38:52+02:00 GitHub noreply@github.com 2024-07-15T09:38:52+02:00 Vulkan MMQ Fix (#8479)
090fca7a073efe9ad3dd2b9ac12491a4f20ac957 aaab2419eaa17ab3aa38f4ba49c7eea406999e99 compilade git@compilade.net 2024-07-14T19:51:21-04:00 GitHub noreply@github.com 2024-07-14T19:51:21-04:00 pydantic : replace uses of __annotations__ with get_type_hints (#8474)
aaab2419eaa17ab3aa38f4ba49c7eea406999e99 73cf442e7bc9baca7b3e213b261551812f1676c9 Georgi Gerganov ggerganov@gmail.com 2024-07-14T18:54:02+03:00 GitHub noreply@github.com 2024-07-14T08:54:02-07:00 flake.lock: Update (#8475)
73cf442e7bc9baca7b3e213b261551812f1676c9 e236528e7628a0e59751eee9addf21fc3c33d376 Georgi Gerganov ggerganov@gmail.com 2024-07-14T14:05:09+03:00 GitHub noreply@github.com 2024-07-14T14:05:09+03:00 llama : fix Gemma-2 Query scaling factors (#8473)
e236528e7628a0e59751eee9addf21fc3c33d376 fa79495bb4897953a75607addd9d2cdd2ec63222 Brian mofosyne@gmail.com 2024-07-14T16:47:14+10:00 GitHub noreply@github.com 2024-07-14T16:47:14+10:00 gguf_hash.py: Add sha256 (#8470)
fa79495bb4897953a75607addd9d2cdd2ec63222 17eb6aa8a992cda37ee65cf848d9289bd6cad860 compilade git@compilade.net 2024-07-13T23:35:10-04:00 GitHub noreply@github.com 2024-07-13T23:35:10-04:00 llama : fix pre-tokenization of non-special added tokens (#8228)
17eb6aa8a992cda37ee65cf848d9289bd6cad860 c917b67f06c42d8ca8391b9bc73f5fe62c83bf70 bandoti 141645996+bandoti@users.noreply.github.com 2024-07-13T13:12:39-03:00 GitHub noreply@github.com 2024-07-13T18:12:39+02:00 vulkan : cmake integration (#8119)
c917b67f06c42d8ca8391b9bc73f5fe62c83bf70 4e24cffd8cccd653634e24ee461c252bd77b1426 Georgi Gerganov ggerganov@gmail.com 2024-07-13T18:32:33+03:00 GitHub noreply@github.com 2024-07-13T18:32:33+03:00 metal : template-ify some of the kernels (#8447)
4e24cffd8cccd653634e24ee461c252bd77b1426 6af51c0d96e6268769fc05c98d5b1a5e832c0017 Georgi Gerganov ggerganov@gmail.com 2024-07-12T14:48:15+03:00 GitHub noreply@github.com 2024-07-12T14:48:15+03:00 server : handle content array in chat API (#8449)
6af51c0d96e6268769fc05c98d5b1a5e832c0017 f53226245f421bd01b47cce43a47e791de82c636 Georgi Gerganov ggerganov@gmail.com 2024-07-12T14:48:04+03:00 GitHub noreply@github.com 2024-07-12T14:48:04+03:00 main : print error on empty input (#8456)
f53226245f421bd01b47cce43a47e791de82c636 c3ebcfa148e867a68e78fd5c4f0c23e8f84c788b Daniel Bevenius daniel.bevenius@gmail.com 2024-07-12T11:05:21+02:00 GitHub noreply@github.com 2024-07-12T12:05:21+03:00 llama : suppress unary minus operator warning (#8448)
c3ebcfa148e867a68e78fd5c4f0c23e8f84c788b 8a4441ea1a2564578134404f31158c318e9c0bf3 Douglas Hanley thesecretaryofwar@gmail.com 2024-07-12T03:14:12-05:00 GitHub noreply@github.com 2024-07-12T11:14:12+03:00 server : ensure batches are either all embed or all completion (#8420)
8a4441ea1a2564578134404f31158c318e9c0bf3 5aefbce27a66473d4b1263ba3f7bdd3d14245975 Armen Kaleshian kriation@users.noreply.github.com 2024-07-12T04:08:19-04:00 GitHub noreply@github.com 2024-07-12T11:08:19+03:00 docker : fix filename for convert-hf-to-gguf.py in tools.sh (#8441)
5aefbce27a66473d4b1263ba3f7bdd3d14245975 71c1121d11f1437be9421fd0cbaa011b9ef49098 Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2024-07-12T10:06:33+02:00 GitHub noreply@github.com 2024-07-12T11:06:33+03:00 convert : remove fsep token from GPTRefactForCausalLM (#8237)
71c1121d11f1437be9421fd0cbaa011b9ef49098 370b1f7e7a7514d9a63daf15f7b0f00319b7f908 Georgi Gerganov ggerganov@gmail.com 2024-07-12T10:46:14+03:00 GitHub noreply@github.com 2024-07-12T10:46:14+03:00 examples : sprintf -> snprintf (#8434)
370b1f7e7a7514d9a63daf15f7b0f00319b7f908 b549a1bbefb2f1fbb8b558bac1f2ae7967e60964 Georgi Gerganov ggerganov@gmail.com 2024-07-12T10:46:02+03:00 GitHub noreply@github.com 2024-07-12T10:46:02+03:00 ggml : minor naming changes (#8433)
b549a1bbefb2f1fbb8b558bac1f2ae7967e60964 368645698ab648e390dcd7c00a2bf60efa654f57 Chen Xi xixichen08@foxmail.com 2024-07-12T00:52:04Z GitHub noreply@github.com 2024-07-12T08:52:04+08:00 [SYCL] fix the mul_mat_id ut issues (#8427)
368645698ab648e390dcd7c00a2bf60efa654f57 b078c619aa4e97fe726d61b0b5499a2e19a418a4 Nicholai Tukanov nicholaitukanov@gmail.com 2024-07-11T11:49:15-05:00 GitHub noreply@github.com 2024-07-11T18:49:15+02:00 ggml : add NVPL BLAS support (#8329) (#8425)
b078c619aa4e97fe726d61b0b5499a2e19a418a4 808aba39161e5d7ca2ff24110b5aa14d2e536988 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-11T17:53:42+02:00 GitHub noreply@github.com 2024-07-11T17:53:42+02:00 cuda : suppress 'noreturn' warn in no_device_code (#8414)
808aba39161e5d7ca2ff24110b5aa14d2e536988 a977c115448e40856fb9cbe3ceb6d8ce802553b0 Johannes Gäßler johannesg@5d6.de 2024-07-11T16:47:47+02:00 GitHub noreply@github.com 2024-07-11T16:47:47+02:00 CUDA: optimize and refactor MMQ (#8416)
a977c115448e40856fb9cbe3ceb6d8ce802553b0 9a55ffe6fbf7f19c865f8f277ca7cd585cc0b094 Georgi Gerganov ggerganov@gmail.com 2024-07-11T11:20:40+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-11T11:20:40+03:00 gitignore : deprecated binaries
9a55ffe6fbf7f19c865f8f277ca7cd585cc0b094 7a221b672e49dfae459b1af27210ba3f2b5419b6 compilade git@compilade.net 2024-07-11T03:41:48-04:00 GitHub noreply@github.com 2024-07-11T10:41:48+03:00 tokenize : add --no-parse-special option (#8423)
7a221b672e49dfae459b1af27210ba3f2b5419b6 278d0e18469aacf505be18ce790a63c7cc31be26 Georgi Gerganov ggerganov@gmail.com 2024-07-11T10:21:30+03:00 GitHub noreply@github.com 2024-07-11T10:21:30+03:00 llama : use F32 precision in Qwen2 attention and no FA (#8412)
278d0e18469aacf505be18ce790a63c7cc31be26 dd07a123b79f9bd9e8a4ba0447427b3083e9347a Clint Herron hanclinto@gmail.com 2024-07-10T20:08:17-04:00 GitHub noreply@github.com 2024-07-10T20:08:17-04:00 Initialize default slot sampling parameters from the global context. (#8418)
dd07a123b79f9bd9e8a4ba0447427b3083e9347a f4444d992c16b6b9442f4770c7c3a10b19a08343 Clint Herron hanclinto@gmail.com 2024-07-10T12:35:18-04:00 GitHub noreply@github.com 2024-07-10T12:35:18-04:00 Name Migration: Build the deprecation-warning 'main' binary every time (#8404)
f4444d992c16b6b9442f4770c7c3a10b19a08343 6b2a849d1f43d46b82d2f9c08c3275137b528784 AidanBeltonS aidan.belton@codeplay.com 2024-07-10T16:10:49+01:00 GitHub noreply@github.com 2024-07-10T16:10:49+01:00 [SYCL] Use multi_ptr to clean up deprecated warnings (#8256)
6b2a849d1f43d46b82d2f9c08c3275137b528784 0f1a39f3439825acf7e3a1663566d410be152170 Georgi Gerganov ggerganov@gmail.com 2024-07-10T15:23:29+03:00 GitHub noreply@github.com 2024-07-10T15:23:29+03:00 ggml : move sgemm sources to llamafile subfolder (#8394)
0f1a39f3439825acf7e3a1663566d410be152170 83321c6958acf20747d288031174cc1bf8816e33 Dibakar Gope dibakar.gope@arm.com 2024-07-10T07:14:51-05:00 GitHub noreply@github.com 2024-07-10T15:14:51+03:00 ggml : add AArch64 optimized GEMV and GEMM Q4 kernels (#5780)
83321c6958acf20747d288031174cc1bf8816e33 cc61948b1f97165b46990f4c2d9699bf9bb64443 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2024-07-10T15:12:35+03:00 GitHub noreply@github.com 2024-07-10T15:12:35+03:00 gguf-py rel pipeline (#8410)
cc61948b1f97165b46990f4c2d9699bf9bb64443 7a80710d93ee0f4e0d335d65984ae747e62c0337 Borislav Stanimirov b.stanimirov@abv.bg 2024-07-10T14:45:44+03:00 GitHub noreply@github.com 2024-07-10T14:45:44+03:00 llama : C++20 compatibility for u8 strings (#8408)
7a80710d93ee0f4e0d335d65984ae747e62c0337 a8be1e6f5995bec3b1d8474d48ce3a139553a8e1 Borislav Stanimirov b.stanimirov@abv.bg 2024-07-10T14:40:53+03:00 GitHub noreply@github.com 2024-07-10T14:40:53+03:00 msvc : silence codecvt c++17 deprecation warnings (#8395)
a8be1e6f5995bec3b1d8474d48ce3a139553a8e1 e4dd31ff892627665d3c53c5d1a03c0d282d9d45 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-07-10T13:38:58+02:00 GitHub noreply@github.com 2024-07-10T14:38:58+03:00 llama : add assert about missing llama_encode() call (#8400)
e4dd31ff892627665d3c53c5d1a03c0d282d9d45 8f0fad42b9589f9b11362c74ea5338c51e4c7e61 RunningLeon maningsheng@sensetime.com 2024-07-10T19:26:40+08:00 GitHub noreply@github.com 2024-07-10T14:26:40+03:00 py : fix converter for internlm2 (#8321)
8f0fad42b9589f9b11362c74ea5338c51e4c7e61 a59f8fdc85e1119d470d8766e29617962549d993 laik laik.lj@me.com 2024-07-10T19:19:10+08:00 GitHub noreply@github.com 2024-07-10T14:19:10+03:00 py : fix extra space in convert_hf_to_gguf.py (#8407)
a59f8fdc85e1119d470d8766e29617962549d993 fd560fe680c72fd0a0af2bc8881add20ad919071 Clint Herron hanclinto@gmail.com 2024-07-09T18:26:40-04:00 GitHub noreply@github.com 2024-07-09T18:26:40-04:00 Server: Enable setting default sampling parameters via command-line (#8402)
fd560fe680c72fd0a0af2bc8881add20ad919071 e500d6135ac42c4a23baf6a9a1a934dc023e5c7d Andy Salerno andysalerno@gmail.com 2024-07-09T11:58:44-07:00 GitHub noreply@github.com 2024-07-09T14:58:44-04:00 Update README.md to fix broken link to docs (#8399)
e500d6135ac42c4a23baf6a9a1a934dc023e5c7d a03e8dd99d3954e7547137936c5a2a3b348bdb7f Clint Herron hanclinto@gmail.com 2024-07-09T11:54:43-04:00 GitHub noreply@github.com 2024-07-09T11:54:43-04:00 Deprecation warning to assist with migration to new binary names (#8283)
a03e8dd99d3954e7547137936c5a2a3b348bdb7f 5b0b8d8cfb5ddf2118f686ba6c30fab3f71b384b Johannes Gäßler johannesg@5d6.de 2024-07-09T17:11:07+02:00 GitHub noreply@github.com 2024-07-09T17:11:07+02:00 make/cmake: LLAMA_NO_CCACHE -> GGML_NO_CCACHE (#8392)
5b0b8d8cfb5ddf2118f686ba6c30fab3f71b384b 9925ca4087a34ab973b07bf06c0b770cb586830b Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-07-09T15:03:15+01:00 GitHub noreply@github.com 2024-07-09T22:03:15+08:00 sycl : Reenabled mmvq path for the SYCL Nvidia Backend (#8372)
9925ca4087a34ab973b07bf06c0b770cb586830b 9beb2dda038e2107a39a95def94a4cf971e048ea Borislav Stanimirov b.stanimirov@abv.bg 2024-07-09T11:38:00+03:00 GitHub noreply@github.com 2024-07-09T11:38:00+03:00 cmake : allow external ggml (#8370)
9beb2dda038e2107a39a95def94a4cf971e048ea 7d0e23d72ef4540d0d4409cb63ae682c17d53926 daghanerdonmez 44506702+daghanerdonmez@users.noreply.github.com 2024-07-09T09:16:00+03:00 GitHub noreply@github.com 2024-07-09T09:16:00+03:00 readme : fix typo [no ci] (#8389)
7d0e23d72ef4540d0d4409cb63ae682c17d53926 7fdb6f73e35605c8dbc39e9f19cd9ed84dbc87f2 compilade git@compilade.net 2024-07-09T01:04:49-04:00 GitHub noreply@github.com 2024-07-09T01:04:49-04:00 gguf-py : do not use internal numpy types (#7472)
7fdb6f73e35605c8dbc39e9f19cd9ed84dbc87f2 a130eccef42b75a84da270411cefeed45c153e30 Georgi Gerganov ggerganov@gmail.com 2024-07-09T01:36:38+03:00 GitHub noreply@github.com 2024-07-08T15:36:38-07:00 flake.lock: Update (#8342)
a130eccef42b75a84da270411cefeed45c153e30 c4dd11d1d3903e1922c06242e189f6310fc4d8c3 Alberto Cabrera Pérez alberto.cabrera@codeplay.com 2024-07-08T21:35:17+01:00 GitHub noreply@github.com 2024-07-08T22:35:17+02:00 labeler : updated sycl to match docs and code refactor (#8373)
c4dd11d1d3903e1922c06242e189f6310fc4d8c3 2ec846d558f6385ea647f7b8e665eb249c1ebce7 b4b4o zwbao@foxmail.com 2024-07-08T22:19:24+08:00 GitHub noreply@github.com 2024-07-08T17:19:24+03:00 readme : fix web link error [no ci] (#8347)
2ec846d558f6385ea647f7b8e665eb249c1ebce7 3f2d538b817112ad8429341c7e8657dcd660f4d3 Alberto Cabrera Pérez alberto.cabrera@intel.com 2024-07-08T14:22:41+01:00 GitHub noreply@github.com 2024-07-08T14:22:41+01:00 sycl : fix powf call in device code (#8368)
3f2d538b817112ad8429341c7e8657dcd660f4d3 2ee44c9a1865a928ccbbc16a2d7841d7513f31c1 Georgi Gerganov ggerganov@gmail.com 2024-07-08T13:51:31+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-08T13:51:31+03:00 scripts : fix sync for sycl
2ee44c9a1865a928ccbbc16a2d7841d7513f31c1 6847d54c4f72f8bf6767b6feae7a95394654335e Georgi Gerganov ggerganov@gmail.com 2024-07-08T10:39:50+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-08T12:23:00+03:00 sync : ggml
6847d54c4f72f8bf6767b6feae7a95394654335e fde13b3bb9f569f07fd8af74696ee48a43d05131 Georgi Gerganov ggerganov@gmail.com 2024-07-08T10:39:36+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-08T12:23:00+03:00 tests : fix whitespace (#0)
fde13b3bb9f569f07fd8af74696ee48a43d05131 470939d483d1c89b7292f78bac1fd27c42c171ce John Balis phobossystems@gmail.com 2024-07-02T11:09:52-05:00 Georgi Gerganov ggerganov@gmail.com 2024-07-08T12:23:00+03:00 feat: cuda implementation for `ggml_conv_transpose_1d` (ggml/854)
470939d483d1c89b7292f78bac1fd27c42c171ce 6f0dbf6ab087bcd286fb78560099ca0458316735 Kevin Wang kevmo314@gmail.com 2024-07-08T03:26:53-04:00 GitHub noreply@github.com 2024-07-08T10:26:53+03:00 common : preallocate sampling token data vector (#8363)
6f0dbf6ab087bcd286fb78560099ca0458316735 ffd00797d81ef7db1528b9e10adbdc333ade6495 Georgi Gerganov ggerganov@gmail.com 2024-07-08T09:34:35+03:00 GitHub noreply@github.com 2024-07-08T09:34:35+03:00 infill : assert prefix/suffix tokens + remove old space logic (#8351)
ffd00797d81ef7db1528b9e10adbdc333ade6495 04ce3a8b19256a155aea4d14eaa87edf274c93c3 Kevin Wang kevmo314@gmail.com 2024-07-08T02:31:55-04:00 GitHub noreply@github.com 2024-07-08T09:31:55+03:00 common : avoid unnecessary logits fetch (#8358)
04ce3a8b19256a155aea4d14eaa87edf274c93c3 3fd62a6b1c9ca7b7c0093e984cc9c133c6f2726d toyer 2042519524@qq.com 2024-07-08T13:57:19+08:00 GitHub noreply@github.com 2024-07-08T08:57:19+03:00 readme : add supported glm models (#8360)
3fd62a6b1c9ca7b7c0093e984cc9c133c6f2726d a8db2a9ce64cd4417f6a312ab61858f17f0f8584 compilade git@compilade.net 2024-07-07T15:04:39-04:00 GitHub noreply@github.com 2024-07-07T15:04:39-04:00 py : type-check all Python scripts with Pyright (#8341)
a8db2a9ce64cd4417f6a312ab61858f17f0f8584 4090ea5501c702cd858095c394ba068919c56cc8 Denis Spasyuk 34203011+dspasyuk@users.noreply.github.com 2024-07-07T09:08:28-06:00 GitHub noreply@github.com 2024-07-07T17:08:28+02:00 Update llama-cli documentation (#8315)
4090ea5501c702cd858095c394ba068919c56cc8 f1948f1e108157d5e7eb60fc8f24a10412e5d4ff Alex Tuddenham 61622354+AlexsCode@users.noreply.github.com 2024-07-07T15:59:14+01:00 GitHub noreply@github.com 2024-07-07T17:59:14+03:00 ci : add checks for cmake,make and ctest in ci/run.sh (#8200)
f1948f1e108157d5e7eb60fc8f24a10412e5d4ff f7cab35ef9e66c57b4a416d09eb6c814e0ba4e4c Andy Tai andy-tai@users.noreply.github.com 2024-07-07T06:21:37-07:00 GitHub noreply@github.com 2024-07-07T16:21:37+03:00 readme : update bindings list (#8222)
f7cab35ef9e66c57b4a416d09eb6c814e0ba4e4c 905942abdba5ba0b28a1b0805e51e4f818c54bc9 Brian mofosyne@gmail.com 2024-07-07T22:58:43+10:00 GitHub noreply@github.com 2024-07-07T22:58:43+10:00 gguf-hash: model wide and per tensor hashing using xxhash and sha1 (#8048)
905942abdba5ba0b28a1b0805e51e4f818c54bc9 b5040086d436e7345e4fa33a5b9558060c75603f toyer 2042519524@qq.com 2024-07-07T20:52:10+08:00 GitHub noreply@github.com 2024-07-07T15:52:10+03:00 llama : support glm3 and glm4 (#8031)
b5040086d436e7345e4fa33a5b9558060c75603f d39130a3985ce0747d7229a6b7ebebb6376b5abf Georgi Gerganov ggerganov@gmail.com 2024-07-07T14:59:02+03:00 GitHub noreply@github.com 2024-07-07T14:59:02+03:00 llama : fix n_rot default (#8348)
d39130a3985ce0747d7229a6b7ebebb6376b5abf b81ba1f96b06c691020429a2d2dfcbada899ca86 compilade git@compilade.net 2024-07-07T07:23:38-04:00 GitHub noreply@github.com 2024-07-07T14:23:38+03:00 py : use cpu-only torch in requirements.txt (#8335)
b81ba1f96b06c691020429a2d2dfcbada899ca86 210eb9ed0ac3979a93e37568d96447ec3e95ad05 standby24x7 standby24x7@gmail.com 2024-07-07T19:38:02+09:00 GitHub noreply@github.com 2024-07-07T13:38:02+03:00 finetune: Rename command name in README.md (#8343)
210eb9ed0ac3979a93e37568d96447ec3e95ad05 cb4d86c4d723af87d3d7e3177e9485f200391384 standby24x7 standby24x7@gmail.com 2024-07-07T19:37:47+09:00 GitHub noreply@github.com 2024-07-07T13:37:47+03:00 finetune: Rename an old command name in finetune.sh (#8344)
cb4d86c4d723af87d3d7e3177e9485f200391384 86e7299ef5dff0f388922dc6fcbce009e99d8005 Bjarke Viksøe 164612031+bviksoe@users.noreply.github.com 2024-07-07T11:10:38+02:00 GitHub noreply@github.com 2024-07-07T11:10:38+02:00 server: Retrieve prompt template in /props (#8337)
86e7299ef5dff0f388922dc6fcbce009e99d8005 60d83a0149849e9217e4b8ae26e277a41aea906e Derrick T. Woolworth dwoolworth@gmail.com 2024-07-06T15:32:04-05:00 GitHub noreply@github.com 2024-07-06T22:32:04+02:00 added support for Authorization Bearer tokens when downloading model (#8307)
60d83a0149849e9217e4b8ae26e277a41aea906e 87e25a1d1bd26eb06d1cab9e2ee4e14a7a0be33c Xuan Son Nguyen thichthat@gmail.com 2024-07-06T19:01:23+02:00 GitHub noreply@github.com 2024-07-06T19:01:23+02:00 update main readme (#8333)
87e25a1d1bd26eb06d1cab9e2ee4e14a7a0be33c 213701b51a17175d0d326b566efc03f30ec7fbe6 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-06T09:22:16+02:00 GitHub noreply@github.com 2024-07-06T10:22:16+03:00 llama : add early return for empty range (#8327)
213701b51a17175d0d326b566efc03f30ec7fbe6 be20e7f49d9e5c6d9e8d9b4871eeba3df7a1639d jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-07-05T19:01:35+02:00 GitHub noreply@github.com 2024-07-05T19:01:35+02:00 Detokenizer fixes (#8039)
be20e7f49d9e5c6d9e8d9b4871eeba3df7a1639d 7ed03b8974269b6c48e55c4245d12fb3264a6cf5 Xuan Son Nguyen thichthat@gmail.com 2024-07-05T18:08:32+02:00 GitHub noreply@github.com 2024-07-05T18:08:32+02:00 Reorganize documentation pages (#8325)
7ed03b8974269b6c48e55c4245d12fb3264a6cf5 1d894a790e62b10d066adcdd9c9feb559455b7d2 Georgi Gerganov ggerganov@gmail.com 2024-07-05T17:32:09+03:00 Georgi Gerganov ggerganov@gmail.com 2024-07-05T17:32:09+03:00 llama : fix compile warning (#8304)
1d894a790e62b10d066adcdd9c9feb559455b7d2 1f3e1b66e21310ed78b964f72f19766549633f0e Natsu chino@hotococoa.moe 2024-07-05T22:29:35+08:00 GitHub noreply@github.com 2024-07-05T17:29:35+03:00 cmake : add GGML_BUILD and GGML_SHARED macro definitions (#8281)
1f3e1b66e21310ed78b964f72f19766549633f0e 148ec970b62c3c5ae0a8bfdaad2fc237aaae350d Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-07-05T13:23:25+01:00 GitHub noreply@github.com 2024-07-05T13:23:25+01:00 Enabled more data types for oneMKL gemm_batch (#8236)
148ec970b62c3c5ae0a8bfdaad2fc237aaae350d 2cccbaa0086c62801f95405131a5b2faf3ba1de8 Georgi Gerganov ggerganov@gmail.com 2024-07-05T10:15:36+03:00 GitHub noreply@github.com 2024-07-05T10:15:36+03:00 convert : remove AWQ remnants (#8320)
2cccbaa0086c62801f95405131a5b2faf3ba1de8 8e558309dc149dc1f9fd159185b0b9071527ffb5 Georgi Gerganov ggerganov@gmail.com 2024-07-05T10:15:24+03:00 GitHub noreply@github.com 2024-07-05T10:15:24+03:00 llama : minor indentation during tensor loading (#8304)
8e558309dc149dc1f9fd159185b0b9071527ffb5 0a423800ffe4e5da3d83527ef3473da88cd78146 Johannes Gäßler johannesg@5d6.de 2024-07-05T09:06:31+02:00 GitHub noreply@github.com 2024-07-05T09:06:31+02:00 CUDA: MMQ support for iq4_nl, iq4_xs (#8278)
0a423800ffe4e5da3d83527ef3473da88cd78146 d12f781074b92589a72a36ffabb583933f7b9dc0 Daniele 57776841+daniandtheweb@users.noreply.github.com 2024-07-05T07:06:09Z GitHub noreply@github.com 2024-07-05T09:06:09+02:00 CUDA: revert part of the RDNA1 optimizations (#8309)
d12f781074b92589a72a36ffabb583933f7b9dc0 bcefa03bc01a41aace2e200ee8e77827d6d39b4f Douglas Hanley thesecretaryofwar@gmail.com 2024-07-05T02:05:56-05:00 GitHub noreply@github.com 2024-07-05T10:05:56+03:00 llama : streamline embeddings from "non-embedding" models (#8087)
bcefa03bc01a41aace2e200ee8e77827d6d39b4f 5a7447c5692c9e3dde1161e8e69edb76d3d34714 Johannes Gäßler johannesg@5d6.de 2024-07-05T09:05:34+02:00 GitHub noreply@github.com 2024-07-05T09:05:34+02:00 CUDA: fix MMQ stream-k rounding if ne00 % 128 != 0 (#8311)
5a7447c5692c9e3dde1161e8e69edb76d3d34714 61ecafa3905a02a299b7ae889b5578cfeb8d79df Pieter Ouwerkerk pieter.ouwerkerk@gmail.com 2024-07-05T02:58:41-04:00 GitHub noreply@github.com 2024-07-05T09:58:41+03:00 readme : fix minor typos [no ci] (#8314)
61ecafa3905a02a299b7ae889b5578cfeb8d79df aa5898dc53afcf77f16222cd719e10b29049f95a Daniel Bevenius daniel.bevenius@gmail.com 2024-07-05T08:14:24+02:00 GitHub noreply@github.com 2024-07-05T09:14:24+03:00 passkey : add short intro to README.md [no-ci] (#8317)
aa5898dc53afcf77f16222cd719e10b29049f95a 6c05752c507f51b88348f16d9e2c8df49f66c028 Georgi Gerganov ggerganov@gmail.com 2024-07-05T09:10:03+03:00 GitHub noreply@github.com 2024-07-05T09:10:03+03:00 llama : prefer n_ over num_ prefix (#8308)
6c05752c507f51b88348f16d9e2c8df49f66c028 a9554e20b66546b0549aebe2e1034bc8afe9d809 Georgi Gerganov ggerganov@gmail.com 2024-07-05T09:09:47+03:00 GitHub noreply@github.com 2024-07-05T09:09:47+03:00 contributing : update guidelines (#8316)
a9554e20b66546b0549aebe2e1034bc8afe9d809 e235b267a2539d043734ff340eff74107722eb57 luoyu-intel yu.luo@intel.com 2024-07-05T05:06:13Z GitHub noreply@github.com 2024-07-05T13:06:13+08:00 [SYCL] Fix WARP_SIZE=16 bug of Intel GPU (#8266)
e235b267a2539d043734ff340eff74107722eb57 f09b7cb609d80b8031803f89255991dc8b35db69 Georgi Gerganov ggerganov@gmail.com 2024-07-05T07:53:33+03:00 GitHub noreply@github.com 2024-07-05T07:53:33+03:00 py : switch to snake_case (#8305)
f09b7cb609d80b8031803f89255991dc8b35db69 a38b884c6c4b0c256583acfaaabdf556c62fabea Neo Zhang Jianyu jianyu.zhang@intel.com 2024-07-05T10:32:29+08:00 GitHub noreply@github.com 2024-07-05T10:32:29+08:00 rm get_work_group_size() by local cache for performance (#8286)
a38b884c6c4b0c256583acfaaabdf556c62fabea d7fd29fff16456ce9c3a23fd2d09a66256b05aff Xuan Son Nguyen thichthat@gmail.com 2024-07-04T20:55:03+02:00 GitHub noreply@github.com 2024-07-04T20:55:03+02:00 cli: add EOT when user hit Ctrl+C (#8296)
d7fd29fff16456ce9c3a23fd2d09a66256b05aff 6f63d646c1a06a6e09f721009a2676864ae04e31 Icecream95 the.real.icecream95@gmail.com 2024-07-05T05:14:21+12:00 GitHub noreply@github.com 2024-07-04T20:14:21+03:00 llama : add OpenELM support (#7359)
6f63d646c1a06a6e09f721009a2676864ae04e31 51d2ebadbbf365b894f3888361df42dbacb12b7a Daniel Bevenius daniel.bevenius@gmail.com 2024-07-04T18:38:58+02:00 GitHub noreply@github.com 2024-07-04T19:38:58+03:00 tokenize : add --show-count (token) option (#8299)
51d2ebadbbf365b894f3888361df42dbacb12b7a 1e920018d38aee270a044cc48eaefe7c64cb8750 ditsuke ditsuke@protonmail.com 2024-07-04T20:54:35+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z build: Export hf-to-gguf as snakecase
1e920018d38aee270a044cc48eaefe7c64cb8750 01a5f06550a866bd63717635e5e7e1ff4203b873 ditsuke ditsuke@protonmail.com 2024-07-03T01:02:56+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z doc: Add context for why we add an explicit pytorch source
01a5f06550a866bd63717635e5e7e1ff4203b873 07786a61a2097306ee496f565f78796f1aa205e6 ditsuke ditsuke@protonmail.com 2024-07-02T15:48:13+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z chore: Remove rebase artifacts
07786a61a2097306ee496f565f78796f1aa205e6 de14e2ea2b542386dcb800226eb360be76f433fd ditsuke ditsuke@protonmail.com 2024-07-02T15:35:43+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z chore: Fixup requirements and build
de14e2ea2b542386dcb800226eb360be76f433fd 821922916f95cc3853fc7f58ea2f1e15a0ffa669 ditsuke ditsuke@protonmail.com 2024-07-02T15:18:13+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z chore: ignore all __pychache__
821922916f95cc3853fc7f58ea2f1e15a0ffa669 b1c3f26e5e882fa46d7a6704d893b31a025e9000 ditsuke ditsuke@protonmail.com 2024-03-10T23:21:46+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z fix: Update script paths in CI scripts
b1c3f26e5e882fa46d7a6704d893b31a025e9000 b0a46993dfbf8b8127598f319d4dcfdd83824ba8 ditsuke ditsuke@protonmail.com 2024-02-29T01:47:15+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z fix: Actually include scripts in build
b0a46993dfbf8b8127598f319d4dcfdd83824ba8 807b0c49ff7071094f97ebc3a0a8e2b9e274f503 ditsuke ditsuke@protonmail.com 2024-02-27T12:01:02+05:30 Someone else@someonex.net 2024-07-04T15:39:13Z build(python): Package scripts with pip-0517 compliance
807b0c49ff7071094f97ebc3a0a8e2b9e274f503 f8c4c0738d72d2162736edd72dd5db8b269adca1 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-07-04T15:46:11+02:00 GitHub noreply@github.com 2024-07-04T15:46:11+02:00 Inference support for T5 and FLAN-T5 model families (#5763)
f8c4c0738d72d2162736edd72dd5db8b269adca1 402d6feffa0572d1c7a957901b6d1702bd188484 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-04T12:53:42+02:00 GitHub noreply@github.com 2024-07-04T13:53:42+03:00 tests : add _CRT_SECURE_NO_WARNINGS for WIN32 (#8231)
402d6feffa0572d1c7a957901b6d1702bd188484 20fc3804bfb727074bc270b6eacb60af8d0bf7d4 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-04T12:50:57+02:00 GitHub noreply@github.com 2024-07-04T13:50:57+03:00 llama : suppress unref var in Windows MSVC (#8150)
20fc3804bfb727074bc270b6eacb60af8d0bf7d4 f619024764e72261f14d7c31d892b8fb976603b4 Georgi Gerganov ggerganov@gmail.com 2024-07-04T10:41:03+03:00 GitHub noreply@github.com 2024-07-04T10:41:03+03:00 convert : fix gemma v1 tokenizer convert (#8248)
f619024764e72261f14d7c31d892b8fb976603b4 d23287f122c34ebef368742116d53a0ccb2041ee AidanBeltonS aidan.belton@codeplay.com 2024-07-04T02:07:19+01:00 GitHub noreply@github.com 2024-07-04T09:07:19+08:00 [SYCL] Remove unneeded semicolons (#8280)
d23287f122c34ebef368742116d53a0ccb2041ee 5f2d4e60e202aabee10051e6615bb821e51787be Daniele 57776841+daniandtheweb@users.noreply.github.com 2024-07-03T23:02:58Z GitHub noreply@github.com 2024-07-04T01:02:58+02:00 Define and optimize RDNA1 (#8085)
5f2d4e60e202aabee10051e6615bb821e51787be 916248af1f3c16abd7408de848e025da095c621c slaren slarengh@gmail.com 2024-07-03T19:33:31+02:00 GitHub noreply@github.com 2024-07-03T20:33:31+03:00 ppl : fix n_seq_max for perplexity (#8277)
916248af1f3c16abd7408de848e025da095c621c f8d6a23804f3798ff2869da68c1223b618df09ec Xuan Son Nguyen thichthat@gmail.com 2024-07-03T16:01:54+02:00 GitHub noreply@github.com 2024-07-03T16:01:54+02:00 fix phi 3 conversion (#8262)
f8d6a23804f3798ff2869da68c1223b618df09ec fadde6713506d9e6c124f5680ab8c7abebe31837 Judd foldl@users.noreply.github.com 2024-07-03T20:40:16+08:00 GitHub noreply@github.com 2024-07-03T14:40:16+02:00 fix typo (#8267)
fadde6713506d9e6c124f5680ab8c7abebe31837 a27152b602b369e76f85b7cb7b872a321b7218f7 AidanBeltonS aidan.belton@codeplay.com 2024-07-03T02:55:34+01:00 GitHub noreply@github.com 2024-07-03T09:55:34+08:00 Dequant improvements rebase (#8255)
a27152b602b369e76f85b7cb7b872a321b7218f7 3e2618bc7bf9e9fbf58c32cc3c8dd7d5df1de27e MistApproach 98988043+MistApproach@users.noreply.github.com 2024-07-02T22:56:46+02:00 GitHub noreply@github.com 2024-07-02T22:56:46+02:00 fix: add missing short command line argument -mli for multiline-input (#8261)
3e2618bc7bf9e9fbf58c32cc3c8dd7d5df1de27e 07a3fc0608a68c0c93a5fbfa9c58f4c9ec64cb81 Clint Herron hanclinto@gmail.com 2024-07-02T13:19:56-04:00 GitHub noreply@github.com 2024-07-02T13:19:56-04:00 Adding step to `clean` target to remove legacy binary names to reduce upgrade / migration confusion arising from #7809. (#8257)
07a3fc0608a68c0c93a5fbfa9c58f4c9ec64cb81 968967376dc2c018d29f897c4883d335bbf384fb Clint Herron hanclinto@gmail.com 2024-07-02T12:18:10-04:00 GitHub noreply@github.com 2024-07-02T12:18:10-04:00 Removes multiple newlines at the end of files that is breaking the editorconfig step of CI. (#8258)
968967376dc2c018d29f897c4883d335bbf384fb 023b8807e10bc3ade24a255f01c1ad2a01bb4228 Faisal Zaghloul faisal.zaghloul@gmail.com 2024-07-02T10:36:00-04:00 GitHub noreply@github.com 2024-07-02T16:36:00+02:00 Add `JAIS` model(s) (#8118)
023b8807e10bc3ade24a255f01c1ad2a01bb4228 0e0590adab9f367b15ae2bf090a6d24f9df47ff1 Daniel Bevenius daniel.bevenius@gmail.com 2024-07-02T08:40:49+02:00 GitHub noreply@github.com 2024-07-02T09:40:49+03:00 convert-hf : print output file name when completed (#8181)
0e0590adab9f367b15ae2bf090a6d24f9df47ff1 a9f3b102157ba992cfe058909b7f6e1906d2d647 slaren slarengh@gmail.com 2024-07-02T08:39:38+02:00 GitHub noreply@github.com 2024-07-02T09:39:38+03:00 cuda : update supports_op for matrix multiplication (#8245)
a9f3b102157ba992cfe058909b7f6e1906d2d647 d08c20eddedb24515a3212e2de66bdff41a26b8c luoyu-intel yu.luo@intel.com 2024-07-02T04:50:07Z GitHub noreply@github.com 2024-07-02T12:50:07+08:00 [SYCL] Fix win build conflict of math library (#8230)
d08c20eddedb24515a3212e2de66bdff41a26b8c 5fac350b9cc49d0446fc291b9c4ad53666c77591 luoyu-intel yu.luo@intel.com 2024-07-02T02:16:00Z GitHub noreply@github.com 2024-07-02T10:16:00+08:00 [SYCL] Fix the sub group size of Intel (#8106)
5fac350b9cc49d0446fc291b9c4ad53666c77591 cb5fad4c6c2cbef92e9b8b63449e1cb7664e4846 Xuan Son Nguyen thichthat@gmail.com 2024-07-02T01:07:23+02:00 GitHub noreply@github.com 2024-07-02T01:07:23+02:00 Fix gemma2 tokenizer convert (#8244)
cb5fad4c6c2cbef92e9b8b63449e1cb7664e4846 dae57a1ebc1c9bd5693ab999e19d77c5506ae559 Johannes Gäßler johannesg@5d6.de 2024-07-01T20:39:06+02:00 GitHub noreply@github.com 2024-07-01T20:39:06+02:00 CUDA: refactor and optimize IQ MMVQ (#8215)
dae57a1ebc1c9bd5693ab999e19d77c5506ae559 49122a873f54615626d1b49a2a39013ed4be98d5 Mateusz Charytoniuk mateusz.charytoniuk@protonmail.com 2024-07-01T19:13:22+02:00 GitHub noreply@github.com 2024-07-01T20:13:22+03:00 readme: add Paddler to the list of projects (#8239)
49122a873f54615626d1b49a2a39013ed4be98d5 0ddeff10230b88f1fa9866bbe5fe0d71ba2323a0 Xuan Son Nguyen thichthat@gmail.com 2024-07-01T18:48:34+02:00 GitHub noreply@github.com 2024-07-01T18:48:34+02:00 gemma2: add sliding window mask (#8227)
0ddeff10230b88f1fa9866bbe5fe0d71ba2323a0 3840b6f593751a0ba636bfda73b630cd6c29d7b5 Roni sulpher@gmx.net 2024-07-01T14:48:16+02:00 GitHub noreply@github.com 2024-07-01T15:48:16+03:00 readme : update tool list (#8209)
3840b6f593751a0ba636bfda73b630cd6c29d7b5 257f8e41e24b5bbfc27d9e907189a3e0cdb650d4 Michael Francis edude03@gmail.com 2024-07-01T07:47:04-04:00 GitHub noreply@github.com 2024-07-01T14:47:04+03:00 nix : enable curl (#8043)
257f8e41e24b5bbfc27d9e907189a3e0cdb650d4 694c59cb42d1ebd6a7d912ca65d3d7363e0f14c9 Georgi Gerganov ggerganov@gmail.com 2024-07-01T14:46:18+03:00 GitHub noreply@github.com 2024-07-01T14:46:18+03:00 nix : remove OpenCL remnants (#8235)
694c59cb42d1ebd6a7d912ca65d3d7363e0f14c9 197fe6c1d7bec6718ce901f0141b2725240f298c iacore 74560659+iacore@users.noreply.github.com 2024-07-01T11:40:58Z GitHub noreply@github.com 2024-07-01T13:40:58+02:00 Document BERT support. (#8205)
197fe6c1d7bec6718ce901f0141b2725240f298c d0a7145ba99ed3a8bc3145aa785b5c86ffe65020 zhentaoyu zhentao.yu@intel.com 2024-07-01T19:39:06+08:00 GitHub noreply@github.com 2024-07-01T19:39:06+08:00 [SYCL] Update SYCL-Rope op and Refactor (#8157)
d0a7145ba99ed3a8bc3145aa785b5c86ffe65020 9ef07800622e4c371605f9419864d15667c3558f Georgi Gerganov ggerganov@gmail.com 2024-07-01T02:09:34+03:00 GitHub noreply@github.com 2024-06-30T16:09:34-07:00 flake.lock: Update (#8218)
9ef07800622e4c371605f9419864d15667c3558f 1c5eba6f8e628fb0a98afb27d8aaeb3b0e136451 Xuan Son Nguyen thichthat@gmail.com 2024-06-30T20:27:13+02:00 GitHub noreply@github.com 2024-06-30T20:27:13+02:00 Fix new line issue with chat template, disable template when in-prefix/suffix is set (#8203)
1c5eba6f8e628fb0a98afb27d8aaeb3b0e136451 72272b83a3878e91251218c981b4c6ec16c33912 Andrei abetlen@gmail.com 2024-06-29T20:44:08-07:00 GitHub noreply@github.com 2024-06-29T23:44:08-04:00 llama: Add attention and final logit soft-capping, update scaling factor to Gemma2 (#8197)
72272b83a3878e91251218c981b4c6ec16c33912 8748d8ac6f172b99826ab18f01d9a3a165987d54 Xuan Son Nguyen thichthat@gmail.com 2024-06-29T00:14:20+02:00 GitHub noreply@github.com 2024-06-29T00:14:20+02:00 fix code typo in llama-cli (#8198)
8748d8ac6f172b99826ab18f01d9a3a165987d54 26a39bbd6b0bbd66118bb68569f0276d7fe7df6c Olivier Chafik ochafik@users.noreply.github.com 2024-06-28T18:02:05+01:00 GitHub noreply@github.com 2024-06-28T18:02:05+01:00 json: attempt to skip slow tests when running under emulator (#8189)
26a39bbd6b0bbd66118bb68569f0276d7fe7df6c 38373cfbab5397cc2ab5c3694a3dee12a9e58f45 Xuan Son Nguyen thichthat@gmail.com 2024-06-28T15:11:44+02:00 GitHub noreply@github.com 2024-06-28T15:11:44+02:00 Add MiniCPM, Deepseek V2 chat template + clean up `llama_chat_apply_template_internal` (#8172)
38373cfbab5397cc2ab5c3694a3dee12a9e58f45 b851b3fba0a1b06a1129189bac300e07dd1648c8 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-06-28T12:53:43+02:00 GitHub noreply@github.com 2024-06-28T12:53:43+02:00 Add SPM infill support (#8016)
b851b3fba0a1b06a1129189bac300e07dd1648c8 139cc621e90b4f61830515c3c124cf35b3d7a6dc slaren slarengh@gmail.com 2024-06-28T12:37:45+02:00 GitHub noreply@github.com 2024-06-28T12:37:45+02:00 cmake : allow user to override default options (#8178)
139cc621e90b4f61830515c3c124cf35b3d7a6dc e57dc62057d41211ac018056c19c02cd544694df Olivier Chafik ochafik@users.noreply.github.com 2024-06-28T09:26:45+01:00 GitHub noreply@github.com 2024-06-28T09:26:45+01:00 `json`: restore default additionalProperties to false, fix some pattern escapes (#8180)
e57dc62057d41211ac018056c19c02cd544694df a27aa50ab7e07fe46aae619076b6e31d5663e914 pculliton phillipculliton@gmail.com 2024-06-28T00:00:43-04:00 GitHub noreply@github.com 2024-06-27T21:00:43-07:00 llama: Add support for Gemma2ForCausalLM (#8156)
a27aa50ab7e07fe46aae619076b6e31d5663e914 cb0b06a8a613f7a2ccb7253b2a3c00fdd397ba1c Xuan Son Nguyen thichthat@gmail.com 2024-06-28T02:19:11+02:00 GitHub noreply@github.com 2024-06-28T02:19:11+02:00 Add missing items in makefile (#8177)
cb0b06a8a613f7a2ccb7253b2a3c00fdd397ba1c 558f44bf83d78242d4e5c4ab98d0be9125cb9780 Olivier Chafik ochafik@users.noreply.github.com 2024-06-27T22:08:42+01:00 GitHub noreply@github.com 2024-06-27T22:08:42+01:00 `json`: update grammars/README w/ examples & note about additionalProperties (#8132)
558f44bf83d78242d4e5c4ab98d0be9125cb9780 8172ee9da9921ca53d698c7438c2d792b3f3f2c8 loonerin 132926317+loonerin@users.noreply.github.com 2024-06-27T15:01:23-04:00 GitHub noreply@github.com 2024-06-27T21:01:23+02:00 CI: fix release build (Ubuntu+Mac) (#8170)
8172ee9da9921ca53d698c7438c2d792b3f3f2c8 16791b8f0b4526aafbf5d0e5bbbd2e99c2253418 slaren slarengh@gmail.com 2024-06-27T20:04:39+02:00 GitHub noreply@github.com 2024-06-27T20:04:39+02:00 cmake : fix deprecated option names not working (#8171)
16791b8f0b4526aafbf5d0e5bbbd2e99c2253418 ab3679112d4c49a215a3d31550a7720b202e9015 Xuan Son Nguyen thichthat@gmail.com 2024-06-27T18:14:19+02:00 GitHub noreply@github.com 2024-06-27T18:14:19+02:00 Add chatml fallback for cpp `llama_chat_apply_template` (#8160)
ab3679112d4c49a215a3d31550a7720b202e9015 97877eb10bd8e7f8023420b5b5300bcbdadd62dc Georgi Gerganov ggerganov@gmail.com 2024-06-27T18:37:29+03:00 GitHub noreply@github.com 2024-06-27T08:37:29-07:00 flake.lock: Update (#8071)
97877eb10bd8e7f8023420b5b5300bcbdadd62dc 387952651a8fc493f8c85ea4c9774bd4a5694f87 jukofyork 69222624+jukofyork@users.noreply.github.com 2024-06-27T15:48:07+01:00 GitHub noreply@github.com 2024-06-27T16:48:07+02:00 Control vector loading fixes (#8137)
387952651a8fc493f8c85ea4c9774bd4a5694f87 6030c61281c8a7eb94eceb7396a608fac8b71555 Raj Hammeer Singh Hada hammeerraj@gmail.com 2024-06-27T20:09:29+05:30 GitHub noreply@github.com 2024-06-27T16:39:29+02:00 Delete examples/llama.android/llama/CMakeLists.txt (#8165)
6030c61281c8a7eb94eceb7396a608fac8b71555 85a267daaa1c6f8fd69160445bcb88717031d10c Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-06-27T16:27:41+02:00 GitHub noreply@github.com 2024-06-27T16:27:41+02:00 Add Qwen2MoE 57B-A14B model identifier (#8158)
85a267daaa1c6f8fd69160445bcb88717031d10c f675b20a3b7f878bf3be766b9a737e2c8321ff0d Johannes Gäßler johannesg@5d6.de 2024-06-27T16:26:05+02:00 GitHub noreply@github.com 2024-06-27T16:26:05+02:00 CUDA: fix MMQ stream-k for --split-mode row (#8167)
f675b20a3b7f878bf3be766b9a737e2c8321ff0d 911e35bb8bb2fd1c7d3f40f27e96ff432eae7e14 kustaaya 58045274+kustaaya@users.noreply.github.com 2024-06-27T11:58:54+03:00 GitHub noreply@github.com 2024-06-27T10:58:54+02:00 Added support for Viking pre-tokenizer (#8135)
911e35bb8bb2fd1c7d3f40f27e96ff432eae7e14 ac146628e47451c531a3c7e62e6a973a2bb467a0 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-06-27T09:46:41+02:00 GitHub noreply@github.com 2024-06-27T10:46:41+03:00 llama : fix CodeLlama FIM token checks (#8144)
ac146628e47451c531a3c7e62e6a973a2bb467a0 9b31a40c6ddabe552875b811d7127aa039ca9703 Raj Hammeer Singh Hada hammeerraj@gmail.com 2024-06-27T07:27:57+05:30 GitHub noreply@github.com 2024-06-27T03:57:57+02:00 Fix llama-android.cpp for error - "common/common.h not found" (#8145)
9b31a40c6ddabe552875b811d7127aa039ca9703 c70d117c37cc7876e775d1e2722208a50c52edb3 Daniel Bevenius daniel.bevenius@gmail.com 2024-06-27T01:50:09+02:00 GitHub noreply@github.com 2024-06-27T01:50:09+02:00 clip : suppress unused variable warnings (#8105)
c70d117c37cc7876e775d1e2722208a50c52edb3 ae5d0f4b899ff2842bfca561370c945ad8d4368b Georgi Gerganov ggerganov@gmail.com 2024-06-26T23:25:22+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-26T23:25:22+03:00 scripts : fix filename sync
ae5d0f4b899ff2842bfca561370c945ad8d4368b 31ec3993f6e050322a249c07af79dbde66ea6ddc slaren slarengh@gmail.com 2024-06-26T21:59:28+02:00 GitHub noreply@github.com 2024-06-26T21:59:28+02:00 ci : publish new docker images only when the files change (#8142)
31ec3993f6e050322a249c07af79dbde66ea6ddc c7ab7b612cbdce04499575e713076a026af4b9c5 slaren slarengh@gmail.com 2024-06-26T21:34:14+02:00 GitHub noreply@github.com 2024-06-26T21:34:14+02:00 ggml : add GGML_CUDA_USE_GRAPHS option, restore GGML_CUDA_FORCE_CUBLAS (cmake) (#8140)
c7ab7b612cbdce04499575e713076a026af4b9c5 f2d48fffde76d959fdb0da37316bdc09e5518eb1 slaren slarengh@gmail.com 2024-06-26T20:20:22+02:00 GitHub noreply@github.com 2024-06-26T21:20:22+03:00 make : fix missing -O3 (#8143)
f2d48fffde76d959fdb0da37316bdc09e5518eb1 4713bf3093d58a3e12368ab2ab5fc3630f27803e Georgi Gerganov ggerganov@gmail.com 2024-06-26T19:39:19+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-26T19:39:19+03:00 sync : ggml
4713bf3093d58a3e12368ab2ab5fc3630f27803e 0e814dfc42b4b57ad19598d239557b6a977ca16c Georgi Gerganov ggerganov@gmail.com 2024-06-26T19:36:44+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-26T19:36:44+03:00 authors : regen
0e814dfc42b4b57ad19598d239557b6a977ca16c a95631ee97bb24861af6bdeec380270459631e8e Georgi Gerganov ggerganov@gmail.com 2024-06-26T19:32:07+03:00 GitHub noreply@github.com 2024-06-26T19:32:07+03:00 devops : remove clblast + LLAMA_CUDA -> GGML_CUDA (#8139)
a95631ee97bb24861af6bdeec380270459631e8e f3f65429c44bb195a9195bfdc19a30a79709db7b Georgi Gerganov ggerganov@gmail.com 2024-06-26T19:26:13+03:00 GitHub noreply@github.com 2024-06-26T19:26:13+03:00 readme : update API notes
f3f65429c44bb195a9195bfdc19a30a79709db7b 88540445615e77a0177fcca43aaa8e9d8eea6864 Georgi Gerganov ggerganov@gmail.com 2024-06-26T18:33:02+03:00 GitHub noreply@github.com 2024-06-26T18:33:02+03:00 llama : reorganize source code + improve CMake (#8006)
88540445615e77a0177fcca43aaa8e9d8eea6864 c8771ab5f89387cdd7d9a8a69280dac46b45e02f Isaac McFadyen isaac@imcf.me 2024-06-26T02:29:28-04:00 GitHub noreply@github.com 2024-06-26T08:29:28+02:00 Clarify default MMQ for CUDA and LLAMA_CUDA_FORCE_MMQ flag (#8115)
c8771ab5f89387cdd7d9a8a69280dac46b45e02f 494165f3b6c4cbcd793123cb57fb3e1f5477f1db Johannes Gäßler johannesg@5d6.de 2024-06-26T08:28:02+02:00 GitHub noreply@github.com 2024-06-26T08:28:02+02:00 CUDA: fix misaligned shared memory read (#8123)
494165f3b6c4cbcd793123cb57fb3e1f5477f1db 9b2f16f8055265c67e074025350736adc1ea0666 Eddie-Wang wangjinheng1120@163.com 2024-06-26T14:27:46+08:00 GitHub noreply@github.com 2024-06-26T09:27:46+03:00 llama : extend llm_build_ffn() to support _scale tensors (#8103)
9b2f16f8055265c67e074025350736adc1ea0666 6777c544bdd8c5d9de3220d6e2557957bbbf2a4f Olivier Chafik ochafik@users.noreply.github.com 2024-06-26T01:46:35+01:00 GitHub noreply@github.com 2024-06-26T01:46:35+01:00 `json`: better support for "type" unions (e.g. nullable arrays w/ typed items) (#7863)
6777c544bdd8c5d9de3220d6e2557957bbbf2a4f 163d50adaf8897d8b734d701ff332de6be63d484 Olivier Chafik ochafik@users.noreply.github.com 2024-06-26T01:45:58+01:00 GitHub noreply@github.com 2024-06-26T01:45:58+01:00 `json`: fix additionalProperties, allow space after enum/const (#7840)
163d50adaf8897d8b734d701ff332de6be63d484 6fcbf6823553efabe52ed83e3c2a3329aa3387d1 jukofyork 69222624+jukofyork@users.noreply.github.com 2024-06-25T21:47:40+01:00 GitHub noreply@github.com 2024-06-25T22:47:40+02:00 fixes #7999 (adds control vectors to all `build_XXX()` functions in `llama.cpp` [needs testing] (#8060)
6fcbf6823553efabe52ed83e3c2a3329aa3387d1 e6bf007744eb06336a231ef39cf08146dd16d2ce fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-06-25T21:14:35+02:00 GitHub noreply@github.com 2024-06-25T21:14:35+02:00 llama : implement Unigram tokenizer needed by T5 and FLAN-T5 model families (#5763)
e6bf007744eb06336a231ef39cf08146dd16d2ce 84631fe1504de40427dc4b4cdac92fa7ebf65955 Daniel Bevenius daniel.bevenius@gmail.com 2024-06-25T21:07:28+02:00 GitHub noreply@github.com 2024-06-25T15:07:28-04:00 llama : return nullptr from llama_grammar_init (#8093)
84631fe1504de40427dc4b4cdac92fa7ebf65955 dd047b476c8b904e0c25e5dbc5bee6ffde2f6e17 Olivier Chafik ochafik@users.noreply.github.com 2024-06-25T20:06:20+01:00 GitHub noreply@github.com 2024-06-25T20:06:20+01:00 `json`: support integer minimum, maximum, exclusiveMinimum, exclusiveMaximum (#7797)
dd047b476c8b904e0c25e5dbc5bee6ffde2f6e17 925c30956dd17723c3a25297bcd0a609aec60663 slaren slarengh@gmail.com 2024-06-25T19:20:06+02:00 GitHub noreply@github.com 2024-06-25T19:20:06+02:00 disable docker CI on pull requests (#8110)
925c30956dd17723c3a25297bcd0a609aec60663 c8ad35955ad2c68db172dcd0e857423ab128518d joecryptotoo 80373433+joecryptotoo@users.noreply.github.com 2024-06-25T08:13:27-07:00 GitHub noreply@github.com 2024-06-25T17:13:27+02:00 Add healthchecks to llama-server containers (#8081)
c8ad35955ad2c68db172dcd0e857423ab128518d 49c03c79cda17913b72260acdc8157b742cee41c Brian mofosyne@gmail.com 2024-06-25T22:03:25+10:00 GitHub noreply@github.com 2024-06-25T22:03:25+10:00 Gguf dump start data offset via --data-offset and some extra refactor (#8054)
49c03c79cda17913b72260acdc8157b742cee41c 48e6b92cc378c937e59719f2c0f482bf76c9ca81 Xuan Son Nguyen thichthat@gmail.com 2024-06-25T13:59:54+02:00 GitHub noreply@github.com 2024-06-25T13:59:54+02:00 cvector: better prompt handling, add "mean vector" method (#8069)
48e6b92cc378c937e59719f2c0f482bf76c9ca81 3791ad219323389106dc3fd80814eb5bbb7b80de Xuan Son Nguyen thichthat@gmail.com 2024-06-25T13:56:49+02:00 GitHub noreply@github.com 2024-06-25T21:56:49+10:00 Add chat template support for llama-cli (#8068)
3791ad219323389106dc3fd80814eb5bbb7b80de f702a90e245499283d6de0b287701c723cda2a87 HanishKVC hanishkvc@gmail.com 2024-06-25T16:57:35+05:30 GitHub noreply@github.com 2024-06-25T21:27:35+10:00 SimpleChat v3.1: Boolean chat request options in Settings UI, cache_prompt (#7950)
f702a90e245499283d6de0b287701c723cda2a87 083bacce14c1aaf9976aa40e8266cdc25ac749d3 HatsuneMikuUwU33 173229399+HatsuneMikuUwU33@users.noreply.github.com 2024-06-25T10:44:48+02:00 GitHub noreply@github.com 2024-06-25T10:44:48+02:00 Update control vector help (#8104)
083bacce14c1aaf9976aa40e8266cdc25ac749d3 2df373ac40ea581ccca8a58c713f03ad9d4b658d Meng, Hengyu hengyu.meng@intel.com 2024-06-25T10:19:20+08:00 GitHub noreply@github.com 2024-06-25T10:19:20+08:00 [SYCL] Re-enabled mul_mat_batched_sycl (#8095)
2df373ac40ea581ccca8a58c713f03ad9d4b658d 3b099bcd9cbf2434f90cbe40eba6fa2189ed1d02 Johannes Gäßler johannesg@5d6.de 2024-06-25T01:22:33+02:00 GitHub noreply@github.com 2024-06-25T01:22:33+02:00 CUDA: fix matrix multiplication algorithm choice (#8102)
3b099bcd9cbf2434f90cbe40eba6fa2189ed1d02 a818f3028d1497a51cb2b8eb7d993ad58784940e Johannes Gäßler johannesg@5d6.de 2024-06-24T22:15:33+02:00 GitHub noreply@github.com 2024-06-24T22:15:33+02:00 CUDA: fix MMQ writeback for int8 tensor cores (#8100)
a818f3028d1497a51cb2b8eb7d993ad58784940e d62e4aaa02540c89be8b59426340b909d02bbc9e Johannes Gäßler johannesg@5d6.de 2024-06-24T17:43:42+02:00 GitHub noreply@github.com 2024-06-24T17:43:42+02:00 CUDA: use MMQ instead of cuBLAS by default (#8075)
d62e4aaa02540c89be8b59426340b909d02bbc9e 9a590c82262dd518137f85406e65e452fdf2aca3 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-06-24T14:13:39+02:00 GitHub noreply@github.com 2024-06-24T14:13:39+02:00 gguf-py : fix tensor groups for encoder-decoder models in gguf-dump.py (#8090)
9a590c82262dd518137f85406e65e452fdf2aca3 52fc8705a0617452df08333e1161838726c322b4 Johannes Gäßler johannesg@5d6.de 2024-06-24T12:41:23+02:00 GitHub noreply@github.com 2024-06-24T12:41:23+02:00 CUDA: optimize MMQ int8 tensor core performance (#8062)
52fc8705a0617452df08333e1161838726c322b4 8cb508d0d5c024e12692370d85237b45469a004b Christian Zhou-Zheng 59622928+christianazinn@users.noreply.github.com 2024-06-24T05:42:03-04:00 GitHub noreply@github.com 2024-06-24T19:42:03+10:00 Option to split during conversion (#6942)
8cb508d0d5c024e12692370d85237b45469a004b 646ef4a9cfb6f40236d3b9ef07ac03700b6efcc7 slaren slarengh@gmail.com 2024-06-24T07:36:11+02:00 GitHub noreply@github.com 2024-06-24T08:36:11+03:00 disable publishing the full-rocm docker image (#8083)
646ef4a9cfb6f40236d3b9ef07ac03700b6efcc7 de0d6a68ac99f307fe889c48e21124bc3b7ca29a Yann Follet 131855179+YannFollet@users.noreply.github.com 2024-06-24T13:30:24+08:00 GitHub noreply@github.com 2024-06-24T08:30:24+03:00 embedding : more cli arguments (#7458)
de0d6a68ac99f307fe889c48e21124bc3b7ca29a 95f57bb5d5b18ef0beb2702a0d6c06e46804075c fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-06-24T07:06:05+02:00 GitHub noreply@github.com 2024-06-24T07:06:05+02:00 gguf-py, convert-hf : model conversion support for T5 and FLAN-T5 model variants (#5763)
95f57bb5d5b18ef0beb2702a0d6c06e46804075c e112b610a1a75cb7fa8351e1a933e2e7a755a5ce slaren slarengh@gmail.com 2024-06-24T03:07:59+02:00 GitHub noreply@github.com 2024-06-24T03:07:59+02:00 ggml : remove ggml_task_type and GGML_PERF (#8017)
e112b610a1a75cb7fa8351e1a933e2e7a755a5ce 6a2f298bd784403c5c33eebb822217ec5d9b5590 Eddie-Wang wangjinheng1120@163.com 2024-06-24T02:27:57+08:00 GitHub noreply@github.com 2024-06-23T21:27:57+03:00 llama : add support for BitnetForCausalLM (#7931)
6a2f298bd784403c5c33eebb822217ec5d9b5590 11318d9aa1f668aa10407a5cb9614371af32f3ce Aarni Koskela akx@iki.fi 2024-06-23T18:03:08+03:00 GitHub noreply@github.com 2024-06-23T11:03:08-04:00 server : fix JSON-Scheme typo (#7975)
11318d9aa1f668aa10407a5cb9614371af32f3ce b6b9a8e606da7764bd3649c2ea574979c85abd43 Daniel Bevenius daniel.bevenius@gmail.com 2024-06-23T15:39:45+02:00 GitHub noreply@github.com 2024-06-23T15:39:45+02:00 Fix typo in llama_set_embeddings comment (#8077)
b6b9a8e606da7764bd3649c2ea574979c85abd43 45c0e2e4c1268c2d7c8c45536f15e3c9a731ecdc slaren slarengh@gmail.com 2024-06-23T13:14:45+02:00 GitHub noreply@github.com 2024-06-23T13:14:45+02:00 fix CI failures (#8066)
45c0e2e4c1268c2d7c8c45536f15e3c9a731ecdc b5a5f34efadec8d8f0057b35cb04742abfeb2ef5 0cc4m picard12@live.de 2024-06-23T10:21:25+02:00 GitHub noreply@github.com 2024-06-23T10:21:25+02:00 Refactor Vulkan backend to allow multiple contexts (#7961)
b5a5f34efadec8d8f0057b35cb04742abfeb2ef5 3e58b0ee355f78be41cf5211b68426761339bc3c Clint Herron hanclinto@gmail.com 2024-06-22T14:28:18-04:00 GitHub noreply@github.com 2024-06-22T14:28:18-04:00 Removing extra blank lines that were breaking Lint. (#8067)
3e58b0ee355f78be41cf5211b68426761339bc3c adf480c3ab3abedc964c8ae381476257583ae134 Xuan Son Nguyen thichthat@gmail.com 2024-06-22T18:11:30+02:00 GitHub noreply@github.com 2024-06-22T18:11:30+02:00 cvector: fix CI + correct help message (#8064)
adf480c3ab3abedc964c8ae381476257583ae134 3aa184a8c7c553b5dfcc142d919f3db695df297a HatsuneMikuUwU33 173229399+HatsuneMikuUwU33@users.noreply.github.com 2024-06-22T17:19:37+02:00 GitHub noreply@github.com 2024-06-22T17:19:37+02:00 cvector-generator: Moe Moe Fixie-Fixie for Lots of Formats~! ♡(ᐢ ᴥ ᐢ)♡ (#8052)
3aa184a8c7c553b5dfcc142d919f3db695df297a 5b48cd53a87928db0c6447f0c9dac4db5802102d 0xspringtime 110655352+0xspringtime@users.noreply.github.com 2024-06-22T09:37:41-04:00 GitHub noreply@github.com 2024-06-22T15:37:41+02:00 convert-hf : change assert to exception (#8015)
5b48cd53a87928db0c6447f0c9dac4db5802102d c5a8d4b749352645afd4c024f85d6eca2ca72c6d ddh0 dylanhalladay02@icloud.com 2024-06-22T07:16:10-06:00 GitHub noreply@github.com 2024-06-22T15:16:10+02:00 Update llama-quantize ppl/file size output from LLaMA-v1 to Llama-3 values (#8058)
c5a8d4b749352645afd4c024f85d6eca2ca72c6d 557b653dc9ed91e8c313e87500e0050c775f81b6 Clint Herron hanclinto@gmail.com 2024-06-21T23:18:36-04:00 GitHub noreply@github.com 2024-06-21T23:18:36-04:00 JSON Schema to GBNF integration tests (#7790)
557b653dc9ed91e8c313e87500e0050c775f81b6 7d5e8777ae1d21af99d4f95be10db4870720da91 k.h.lai adrian.k.h.lai@outlook.com 2024-06-21T16:28:20+08:00 GitHub noreply@github.com 2024-06-21T10:28:20+02:00 vulkan: detect multiple devices by deviceUUID instead of deviceID (#8022)
7d5e8777ae1d21af99d4f95be10db4870720da91 a927b0f3dd9a86ee042cd2bdcc8c9da4a855926b Eve 139727413+netrunnereve@users.noreply.github.com 2024-06-21T05:57:36Z GitHub noreply@github.com 2024-06-21T08:57:36+03:00 ggml : AVX IQ quants (#7845)
a927b0f3dd9a86ee042cd2bdcc8c9da4a855926b 80ea089d771f0c2d97afa8bead80ded412f600d7 Georgi Gerganov ggerganov@gmail.com 2024-06-21T08:51:28+03:00 GitHub noreply@github.com 2024-06-21T08:51:28+03:00 llama : optimize long word tokenization with WPM (#8034)
80ea089d771f0c2d97afa8bead80ded412f600d7 0e64591e8290037db6412665a56354b789a0597e Douglas Hanley thesecretaryofwar@gmail.com 2024-06-21T00:38:22-05:00 GitHub noreply@github.com 2024-06-21T08:38:22+03:00 llama : allow pooled embeddings on any model (#7477)
0e64591e8290037db6412665a56354b789a0597e b1ef562bc17fbf8ba436ddf2d89b78efd10d3e03 Shuichi Tsutsumi shuichi0526@gmail.com 2024-06-21T14:30:58+09:00 GitHub noreply@github.com 2024-06-21T08:30:58+03:00 swiftui : enable stream updating (#7754)
b1ef562bc17fbf8ba436ddf2d89b78efd10d3e03 17b291a6a581c47f24f99bad926b42617894f99f Hamdoud Hakem 90524568+hamdoudhakem@users.noreply.github.com 2024-06-20T21:01:15+01:00 GitHub noreply@github.com 2024-06-20T22:01:15+02:00 requirements : Bump torch and numpy for python3.12 (#8041)
17b291a6a581c47f24f99bad926b42617894f99f abd894ad96a242043b8e197ec130d8649eead22e Hamdoud Hakem 90524568+hamdoudhakem@users.noreply.github.com 2024-06-20T20:59:59+01:00 GitHub noreply@github.com 2024-06-20T21:59:59+02:00 convert-hf : Fix the encoding in the convert-hf-to-gguf-update.py (#8040)
abd894ad96a242043b8e197ec130d8649eead22e de391e4c803383bbea054b6edd016e78c024a74d Johannes Gäßler johannesg@5d6.de 2024-06-20T16:40:13+02:00 GitHub noreply@github.com 2024-06-20T16:40:13+02:00 common: fix warning (#8036)
de391e4c803383bbea054b6edd016e78c024a74d d50f8897a797a5a03f31228d1b5a7b8130ee1bc2 luoyu-intel yu.luo@intel.com 2024-06-20T13:19:05Z GitHub noreply@github.com 2024-06-20T21:19:05+08:00 [SYCL] Fix windows build and inference (#8003)
d50f8897a797a5a03f31228d1b5a7b8130ee1bc2 2075a66a96cc1b04eabec7cf4b3051193d6f719e Johannes Gäßler johannesg@5d6.de 2024-06-20T14:39:21+02:00 GitHub noreply@github.com 2024-06-20T14:39:21+02:00 CUDA: stream-k decomposition for MMQ (#8018)
2075a66a96cc1b04eabec7cf4b3051193d6f719e ba5899315283eb1df3902363daf79bdc5eefe426 Michael de Gans michael.john.degans@gmail.com 2024-06-19T22:32:01-07:00 GitHub noreply@github.com 2024-06-20T08:32:01+03:00 metal : fix `ggml_metal_supports_op` for BF16 (#8021)
ba5899315283eb1df3902363daf79bdc5eefe426 a7854743c5e6216a6178824a603aa9479728ddd5 sasha0552 admin@sasha0552.org 2024-06-19T23:57:10Z GitHub noreply@github.com 2024-06-20T09:57:10+10:00 server : fix smart slot selection (#8020)
a7854743c5e6216a6178824a603aa9479728ddd5 9c77ec1d74874ee22bdef8f110e8e8d41389abf2 Michael de Gans michael.john.degans@gmail.com 2024-06-19T13:10:42-07:00 GitHub noreply@github.com 2024-06-19T22:10:42+02:00 un-ignore `build-info.cmake` and `build-info.sh` (#7996)
9c77ec1d74874ee22bdef8f110e8e8d41389abf2 a04a953cab583f0229e7b4b506346e3e9a85c8d0 slaren slarengh@gmail.com 2024-06-19T15:04:15+02:00 GitHub noreply@github.com 2024-06-19T15:04:15+02:00 ggml : synchronize threads using barriers (#7993)
a04a953cab583f0229e7b4b506346e3e9a85c8d0 623494a478134432fd2d7ee40135770a3340674f Georgi Gerganov ggerganov@gmail.com 2024-06-19T13:04:36+03:00 GitHub noreply@github.com 2024-06-19T13:04:36+03:00 codecov : remove (#8004)
623494a478134432fd2d7ee40135770a3340674f 37bef8943312d91183ff06d8f1214082a17344a5 Meng, Hengyu hengyu.meng@intel.com 2024-06-19T09:11:51+08:00 GitHub noreply@github.com 2024-06-19T09:11:51+08:00 [SYCL] refactor (#6408)
37bef8943312d91183ff06d8f1214082a17344a5 91c188d6c296bd3384f2a02a83b71187aa3d18b3 jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-06-18T18:40:52+02:00 GitHub noreply@github.com 2024-06-18T18:40:52+02:00 tokenizer : BPE fixes (#7530)
91c188d6c296bd3384f2a02a83b71187aa3d18b3 84f6de17f6a8602e7ff7f7c7bda36a73f510a2dd Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-06-18T14:19:45+02:00 GitHub noreply@github.com 2024-06-18T22:19:45+10:00 Only use FIM middle token if it exists (#7648)
84f6de17f6a8602e7ff7f7c7bda36a73f510a2dd 61665277afde2add00c0d387acb94ed5feb95917 jojorne jojorne@users.noreply.github.com 2024-06-18T09:18:32-03:00 GitHub noreply@github.com 2024-06-18T22:18:32+10:00 Fix no gcc pragma on Windows (#7751)
61665277afde2add00c0d387acb94ed5feb95917 b96f9afb0d58b003ac8d1d0c94cd99393a3bc437 Ulrich Drepper drepper@gmail.com 2024-06-18T14:00:14+02:00 GitHub noreply@github.com 2024-06-18T14:00:14+02:00 Allow compiling with CUDA without CUDA runtime installed (#7989)
b96f9afb0d58b003ac8d1d0c94cd99393a3bc437 1193778105c9a81bd38f72c61aaafbaf85dc9c04 Frank Mai thxcode0824@gmail.com 2024-06-18T15:11:40+08:00 GitHub noreply@github.com 2024-06-18T10:11:40+03:00 chore: clean useless beam search param (#7985)
1193778105c9a81bd38f72c61aaafbaf85dc9c04 5326bcceeb7dd34f16d0fe61b134d1e074a8e65d Abheek Gulati abheekg@hotmail.com 2024-06-17T23:57:41-07:00 GitHub noreply@github.com 2024-06-18T09:57:41+03:00 readme : update UI list (#7943)
5326bcceeb7dd34f16d0fe61b134d1e074a8e65d e6ecc2be470e3c5c6c5c9d8b90aa83a1f7725084 Georgi Gerganov ggerganov@gmail.com 2024-06-18T09:50:45+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-18T09:50:45+03:00 ggml : sync
e6ecc2be470e3c5c6c5c9d8b90aa83a1f7725084 a94e6ff8774b7c9f950d9545baf0ce35e8d1ed2f Georgi Gerganov ggerganov@gmail.com 2024-06-18T09:37:20+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-18T09:50:40+03:00 whisper : use ggml_backend_sched (whisper/2239)
a94e6ff8774b7c9f950d9545baf0ce35e8d1ed2f 5b6da187508f49a9fa9d95fa22ae804a0780d256 Ștefan-Gabriel Muscalu legraphista@users.noreply.github.com 2024-06-17T22:08:46+03:00 GitHub noreply@github.com 2024-06-17T21:08:46+02:00 update: support Qwen2-57B-A14B (#7835)
5b6da187508f49a9fa9d95fa22ae804a0780d256 7c26775adb579e92b59c82e8084c07a1d0f75e9c Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-06-17T23:53:17+05:30 GitHub noreply@github.com 2024-06-17T20:23:17+02:00 Make updates to type cast based on compiler instead of OS (#7851)
7c26775adb579e92b59c82e8084c07a1d0f75e9c b473e95084c286780165568cf0f385f21141d68d Georgi Gerganov ggerganov@gmail.com 2024-06-17T19:40:01+03:00 GitHub noreply@github.com 2024-06-17T19:40:01+03:00 llama : disable FA if KV head size do not match (#7982)
b473e95084c286780165568cf0f385f21141d68d 99052cd227c7182fcf53343d2e7d33bfa180a9cf Bryan Honof bryanhonof@gmail.com 2024-06-17T17:37:55+02:00 GitHub noreply@github.com 2024-06-17T09:37:55-06:00 Add Nix and Flox install instructions (#7899)
99052cd227c7182fcf53343d2e7d33bfa180a9cf c637fcd34d135a9ff4f97d3a53ad03a910a4a31f slaren slarengh@gmail.com 2024-06-17T16:51:42+02:00 GitHub noreply@github.com 2024-06-17T16:51:42+02:00 sched : offload_op also requires supports_op (#7977)
c637fcd34d135a9ff4f97d3a53ad03a910a4a31f 6a2f0b3474d479bda4ac2ee7cfd5dcdcf0be1f79 Frank Mai thxcode0824@gmail.com 2024-06-17T22:11:08+08:00 GitHub noreply@github.com 2024-06-17T16:11:08+02:00 fix: divide 0 exception in mamba (#7932)
6a2f0b3474d479bda4ac2ee7cfd5dcdcf0be1f79 21be9cab94e0b5b53cb6edeeebf8c8c799baad03 Markus Tavenrath mtavenrath@users.noreply.github.com 2024-06-17T16:10:15+02:00 GitHub noreply@github.com 2024-06-17T16:10:15+02:00 Implement non-mapped async IO for CUDA on Windows. (#7896)
21be9cab94e0b5b53cb6edeeebf8c8c799baad03 006167aaf6b6aaa4daa52961035f7460af19f469 Georgi Gerganov ggerganov@gmail.com 2024-06-17T11:09:20+03:00 GitHub noreply@github.com 2024-06-17T11:09:20+03:00 rpc : fix load/store misaligned addresses (#7948)
006167aaf6b6aaa4daa52961035f7460af19f469 df68d4fa5dc929217d3e64d673e099d7a417b206 Brian mofosyne@gmail.com 2024-06-17T15:25:20+10:00 GitHub noreply@github.com 2024-06-17T15:25:20+10:00 gguf-dump.py: add --markdown dump output (#7853)
df68d4fa5dc929217d3e64d673e099d7a417b206 43b35e38ba371f9a7faa6dca4c5d1e8f698ffd87 Neo Zhang zhang.jianyu@outlook.com 2024-06-17T11:17:07+08:00 GitHub noreply@github.com 2024-06-17T11:17:07+08:00 [SYCL] Update README-sycl.md for Chapter "Recommended release" and "News" (#7946)
43b35e38ba371f9a7faa6dca4c5d1e8f698ffd87 19b7a836f6658e18e973af532a5cc6ad6b3a27f8 Calvin Laurenson calvin@laurenson.dev 2024-06-16T15:23:04-07:00 GitHub noreply@github.com 2024-06-17T00:23:04+02:00 Add support for sqrt on CUDA (#7953)
19b7a836f6658e18e973af532a5cc6ad6b3a27f8 b5fcf8ef5c29df53cfff60e180b4992a3b2332a6 Georgi Gerganov ggerganov@gmail.com 2024-06-11T17:39:01+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-16T20:32:49+03:00 cuda : fix bounds check for src0 rows in MMVQ kernel (whisper/2231)
b5fcf8ef5c29df53cfff60e180b4992a3b2332a6 398105ff4373eea385ea8e8625cb417b2ae51134 Hong Bo PENG penghb@cn.ibm.com 2024-06-16T16:53:11+08:00 Georgi Gerganov ggerganov@gmail.com 2024-06-16T20:32:49+03:00 ggml : fix and optimize ppc64le (ggml/849)
398105ff4373eea385ea8e8625cb417b2ae51134 bc6c457fa35f6791e9a2bb61108e7d49e8fc98bd Daniel Bevenius daniel.bevenius@gmail.com 2024-06-16T10:51:18+02:00 Georgi Gerganov ggerganov@gmail.com 2024-06-16T20:32:49+03:00 ggml : remove duplicate include of ggml-common.h (ggml/853)
bc6c457fa35f6791e9a2bb61108e7d49e8fc98bd 52399254b3bceda279b4ea9111a983e32310166e Georgi Gerganov ggerganov@gmail.com 2024-06-16T19:16:21+03:00 GitHub noreply@github.com 2024-06-16T09:16:21-07:00 flake.lock: Update (#7951)
52399254b3bceda279b4ea9111a983e32310166e 6fe1c627413725ddc1f9e323f6b13fe388c53e0a Georgi Gerganov ggerganov@gmail.com 2024-06-16T14:51:40+03:00 GitHub noreply@github.com 2024-06-16T14:51:40+03:00 unicode : avoid char32_t (#7957)
6fe1c627413725ddc1f9e323f6b13fe388c53e0a cddaf028adc738b5a7ecc60809cb78e0ba0f97c1 hopkins385 98618192+hopkins385@users.noreply.github.com 2024-06-16T13:51:18+02:00 GitHub noreply@github.com 2024-06-16T14:51:18+03:00 readme : update UI list [no ci] (#7958)
cddaf028adc738b5a7ecc60809cb78e0ba0f97c1 c8a82194a888f68f259e0d0fa96f3332ac7c5cb6 Georgi Gerganov ggerganov@gmail.com 2024-06-16T14:50:12+03:00 GitHub noreply@github.com 2024-06-16T14:50:12+03:00 ggml : fix handling of zero blocks in IQ quants (#7955)
c8a82194a888f68f259e0d0fa96f3332ac7c5cb6 7c7836d9d4062d6858e3fb337b135c417ccee6ce Georgi Gerganov ggerganov@gmail.com 2024-06-16T10:46:51+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-16T10:46:51+03:00 github : update pr template
7c7836d9d4062d6858e3fb337b135c417ccee6ce 0c7b3595b9e5ad2355818e259f06b0dc3f0065b3 0cc4m picard12@live.de 2024-06-16T07:17:31+02:00 GitHub noreply@github.com 2024-06-16T07:17:31+02:00 Vulkan Shader Refactor, Memory Debugging Option (#7947)
0c7b3595b9e5ad2355818e259f06b0dc3f0065b3 7b2f4a7d193ef2475259bbe7656fcccfab4b1217 Xuan Son Nguyen thichthat@gmail.com 2024-06-15T18:53:40+02:00 GitHub noreply@github.com 2024-06-15T18:53:40+02:00 Add `cvector-generator` example (#7514)
7b2f4a7d193ef2475259bbe7656fcccfab4b1217 f8ec8877b75774fc6c47559d529dac423877bcad Meng, Hengyu hengyu.meng@intel.com 2024-06-15T14:05:10+08:00 GitHub noreply@github.com 2024-06-15T14:05:10+08:00 [SYCL] remove global variables (#7710)
f8ec8877b75774fc6c47559d529dac423877bcad 76d66ee0be91e2bec93206e821ee1db8d023cff5 olexiyb olexiyb@gmail.com 2024-06-14T20:28:34+03:00 GitHub noreply@github.com 2024-06-14T20:28:34+03:00 ci : fix macos x86 build (#7940)
76d66ee0be91e2bec93206e821ee1db8d023cff5 66ef1ceedf983773c8ceb4d925285d41d4e50e2a Johannes Gäßler johannesg@5d6.de 2024-06-14T18:41:49+02:00 GitHub noreply@github.com 2024-06-14T18:41:49+02:00 CUDA: faster q2_K, q3_K MMQ + int8 tensor cores (#7921)
66ef1ceedf983773c8ceb4d925285d41d4e50e2a e65bbf606c61f49dc06c7ac060cd5ba7ae446025 Georgi Gerganov ggerganov@gmail.com 2024-06-14T17:14:09+03:00 GitHub noreply@github.com 2024-06-14T17:14:09+03:00 metal : utilize max shared memory for mul_mat_id (#7935)
e65bbf606c61f49dc06c7ac060cd5ba7ae446025 6fcd1331efbfbb89c8c96eba2321bb7b4d0c40e4 Radoslav Gerganov rgerganov@gmail.com 2024-06-14T16:47:41+03:00 GitHub noreply@github.com 2024-06-14T16:47:41+03:00 llama-bench : fix RPC indication (#7936)
6fcd1331efbfbb89c8c96eba2321bb7b4d0c40e4 41b9260f18eb7f325c952006ac46afc1d0d8ad2f Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-06-14T12:20:04+02:00 GitHub noreply@github.com 2024-06-14T13:20:04+03:00 llama : more checks before assuming FIM tokens (#7644)
41b9260f18eb7f325c952006ac46afc1d0d8ad2f 172c8256840ffd882ab9992ecedbb587d9b21f15 Elaine elaine.zosa@gmail.com 2024-06-14T13:16:49+03:00 GitHub noreply@github.com 2024-06-14T13:16:49+03:00 convert : add Poro-34B-chat tokenizer support (#7713)
172c8256840ffd882ab9992ecedbb587d9b21f15 a55eb1bf0fa2fd84147bdfd384391e029d988253 Radoslav Gerganov rgerganov@gmail.com 2024-06-13T15:18:44+03:00 GitHub noreply@github.com 2024-06-13T15:18:44+03:00 rpc : fix ggml_backend_rpc_supports_buft() (#7918)
a55eb1bf0fa2fd84147bdfd384391e029d988253 f578b86b2123d0f92afbaa98a031df4d4464e582 Galunid karolek1231456@gmail.com 2024-06-13T09:42:41+02:00 GitHub noreply@github.com 2024-06-13T09:42:41+02:00 readme : Remove outdated instructions from README.md (#7914) [no ci]
f578b86b2123d0f92afbaa98a031df4d4464e582 1c641e6aac5c18b964e7b32d9dbbb4bf5301d0d7 slaren slarengh@gmail.com 2024-06-13T03:11:35+02:00 GitHub noreply@github.com 2024-06-13T03:11:35+02:00 move BLAS to a separate backend (#6210)
1c641e6aac5c18b964e7b32d9dbbb4bf5301d0d7 963552903f51043ee947a8deeaaa7ec00bc3f1a4 Olivier Chafik ochafik@users.noreply.github.com 2024-06-13T00:41:52+01:00 GitHub noreply@github.com 2024-06-13T00:41:52+01:00 `build`: rename main → llama-cli, server → llama-server, llava-cli → llama-llava-cli, etc... (#7809)
963552903f51043ee947a8deeaaa7ec00bc3f1a4 a9cae48003dfc4fe95b8f5c81682fc6e63425235 Johannes Gäßler johannesg@5d6.de 2024-06-12T17:41:51+02:00 GitHub noreply@github.com 2024-06-12T17:41:51+02:00 CUDA: fix broken oob check for FA vec f32 kernel (#7904)
a9cae48003dfc4fe95b8f5c81682fc6e63425235 bfaa676b0841617d4ef3596e63aca6be1a8eb1b5 Georgi Gerganov ggerganov@gmail.com 2024-06-12T16:00:22+03:00 GitHub noreply@github.com 2024-06-12T16:00:22+03:00 tests : add non-cont unary tests (#7857)
bfaa676b0841617d4ef3596e63aca6be1a8eb1b5 704a35b183748954013bd875bbbfdd9eaca14e62 Georgi Gerganov ggerganov@gmail.com 2024-06-12T15:24:20+03:00 GitHub noreply@github.com 2024-06-12T15:24:20+03:00 ggml : improve ggml_is_contiguous logic (#7856)
704a35b183748954013bd875bbbfdd9eaca14e62 dcf752707d96eb305f546526c7bc5d01f0831130 Georgi Gerganov ggerganov@gmail.com 2024-06-12T14:42:29+03:00 GitHub noreply@github.com 2024-06-12T14:42:29+03:00 server : restore numeric prompts (#7883)
dcf752707d96eb305f546526c7bc5d01f0831130 f2b5764beb35583295e2475479c18f249b139b58 Meng, Hengyu hengyu.meng@intel.com 2024-06-12T17:05:35+08:00 GitHub noreply@github.com 2024-06-12T19:05:35+10:00 update intel docker oneapi-basekit to 2024.1.1-devel-ubuntu22.04 (#7894)
f2b5764beb35583295e2475479c18f249b139b58 73bac2b11d7d3e20982fc9ee607625836387db8b Patrice Ferlet metal3d@gmail.com 2024-06-12T03:18:16+02:00 GitHub noreply@github.com 2024-06-12T11:18:16+10:00 Fix a typo and add Fedora 40 pacakge to install for Vulkan (#7794) [no ci]
73bac2b11d7d3e20982fc9ee607625836387db8b ef52d1d16afc695d798396cdd13594ea5e45a9dd k.h.lai adrian.k.h.lai@outlook.com 2024-06-12T03:26:05+08:00 GitHub noreply@github.com 2024-06-11T21:26:05+02:00 vulkan: select only one device for single gpu with multiple drivers (#7582)
ef52d1d16afc695d798396cdd13594ea5e45a9dd 14f83526cd27f638c856ea6eff08110b9860eb2a 0cc4m picard12@live.de 2024-06-11T21:20:29+02:00 GitHub noreply@github.com 2024-06-11T21:20:29+02:00 Update Vulkan RoPE implementation (#7818)
14f83526cd27f638c856ea6eff08110b9860eb2a 6fe42d073f0554eada93ac9d40574025aeedb703 Deven Mistry 31466137+deven367@users.noreply.github.com 2024-06-11T12:18:58-04:00 GitHub noreply@github.com 2024-06-12T02:18:58+10:00 fix broken link in pr template (#7880) [no ci]
6fe42d073f0554eada93ac9d40574025aeedb703 148995e5e57b313cce2672f75610db58c6327a51 Brian mofosyne@gmail.com 2024-06-12T00:43:41+10:00 GitHub noreply@github.com 2024-06-11T17:43:41+03:00 github: move PR template to .github/ root (#7868)
148995e5e57b313cce2672f75610db58c6327a51 4bfe50f741479c1df1c377260c3ff5702586719e Johannes Gäßler johannesg@5d6.de 2024-06-11T14:45:40+02:00 GitHub noreply@github.com 2024-06-11T14:45:40+02:00 llama-bench: more compact markdown tables (#7879)
4bfe50f741479c1df1c377260c3ff5702586719e bdcb8f42221bc40c411150a009a3d3a30fa74722 Georgi Gerganov ggerganov@gmail.com 2024-06-11T10:10:20+03:00 GitHub noreply@github.com 2024-06-11T10:10:20+03:00 tests : check the Python version (#7872)
bdcb8f42221bc40c411150a009a3d3a30fa74722 c2ce6c47e4f2d891bf29d8810832a3b310a8f205 Johannes Gäßler johannesg@5d6.de 2024-06-11T08:26:07+02:00 GitHub noreply@github.com 2024-06-11T08:26:07+02:00 CUDA: int8 tensor cores for MMQ (q4_K, q5_K, q6_K) (#7860)
c2ce6c47e4f2d891bf29d8810832a3b310a8f205 b61eb9644d64e90123ac805436d95b94b3b4cc3f slaren slarengh@gmail.com 2024-06-11T07:59:20+02:00 GitHub noreply@github.com 2024-06-11T08:59:20+03:00 fix CUDA CI by using a windows-2019 image (#7861)
b61eb9644d64e90123ac805436d95b94b3b4cc3f 396b18dfec2c56846e80362db70af09b9e1d70ba Olivier Chafik ochafik@users.noreply.github.com 2024-06-11T02:22:57+01:00 GitHub noreply@github.com 2024-06-11T02:22:57+01:00 json: refine constraint for whitespace to avoid runaways yet allow pretty print (#7866)
396b18dfec2c56846e80362db70af09b9e1d70ba 864a99e7a01d9422d2f55618dbe62c8099a2175c Olivier Chafik ochafik@users.noreply.github.com 2024-06-11T01:00:30+01:00 GitHub noreply@github.com 2024-06-11T01:00:30+01:00 `json`: document schema conversion in GBNF readme, align manual grammar examples & converters (#7841)
864a99e7a01d9422d2f55618dbe62c8099a2175c fd5ea0f897ecb3659d6c269ef6f3d833e865ead7 Jared Van Bortel jared@nomic.ai 2024-06-10T18:32:10-04:00 GitHub noreply@github.com 2024-06-10T18:32:10-04:00 cmake : fix CMake requirement for CUDA (#7821)
fd5ea0f897ecb3659d6c269ef6f3d833e865ead7 c28a83902cf6ab6a9e085ad6d4cc2e95c4ccfe40 slaren slarengh@gmail.com 2024-06-10T14:18:41+02:00 GitHub noreply@github.com 2024-06-10T15:18:41+03:00 ci : try win-2019 on server windows test (#7854)
c28a83902cf6ab6a9e085ad6d4cc2e95c4ccfe40 d9da0e4986f121c727bdd9579a6688097b11602c Georgi Gerganov ggerganov@gmail.com 2024-06-10T15:00:15+03:00 GitHub noreply@github.com 2024-06-10T15:00:15+03:00 examples : remove --instruct remnants (#7846)
d9da0e4986f121c727bdd9579a6688097b11602c 1f0dabda8d5c131f9d4632aa41de74317cdd61fb Georgi Gerganov ggerganov@gmail.com 2024-06-10T14:59:55+03:00 GitHub noreply@github.com 2024-06-10T14:59:55+03:00 server : improve "prompt" handling (#7847)
1f0dabda8d5c131f9d4632aa41de74317cdd61fb af4ae502ddaeb03cd5861273ca2e9a5ae4551db7 Johannes Gäßler johannesg@5d6.de 2024-06-10T11:45:13+02:00 GitHub noreply@github.com 2024-06-10T11:45:13+02:00 CUDA: use tensor cores for MMQ (#7676)
af4ae502ddaeb03cd5861273ca2e9a5ae4551db7 10ceba354a3b152ff425e9fa97f9caaef99a46b1 Ben Ashbaugh ben.ashbaugh@intel.com 2024-06-10T02:21:31-07:00 GitHub noreply@github.com 2024-06-10T10:21:31+01:00 use the correct SYCL context for host USM allocations (#7777)
10ceba354a3b152ff425e9fa97f9caaef99a46b1 e95beeb1fc4621826ddd616776dbdf717366bf5c Georgi Gerganov ggerganov@gmail.com 2024-06-10T02:04:50+03:00 GitHub noreply@github.com 2024-06-09T16:04:50-07:00 flake.lock: Update (#7838)
e95beeb1fc4621826ddd616776dbdf717366bf5c 57bf62ce7cb75cca589943e2050d29bff4026e76 Georgi Gerganov ggerganov@gmail.com 2024-06-09T20:19:35+03:00 GitHub noreply@github.com 2024-06-09T20:19:35+03:00 imatrix : handle partial entries (#7833)
57bf62ce7cb75cca589943e2050d29bff4026e76 3e2ee443159724e2d3a0741f6b167e599ec088aa Nicolás Pérez nicolas_perez@brown.edu 2024-06-09T11:24:29-04:00 GitHub noreply@github.com 2024-06-10T01:24:29+10:00 docs: Added initial PR template with directions for doc only changes and squash merges [no ci] (#7700)
3e2ee443159724e2d3a0741f6b167e599ec088aa 42b53d192f4e3abf1b7c8e424628424504ea5dc5 mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2024-06-09T12:50:35+02:00 GitHub noreply@github.com 2024-06-09T20:50:35+10:00 server: do not remove whitespace at the start of a completion chunk (#7830)
42b53d192f4e3abf1b7c8e424628424504ea5dc5 2decf57bc6e4a6b45176c3727d964a01161beecc Johannes Gäßler johannesg@5d6.de 2024-06-09T09:42:25+02:00 GitHub noreply@github.com 2024-06-09T09:42:25+02:00 CUDA: revise q8_1 data layout for mul_mat_q (#7824)
2decf57bc6e4a6b45176c3727d964a01161beecc 5795b941827fdec6c1662986de962badff456718 sasha0552 admin@sasha0552.org 2024-06-09T06:39:25Z GitHub noreply@github.com 2024-06-09T16:39:25+10:00 convert-hf : set the model name based on cli arg, if present (#7693)
5795b941827fdec6c1662986de962badff456718 ed9f2521185706481501a5e6d5315397b11802ff compilade git@compilade.net 2024-06-08T22:47:25-04:00 GitHub noreply@github.com 2024-06-09T12:47:25+10:00 convert-hf : match model part name prefix and suffix (#7687)
ed9f2521185706481501a5e6d5315397b11802ff fe1e3917cfa0f9397a765cfd0aef880674d938d5 compilade git@compilade.net 2024-06-08T22:34:29-04:00 GitHub noreply@github.com 2024-06-09T12:34:29+10:00 gguf-py : decouple adding metadata from writing in GGUFWriter (#7827)
fe1e3917cfa0f9397a765cfd0aef880674d938d5 d4d915d351d1f1270d56184bdd46672893e8a5d8 slaren slarengh@gmail.com 2024-06-09T01:43:39+02:00 GitHub noreply@github.com 2024-06-09T01:43:39+02:00 Revert "[SYCL] Update rpc-server.cpp to include SYCL backend (#7682)" (#7808)
d4d915d351d1f1270d56184bdd46672893e8a5d8 7a16ce7db2a74a223f0f3b9cee66d4539c5bce8f Olivier Chafik ochafik@users.noreply.github.com 2024-06-08T20:21:08+01:00 GitHub noreply@github.com 2024-06-08T21:21:08+02:00 url: save -mu downloads to new cache location (#7826)
7a16ce7db2a74a223f0f3b9cee66d4539c5bce8f da799b41891e34aac86ce4e173f9c4c0afd4fab3 sasha0552 admin@sasha0552.org 2024-06-08T07:50:31Z GitHub noreply@github.com 2024-06-08T10:50:31+03:00 server : smart slot selection using Longest Common Prefix (#7728)
da799b41891e34aac86ce4e173f9c4c0afd4fab3 c00fad71e507ff386d42bd74846fe06d19dd63a4 slaren slarengh@gmail.com 2024-06-07T19:47:49+02:00 GitHub noreply@github.com 2024-06-07T19:47:49+02:00 vulkan : reuse parent extra for views (#7806)
c00fad71e507ff386d42bd74846fe06d19dd63a4 27615f5ab21060d96953c9c1e223051ab2188f57 Christian Zhou-Zheng 59622928+christianazinn@users.noreply.github.com 2024-06-07T08:56:01-04:00 GitHub noreply@github.com 2024-06-07T15:56:01+03:00 gguf-split : change binary multi-byte units to decimal (#7803)
27615f5ab21060d96953c9c1e223051ab2188f57 7027b27d765db95d4ac6b569d976e387a8715881 intelmatt 61025942+intelmatt@users.noreply.github.com 2024-06-07T05:15:07-07:00 GitHub noreply@github.com 2024-06-07T15:15:07+03:00 cmake : fix BUILD_SHARED_LIBS=ON build (#7784)
7027b27d765db95d4ac6b569d976e387a8715881 a5cabd76491f07494c5b8267f921c73f5e2bbfb4 Johannes Gäßler johannesg@5d6.de 2024-06-07T11:15:49+02:00 GitHub noreply@github.com 2024-06-07T11:15:49+02:00 server: update cache_prompt documentation [no ci] (#7745)
a5cabd76491f07494c5b8267f921c73f5e2bbfb4 d5c938cd7716b9a2ace49a43a469dfbffcff4d28 woodx 124784234+woodx9@users.noreply.github.com 2024-06-07T15:09:45+08:00 GitHub noreply@github.com 2024-06-07T10:09:45+03:00 server : do not get prompt in infill mode (#7286)
d5c938cd7716b9a2ace49a43a469dfbffcff4d28 c9ee7118d5644dd3df70ea6878b36a9761616aab pengxin99 pengxin.yuan@intel.com 2024-06-07T14:28:26+08:00 GitHub noreply@github.com 2024-06-07T14:28:26+08:00 [SYCL] fix softmax r2r result wrong issue (#7811)
c9ee7118d5644dd3df70ea6878b36a9761616aab ee459f40f65810a810151b24eba5b8bd174ceffe slaren slarengh@gmail.com 2024-06-07T08:01:29+02:00 GitHub noreply@github.com 2024-06-07T09:01:29+03:00 check for nans in imatrix and quantize (#7807)
ee459f40f65810a810151b24eba5b8bd174ceffe f83351f9a62a6262f1fc3d08f320033089cddfb5 Georgi Gerganov ggerganov@gmail.com 2024-06-06T19:19:59+03:00 GitHub noreply@github.com 2024-06-06T19:19:59+03:00 server : fix --threads-http arg (#7801)
f83351f9a62a6262f1fc3d08f320033089cddfb5 ad675e1c67a05b16e4e12abe30dbecfc808e7b7e Georgi Gerganov ggerganov@gmail.com 2024-06-06T16:30:58+03:00 GitHub noreply@github.com 2024-06-06T16:30:58+03:00 imatrix : migrate to gpt_params (#7771)
ad675e1c67a05b16e4e12abe30dbecfc808e7b7e a143c04375828b1f72eb1a326115791b63e79345 Clint Herron hanclinto@gmail.com 2024-06-06T06:08:52-07:00 GitHub noreply@github.com 2024-06-06T06:08:52-07:00 Added support for . (any character) token in grammar engine. (#6467)
a143c04375828b1f72eb1a326115791b63e79345 55b2d0849d3ec9e45e4a4d9e480f5fa7977872a6 Mattheus Chediak shammcity00@gmail.com 2024-06-06T09:17:54-03:00 GitHub noreply@github.com 2024-06-06T22:17:54+10:00 README minor fixes (#7798) [no ci]
55b2d0849d3ec9e45e4a4d9e480f5fa7977872a6 f5d7b268ec4bf8628aa6ccc9f6631d0230dde76f Olivier Chafik ochafik@users.noreply.github.com 2024-06-06T10:07:06+01:00 GitHub noreply@github.com 2024-06-06T10:07:06+01:00 grammars: x{min,max} repetition operator (#6640)
f5d7b268ec4bf8628aa6ccc9f6631d0230dde76f 2d08b7fbb483c14bd2b173d4cd51ea3a4f862e8f Joan Fontanals joan.fontanals.martinez@jina.ai 2024-06-06T09:22:41+02:00 GitHub noreply@github.com 2024-06-06T10:22:41+03:00 llama : add jina v2 base code (#7596)
2d08b7fbb483c14bd2b173d4cd51ea3a4f862e8f d67caea0d6e6c303d31b01d0a010973e6c908dff slaren slarengh@gmail.com 2024-06-06T07:19:49+02:00 GitHub noreply@github.com 2024-06-06T08:19:49+03:00 docker : build only main and server in their images (#7782)
d67caea0d6e6c303d31b01d0a010973e6c908dff 7672adeec7a79ea271058c63106c142ba84f951a slaren slarengh@gmail.com 2024-06-06T07:17:21+02:00 GitHub noreply@github.com 2024-06-06T08:17:21+03:00 docker : add openmp lib (#7780)
7672adeec7a79ea271058c63106c142ba84f951a 7d1a378b8fb266782d9248538a661405aad80768 Galunid karolek1231456@gmail.com 2024-06-05T19:07:24+02:00 GitHub noreply@github.com 2024-06-06T03:07:24+10:00 Fix encoding in python scripts (#7733)
7d1a378b8fb266782d9248538a661405aad80768 2b3389677a833cee0880226533a1768b1a9508d2 Johannes Gäßler johannesg@5d6.de 2024-06-05T16:53:00+02:00 GitHub noreply@github.com 2024-06-05T16:53:00+02:00 CUDA: refactor mmq, dmmv, mmvq (#7716)
2b3389677a833cee0880226533a1768b1a9508d2 9973e81c5ccf4f31b3980f5aa73f5cfea8699860 Georgi Gerganov ggerganov@gmail.com 2024-06-05T11:29:20+03:00 GitHub noreply@github.com 2024-06-05T11:29:20+03:00 ggml : refactor rope norm/neox (#7634)
9973e81c5ccf4f31b3980f5aa73f5cfea8699860 c90dbe026b456a233f8f0fbe752212e6a0503ca2 arch-btw 57669023+arch-btw@users.noreply.github.com 2024-06-04T23:40:49-07:00 GitHub noreply@github.com 2024-06-05T09:40:49+03:00 readme : remove -ins (#7759)
c90dbe026b456a233f8f0fbe752212e6a0503ca2 b90dc566c1c615289b05b50d61680f23744a21e7 jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-06-05T01:26:14+02:00 GitHub noreply@github.com 2024-06-05T01:26:14+02:00 Fix per token atrributes bits (#7749)
b90dc566c1c615289b05b50d61680f23744a21e7 1442677f92e45a475be7b4d056e3633d1d6f813b agray3 agray3@users.noreply.github.com 2024-06-04T21:06:49+01:00 GitHub noreply@github.com 2024-06-04T22:06:49+02:00 Allow number of nodes in CUDA graph to change (#7738)
1442677f92e45a475be7b4d056e3633d1d6f813b 554c247caffed64465f372661f2826640cb10430 Georgi Gerganov ggerganov@gmail.com 2024-06-04T21:23:39+03:00 GitHub noreply@github.com 2024-06-04T21:23:39+03:00 common : refactor cli arg parsing (#7675)
554c247caffed64465f372661f2826640cb10430 0cd6bd3483fa66124b76a8a8ac794d9ee18c70c1 Georgi Gerganov ggerganov@gmail.com 2024-06-04T21:23:20+03:00 GitHub noreply@github.com 2024-06-04T21:23:20+03:00 ggml : remove OpenCL (#7735)
0cd6bd3483fa66124b76a8a8ac794d9ee18c70c1 5ca0944a153b65724d51b2f484139aa25ccb7a8b Georgi Gerganov ggerganov@gmail.com 2024-06-04T21:23:05+03:00 GitHub noreply@github.com 2024-06-04T21:23:05+03:00 llama : remove beam search (#7736)
5ca0944a153b65724d51b2f484139aa25ccb7a8b adc9ff384121f4d550d28638a646b336d051bf42 Georgi Gerganov ggerganov@gmail.com 2024-06-04T19:43:01+03:00 Georgi Gerganov ggerganov@gmail.com 2024-06-04T19:43:01+03:00 readme : remove obsolete Zig instructions (#7471)
adc9ff384121f4d550d28638a646b336d051bf42 987d743d6bc4cee4bde6820733ea33a2abc0afac slaren slarengh@gmail.com 2024-06-04T14:32:42+02:00 GitHub noreply@github.com 2024-06-04T14:32:42+02:00 llama-bench : allow using a different printer for stderr with -oe (#7722)
987d743d6bc4cee4bde6820733ea33a2abc0afac b226c1227bcf6412076ecf787421135fd2c42ef0 Daniele 57776841+daniandtheweb@users.noreply.github.com 2024-06-04T12:09:15Z GitHub noreply@github.com 2024-06-04T14:09:15+02:00 Improve hipBLAS support in CMake (#7696)
b226c1227bcf6412076ecf787421135fd2c42ef0 3b38d48609280aa5f8ab7ea135a4351b2a5ee240 zhouwg zhouwg2000@gmail.com 2024-06-04T19:21:26+08:00 GitHub noreply@github.com 2024-06-04T21:21:26+10:00 refine .gitignore (#7688)
3b38d48609280aa5f8ab7ea135a4351b2a5ee240 6d1616944d9efd342ed2a4fd318722adfc9febcd jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-06-04T09:17:17+02:00 GitHub noreply@github.com 2024-06-04T09:17:17+02:00 Per token attributes (#7685)
6d1616944d9efd342ed2a4fd318722adfc9febcd bde7cd3cd949c1a85d3a199498ac98e78039d46f Georgi Gerganov ggerganov@gmail.com 2024-06-04T10:01:09+03:00 GitHub noreply@github.com 2024-06-04T17:01:09+10:00 ggml : prevent builds with -ffinite-math-only (#7726)
bde7cd3cd949c1a85d3a199498ac98e78039d46f a5735e4426b19a3ebd0c653ad8ac01420458ee95 Radoslav Gerganov rgerganov@gmail.com 2024-06-03T20:03:26+03:00 GitHub noreply@github.com 2024-06-03T20:03:26+03:00 llama : offload to RPC in addition to other backends (#7640)
a5735e4426b19a3ebd0c653ad8ac01420458ee95 0b832d53ba0ffcc759c8d62ede3772dd62321f8e Masaya, Kato 62578291+msy-kato@users.noreply.github.com 2024-06-04T00:14:15+09:00 GitHub noreply@github.com 2024-06-03T17:14:15+02:00 ggml : use OpenMP as a thread pool (#7606)
0b832d53ba0ffcc759c8d62ede3772dd62321f8e 3d7ebf63123b8652fb7bbecef7ba731202309901 Johannes Gäßler johannesg@5d6.de 2024-06-03T16:28:58+02:00 GitHub noreply@github.com 2024-06-03T16:28:58+02:00 make: fix debug options not being applied to NVCC (#7714)
3d7ebf63123b8652fb7bbecef7ba731202309901 a10cda58d3199cd85305e0f03a8c6056714ae2e8 0cc4m picard12@live.de 2024-06-03T10:59:14+02:00 GitHub noreply@github.com 2024-06-03T10:59:14+02:00 Vulkan Mixture of Experts (MoE) support (#7628)
a10cda58d3199cd85305e0f03a8c6056714ae2e8 6f28a333c1e3fdfdc7b4f9d0367f2b41a9b7e9d4 Andy Tai andy-tai@users.noreply.github.com 2024-06-03T01:06:24-07:00 GitHub noreply@github.com 2024-06-03T11:06:24+03:00 cmake : add pkg-config spec file for llama.cpp (#7702)
6f28a333c1e3fdfdc7b4f9d0367f2b41a9b7e9d4 549279d8049d78620a2b081e26edb654f83c3bbd zhangkaihuo zhangkaihuo@gmail.com 2024-06-03T15:49:30+08:00 GitHub noreply@github.com 2024-06-03T10:49:30+03:00 llama : MiniCPM support tied embeddings (#7664)
549279d8049d78620a2b081e26edb654f83c3bbd 9e405b6e2ecb888e860f7b92720b4809e21b3915 Georgi Gerganov ggerganov@gmail.com 2024-06-03T08:34:43+03:00 GitHub noreply@github.com 2024-06-03T08:34:43+03:00 llama : avoid double token-to-piece cache (#7654)
9e405b6e2ecb888e860f7b92720b4809e21b3915 3413ae2193d0693f14bead02e5018f442cbf579b woachk 24752637+woachk@users.noreply.github.com 2024-06-03T07:32:16+02:00 GitHub noreply@github.com 2024-06-03T08:32:16+03:00 kompute : implement op_getrows_f32 (#6403)
3413ae2193d0693f14bead02e5018f442cbf579b 1669810d7c2446af8425aa54ff6611bf6f14646c Dave Airlie airlied@redhat.com 2024-06-03T07:59:54+10:00 GitHub noreply@github.com 2024-06-02T17:59:54-04:00 fix bug introduced in using calloc (#7701)
1669810d7c2446af8425aa54ff6611bf6f14646c 7c4e5b7eae26581869e782015d9deca947c34997 Georgi Gerganov ggerganov@gmail.com 2024-06-03T00:13:12+03:00 GitHub noreply@github.com 2024-06-02T14:13:12-07:00 flake.lock: Update (#7686)
7c4e5b7eae26581869e782015d9deca947c34997 9422c5e34bbd302493b77a8f6d546154a1f4fe82 Austin 77757836+teleprint-me@users.noreply.github.com 2024-06-02T13:39:08-04:00 GitHub noreply@github.com 2024-06-02T20:39:08+03:00 chore : add ignore rule for generated server themes (#7689)
9422c5e34bbd302493b77a8f6d546154a1f4fe82 e141ce624af57bdffbaf57014a044eb1d9689230 nickp27 nb.porter@gmail.com 2024-06-02T19:13:54+10:00 GitHub noreply@github.com 2024-06-02T12:13:54+03:00 [SYCL] Update rpc-server.cpp to include SYCL backend (#7682)
e141ce624af57bdffbaf57014a044eb1d9689230 2e666832e6ac78194edf030bd1c295e21bdb022c Johannes Gäßler johannesg@5d6.de 2024-06-01T23:26:10+02:00 GitHub noreply@github.com 2024-06-01T23:26:10+02:00 Fix FlashAttention debug test, FP32 assert (#7684)
2e666832e6ac78194edf030bd1c295e21bdb022c 2ac95c9d5678d05e253691fb1f26471675bff5ad Yazan Agha-Schrader mountaiin@icloud.com 2024-06-01T21:31:48+02:00 GitHub noreply@github.com 2024-06-01T22:31:48+03:00 server : new UI (#7633)
2ac95c9d5678d05e253691fb1f26471675bff5ad 750f60c03e4d3f53fa51910551ce87a3d508d2d7 HanishKVC hanishkvc@gmail.com 2024-06-01T21:50:18+05:30 GitHub noreply@github.com 2024-06-02T02:20:18+10:00 SimpleChat: Simple histogram/repeatMatching driven garbageTrimming, Settings UI, Streaming mode, OpenAi Compat (Model, Authorization Bearer), Save/Restore session, Auto Settings UI (#7548)
750f60c03e4d3f53fa51910551ce87a3d508d2d7 9b596417af11c9ac44fcae0fcfbc6f3665089083 Johannes Gäßler johannesg@5d6.de 2024-06-01T15:47:04+02:00 GitHub noreply@github.com 2024-06-01T15:47:04+02:00 CUDA: fix Pascal FA, deq. KV to FP16 for batch > 8 (#7681)
9b596417af11c9ac44fcae0fcfbc6f3665089083 a323ec60af14a33d560df98f2cc41b4112cb4f80 Johannes Gäßler johannesg@5d6.de 2024-06-01T08:44:14+02:00 GitHub noreply@github.com 2024-06-01T08:44:14+02:00 CUDA: quantized KV support for FA vec (#7527)
a323ec60af14a33d560df98f2cc41b4112cb4f80 0515ad93f48df63bbff204eddb0cac75e8585c65 Georgi Gerganov ggerganov@gmail.com 2024-05-31T22:23:04+03:00 GitHub noreply@github.com 2024-05-31T22:23:04+03:00 server : update js (#7670)
0515ad93f48df63bbff204eddb0cac75e8585c65 c8047d538f3addab40e3112be60bb92e70ce1a50 Galunid karolek1231456@gmail.com 2024-05-31T17:42:33+02:00 GitHub noreply@github.com 2024-05-31T17:42:33+02:00 convert-hf : Handle NotImplementedError in convert-hf-to-gguf (#7660)
c8047d538f3addab40e3112be60bb92e70ce1a50 30e238b246f8002cc6eb7cb79afe242243f1f66d Johannes Gäßler johannesg@5d6.de 2024-05-31T16:26:21+02:00 GitHub noreply@github.com 2024-05-31T16:26:21+02:00 scripts: update compare_llama_bench.py [no ci] (#7673)
30e238b246f8002cc6eb7cb79afe242243f1f66d 16926dff92d6d0efa8cbc0f44d30d63349532b38 Daniele 57776841+daniandtheweb@users.noreply.github.com 2024-05-31T14:00:29Z GitHub noreply@github.com 2024-05-31T16:00:29+02:00 Improve HIP compatibility (#7672)
16926dff92d6d0efa8cbc0f44d30d63349532b38 0c27e6f62eea80140daf152d7b6c154466614e5c Georgi Gerganov ggerganov@gmail.com 2024-05-31T15:04:58+03:00 GitHub noreply@github.com 2024-05-31T15:04:58+03:00 readme : link homebrew discussion
0c27e6f62eea80140daf152d7b6c154466614e5c 2e32f874e675f7bc5307cb7b4470ddbe090bab8f Georgi Gerganov ggerganov@gmail.com 2024-05-31T14:17:10+03:00 GitHub noreply@github.com 2024-05-31T14:17:10+03:00 ggml : fix loongson compile warnings (#7537)
2e32f874e675f7bc5307cb7b4470ddbe090bab8f 1af511fc22cba4959dd8bced5501df9e8af6ddf9 Galunid karolek1231456@gmail.com 2024-05-31T10:24:41+02:00 GitHub noreply@github.com 2024-05-31T18:24:41+10:00 Somehow '**' got lost (#7663)
1af511fc22cba4959dd8bced5501df9e8af6ddf9 0541f06296753dbc59a57379eb54cec865a4c9f9 Galunid karolek1231456@gmail.com 2024-05-31T10:09:20+02:00 GitHub noreply@github.com 2024-05-31T10:09:20+02:00 Add convert.py removal to hot topics (#7662)
0541f06296753dbc59a57379eb54cec865a4c9f9 9022c33646fbf78da35f40c3f98576cc08c40ddf Sertaç Özercan 852750+sozercan@users.noreply.github.com 2024-05-30T16:57:16-07:00 GitHub noreply@github.com 2024-05-31T09:57:16+10:00 [no ci] docs: add aikit to readme (#7650)
9022c33646fbf78da35f40c3f98576cc08c40ddf 5921b8f089d3b7bda86aac5a66825df6a6c10603 JohnnyB jboero@users.noreply.github.com 2024-05-30T21:32:38+01:00 GitHub noreply@github.com 2024-05-30T22:32:38+02:00 Fixed painfully slow single process builds. (#7326)
5921b8f089d3b7bda86aac5a66825df6a6c10603 5dcdf946764fae49a8e2a90bf2f0960bde1c44e8 Georgi Gerganov ggerganov@gmail.com 2024-05-30T19:01:41+03:00 GitHub noreply@github.com 2024-05-31T02:01:41+10:00 llama : cache llama_token_to_piece (#7587)
5dcdf946764fae49a8e2a90bf2f0960bde1c44e8 2e2340de1740b07f2418c04792607387d4dc1442 Martin Delille martin@delille.org 2024-05-30T17:07:39+02:00 GitHub noreply@github.com 2024-05-31T01:07:39+10:00 Fix conan badge display [no ci] (#7645)
2e2340de1740b07f2418c04792607387d4dc1442 7846540bd291e52cd4eee53882315760e05239be Manuel 44313466+makuche@users.noreply.github.com 2024-05-30T16:58:15+02:00 GitHub noreply@github.com 2024-05-31T00:58:15+10:00 Add brew installation instruction to README [no ci] (#7616)
7846540bd291e52cd4eee53882315760e05239be e6157f94c8f835f7f774b98409078867472a34fe Martin Delille martin@delille.org 2024-05-30T14:52:50+02:00 GitHub noreply@github.com 2024-05-30T15:52:50+03:00 readme : add Conan badge (#7638)
e6157f94c8f835f7f774b98409078867472a34fe 9c4c9cc83f7297a10bb3b2af54a22ac154fd5b20 Brian mofosyne@gmail.com 2024-05-30T21:55:36+10:00 GitHub noreply@github.com 2024-05-30T21:55:36+10:00 github: add contact links to issues and convert question into research [no ci] (#7612)
9c4c9cc83f7297a10bb3b2af54a22ac154fd5b20 59b0d077662fab430446b3119fa142f3291c45b2 Galunid karolek1231456@gmail.com 2024-05-30T13:40:00+02:00 GitHub noreply@github.com 2024-05-30T21:40:00+10:00 Move convert.py to examples/convert-legacy-llama.py (#7430)
59b0d077662fab430446b3119fa142f3291c45b2 d5c05821f3c3d6cabe8ac45776fe0ecb0da13eca Chris Elrod elrodc@gmail.com 2024-05-30T07:32:55-04:00 GitHub noreply@github.com 2024-05-30T21:32:55+10:00 faster avx512 exp implementation (#7551)
d5c05821f3c3d6cabe8ac45776fe0ecb0da13eca 972b555ab935705f3437abd5909a5c46852811f6 junchao-loongson 68935141+junchao-loongson@users.noreply.github.com 2024-05-30T17:30:10+08:00 GitHub noreply@github.com 2024-05-30T12:30:10+03:00 ggml : fix loongarch build (O2 issue) (#7636)
972b555ab935705f3437abd5909a5c46852811f6 3854c9d07f67de7f8cd6d86117bfaef47549b05a Johannes Gäßler johannesg@5d6.de 2024-05-30T09:52:39+02:00 GitHub noreply@github.com 2024-05-30T09:52:39+02:00 README: explain parallel build [no ci] (#7618)
3854c9d07f67de7f8cd6d86117bfaef47549b05a eb57fee51f7b4d78039f003249873c2eb46f12f6 Meng, Hengyu hengyu.meng@intel.com 2024-05-30T14:19:08+08:00 GitHub noreply@github.com 2024-05-30T16:19:08+10:00 [SYCL] fix intel docker (#7630)
eb57fee51f7b4d78039f003249873c2eb46f12f6 55d62262a99cd8bc28a1492975791fe433c8cc0f Galunid karolek1231456@gmail.com 2024-05-30T02:10:40+02:00 GitHub noreply@github.com 2024-05-30T02:10:40+02:00 gguf-py : Add tokenizer.ggml.pre to gguf-new-metadata.py (#7627)
55d62262a99cd8bc28a1492975791fe433c8cc0f 975ec63ff26cdf96156d1126d86f75a395fdc43a Georgi Gerganov ggerganov@gmail.com 2024-05-29T22:20:40+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-29T22:21:20+03:00 metal : remove invalid asserts (#7617)
975ec63ff26cdf96156d1126d86f75a395fdc43a fb76ec31a9914b7761c1727303ab30380fd4f05c Georgi Gerganov ggerganov@gmail.com 2024-05-29T20:45:25+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-29T20:45:25+03:00 metal : add missing asserts (#7617)
fb76ec31a9914b7761c1727303ab30380fd4f05c cce3dcffc5695bd24835f04e6080070a2a119873 Georgi Gerganov ggerganov@gmail.com 2024-05-29T20:17:31+03:00 GitHub noreply@github.com 2024-05-29T20:17:31+03:00 ggml : fix YARN + add tests + add asserts (#7617)
cce3dcffc5695bd24835f04e6080070a2a119873 210d99173dc82aafb48f6e39d787c387951fe3a9 Georgi Gerganov ggerganov@gmail.com 2024-05-29T15:38:26+03:00 GitHub noreply@github.com 2024-05-29T15:38:26+03:00 cuda : non-cont concat support (#7610)
210d99173dc82aafb48f6e39d787c387951fe3a9 87bdf2a199acd62e19814d7a4d0500a04a7f09f3 Radoslav Gerganov rgerganov@gmail.com 2024-05-29T14:45:44+03:00 GitHub noreply@github.com 2024-05-29T14:45:44+03:00 llama-bench : add support for the RPC backend (#7435)
87bdf2a199acd62e19814d7a4d0500a04a7f09f3 00281b7be32462754618c42ed93f95743af46627 slaren slarengh@gmail.com 2024-05-29T13:36:39+02:00 GitHub noreply@github.com 2024-05-29T13:36:39+02:00 ggml : use atomic_flag for critical section (#7598)
00281b7be32462754618c42ed93f95743af46627 2ab977282b02ccd6783fbbaec393c96886cf33b1 Georgi Gerganov ggerganov@gmail.com 2024-05-29T14:31:18+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-29T14:31:18+03:00 scripts : remove mpi remnants
2ab977282b02ccd6783fbbaec393c96886cf33b1 72de268bec49f67e2883880f573c55cea32de736 Georgi Gerganov ggerganov@gmail.com 2024-05-29T14:29:52+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-29T14:29:52+03:00 sync : ggml
72de268bec49f67e2883880f573c55cea32de736 0e8d8bfd6caf1d0a8cbdf9d3d5c06fbbb9dfced8 Georgi Gerganov ggerganov@gmail.com 2024-05-26T18:35:23+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-29T14:29:33+03:00 ggml : restore ggml_rope_xpos_inplace (ggml/0)
0e8d8bfd6caf1d0a8cbdf9d3d5c06fbbb9dfced8 504f0c340f6b5e04de682f6ddefdd3b81208df5d Akarshan Biswas akarshanbiswas@fedoraproject.org 2024-05-29T12:23:47+05:30 GitHub noreply@github.com 2024-05-29T16:53:47+10:00 Add Arc A750 and Arch linux to readme-sycl.md as verified GPU model and Linux distro (#7605)
504f0c340f6b5e04de682f6ddefdd3b81208df5d b864b50ce5e2beefc8c2fd31733e4e1a978b7754 zhouwg zhouwg2000@gmail.com 2024-05-29T10:09:31+08:00 GitHub noreply@github.com 2024-05-29T04:09:31+02:00 ggml : fix typo in ggml.c (#7603)
b864b50ce5e2beefc8c2fd31733e4e1a978b7754 02c1ecad07f0e2d2febe8196271bcc64bdc9c006 Meng, Hengyu hengyu.meng@intel.com 2024-05-29T07:00:24+08:00 GitHub noreply@github.com 2024-05-29T07:00:24+08:00 [SYCL] Align GEMM dispatch (#7566)
02c1ecad07f0e2d2febe8196271bcc64bdc9c006 6bd12ce409f949012935b7d1b15a21ffa473a565 jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-05-28T21:46:34+02:00 GitHub noreply@github.com 2024-05-28T21:46:34+02:00 Tokenizer WPM fixes (#7500)
6bd12ce409f949012935b7d1b15a21ffa473a565 5442939fcc5e6ae41abf40612a95fd71377e487e Georgi Gerganov ggerganov@gmail.com 2024-05-28T22:22:50+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-28T22:22:50+03:00 sycl : fix assert (#7563)
5442939fcc5e6ae41abf40612a95fd71377e487e 56411a950f255b523a9edd684fd1632752474399 Giuseppe Scrivano giuseppe@scrivano.org 2024-05-28T20:49:49+02:00 GitHub noreply@github.com 2024-05-28T21:49:49+03:00 llama : support small Granite models (#7481)
56411a950f255b523a9edd684fd1632752474399 2b737caae100cf0ac963206984332e422058f2b9 k.h.lai adrian.k.h.lai@outlook.com 2024-05-29T01:25:08+08:00 GitHub noreply@github.com 2024-05-28T19:25:08+02:00 vulkan: properly initialize vulkan devices for LLAMA_SPLIT_MODE_NONE (#7552)
2b737caae100cf0ac963206984332e422058f2b9 ee3dff6b8e39bb8c1cdea1782a7b95ef0118f970 Radoslav Gerganov rgerganov@gmail.com 2024-05-28T18:13:36+03:00 GitHub noreply@github.com 2024-05-28T18:13:36+03:00 rpc : resource management rework (#7562)
ee3dff6b8e39bb8c1cdea1782a7b95ef0118f970 edc29433fa08b4e5aeb67649a29fc7713af13d04 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-05-28T17:07:05+02:00 GitHub noreply@github.com 2024-05-28T17:07:05+02:00 Add support for DeepseekV2ForCausalLM (#7519)
edc29433fa08b4e5aeb67649a29fc7713af13d04 8b99e2aa66ba39e4e1114effea6ef7430881eca4 Georgi Gerganov ggerganov@gmail.com 2024-05-28T15:04:09+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-28T15:04:09+03:00 tests : fix test-tokenizer-0.sh
8b99e2aa66ba39e4e1114effea6ef7430881eca4 271ff3fc44a6ecfcea3ebc192e67567d578b7772 Georgi Gerganov ggerganov@gmail.com 2024-05-28T13:55:35+03:00 GitHub noreply@github.com 2024-05-28T13:55:35+03:00 llama : handle unknown utf8 bytes (#7588)
271ff3fc44a6ecfcea3ebc192e67567d578b7772 e2b065071c5fc8ac5697d12ca343551faee465cc Brian mofosyne@gmail.com 2024-05-28T20:27:27+10:00 GitHub noreply@github.com 2024-05-28T20:27:27+10:00 github: add refactor to issue template (#7561)
e2b065071c5fc8ac5697d12ca343551faee465cc 0548a4187f2e53b8fc6d9ff0f4c71988f708ff42 Neo Zhang 14088817+arthw@users.noreply.github.com 2024-05-28T17:53:37+08:00 GitHub noreply@github.com 2024-05-28T10:53:37+01:00 [SYCL]fix ggml_sycl_mul_mat_id() to match the change of api (#7436)
0548a4187f2e53b8fc6d9ff0f4c71988f708ff42 9335b969e86a222e247adacedf814d8abfff8847 Georgi Gerganov ggerganov@gmail.com 2024-05-28T11:04:19+03:00 GitHub noreply@github.com 2024-05-28T11:04:19+03:00 ggml : generalize GGML_OP_CONCAT (#7563)
9335b969e86a222e247adacedf814d8abfff8847 c41767154eb82aa3fe7568fc816c3402b78eae94 mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2024-05-28T06:55:51+02:00 GitHub noreply@github.com 2024-05-28T14:55:51+10:00 server: do not remove whitespace at the start of a completion chunk (#7524)
c41767154eb82aa3fe7568fc816c3402b78eae94 74b239b3d5f067470d7ef5e26e2e059720572e32 Nathan Epstein nate2@umbc.edu 2024-05-28T00:41:14-04:00 GitHub noreply@github.com 2024-05-28T14:41:14+10:00 Markdownish code block fix (#7571)
74b239b3d5f067470d7ef5e26e2e059720572e32 852aafb163d32d5bad63c10bc323a02c28fec59d Ikko Eltociear Ashimine eltociear@gmail.com 2024-05-28T11:48:16+09:00 GitHub noreply@github.com 2024-05-28T12:48:16+10:00 llava : update clip.h (#7580)
852aafb163d32d5bad63c10bc323a02c28fec59d 0136966dafb452601c23f30395878d5a65ddc559 Djip007 djip.perois@free.fr 2024-05-28T01:40:47+02:00 GitHub noreply@github.com 2024-05-28T01:40:47+02:00 update HIP_UMA #7399 (#7414)
0136966dafb452601c23f30395878d5a65ddc559 10b1e4587670feba2c7730a645accf8234873113 kunnis kunnis@users.noreply.github.com 2024-05-27T18:40:12-05:00 GitHub noreply@github.com 2024-05-28T01:40:12+02:00 adding in x64 targets to cmake presets (#7574)
10b1e4587670feba2c7730a645accf8234873113 197c00681b80f9dea17d11a4436b6b8ef1be0ce8 Johannes Gäßler johannesg@5d6.de 2024-05-27T19:34:40+02:00 GitHub noreply@github.com 2024-05-27T19:34:40+02:00 make: add --device-debug to NVCC debug flags (#7542)
197c00681b80f9dea17d11a4436b6b8ef1be0ce8 95f84d5ce8b449a9b16009434aca800df504a02e agray3 agray3@users.noreply.github.com 2024-05-27T18:33:42+01:00 GitHub noreply@github.com 2024-05-27T19:33:42+02:00 Allow multiple copy function pointers for CUDA graph kernel param updates (#7565)
95f84d5ce8b449a9b16009434aca800df504a02e 5487593bc7ee0b65b9d2e2985b4b61dc77043101 AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-05-27T17:34:51+01:00 GitHub noreply@github.com 2024-05-27T22:04:51+05:30 Fix q_xxs using mul_mat_q (#7459)
5487593bc7ee0b65b9d2e2985b4b61dc77043101 1d8fca72ae9154eec0e1c0a75cfaac3c50f08e4a AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-05-27T13:34:09+01:00 GitHub noreply@github.com 2024-05-27T18:04:09+05:30 Add freq factors (#7495)
1d8fca72ae9154eec0e1c0a75cfaac3c50f08e4a 62bfef5194d5582486d62da3db59bf44981b7912 Georgi Gerganov ggerganov@gmail.com 2024-05-27T12:10:19+03:00 GitHub noreply@github.com 2024-05-27T12:10:19+03:00 metal : add GGML_OP_REPEAT kernels (#7557)
62bfef5194d5582486d62da3db59bf44981b7912 eaf6e031741ca2d3aafeff3e0f4dd7557a974d2b Georgi Gerganov ggerganov@gmail.com 2024-05-27T10:38:39+03:00 GitHub noreply@github.com 2024-05-27T10:38:39+03:00 metal : disable FA kernel for HS=256 (#7556)
eaf6e031741ca2d3aafeff3e0f4dd7557a974d2b d6ef0e77dd25f54fb5856af47e3926cf6f36c281 Georgi Gerganov ggerganov@gmail.com 2024-05-27T09:24:13+03:00 GitHub noreply@github.com 2024-05-27T09:24:13+03:00 llama : add comments about experimental flags (#7544)
d6ef0e77dd25f54fb5856af47e3926cf6f36c281 dff451cfa1f297348751ce6b538670e1ae9a7d5b Brian mofosyne@gmail.com 2024-05-27T10:54:30+10:00 GitHub noreply@github.com 2024-05-27T10:54:30+10:00 github: add self sorted issue ticket forms (#7543)
dff451cfa1f297348751ce6b538670e1ae9a7d5b d298382ad977ec89c8de7b57459b9d7965d2c272 Georgi Gerganov ggerganov@gmail.com 2024-05-26T18:54:56+03:00 GitHub noreply@github.com 2024-05-26T08:54:56-07:00 flake.lock: Update (#7540)
d298382ad977ec89c8de7b57459b9d7965d2c272 32a28217f475119926c603341e8273b26932b56a Brian mofosyne@gmail.com 2024-05-27T00:10:17+10:00 GitHub noreply@github.com 2024-05-27T00:10:17+10:00 main: replace --no-special with --special (#7534)
32a28217f475119926c603341e8273b26932b56a c429b33beb35f13934a4dfbe0c138d30b45e5d54 Galunid karolek1231456@gmail.com 2024-05-26T16:02:34+02:00 GitHub noreply@github.com 2024-05-26T16:02:34+02:00 Fix aya-23 conversion scripts (#7539)
c429b33beb35f13934a4dfbe0c138d30b45e5d54 9146d36fe7e3e911a07438c07efc1bae082f6390 Bartowski ckealty1182@gmail.com 2024-05-26T08:28:35-04:00 GitHub noreply@github.com 2024-05-26T15:28:35+03:00 llama : add Smaug 70B support (#7402)
9146d36fe7e3e911a07438c07efc1bae082f6390 b9adcbbf92fc7096bee23fe61496d25652ebf765 Aarni Koskela akx@iki.fi 2024-05-26T15:09:42+03:00 GitHub noreply@github.com 2024-05-26T22:09:42+10:00 Readme: add akx/ggify to tools (#1484)
b9adcbbf92fc7096bee23fe61496d25652ebf765 9588f196b1d7b21bdff013fcf958c249576b2619 HanishKVC hanishkvc@gmail.com 2024-05-26T06:26:34+05:30 GitHub noreply@github.com 2024-05-26T10:56:34+10:00 SimpleChat Completion Mode flexibility and cleanup, Settings gMe, Optional sliding window (#7480)
9588f196b1d7b21bdff013fcf958c249576b2619 3cbd23ed88c03a27e1eb6090ac4a8186ca9ac29a Georgi Gerganov ggerganov@gmail.com 2024-05-25T15:21:30+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-25T15:22:35+03:00 train : change default FA argument (#7528)
3cbd23ed88c03a27e1eb6090ac4a8186ca9ac29a 00c63907931bb08a0ed2b7e38cf44dd290143cb9 Brian mofosyne@gmail.com 2024-05-25T19:30:42+10:00 GitHub noreply@github.com 2024-05-25T19:30:42+10:00 labeler: added Apple Metal detector (+Kompute) (#7529)
00c63907931bb08a0ed2b7e38cf44dd290143cb9 faa0e6979a11dcb731e9d778ad42ceaa0302015e Justine Tunney jtunney@mozilla.com 2024-05-25T05:04:03-04:00 GitHub noreply@github.com 2024-05-25T19:04:03+10:00 main : don't print special tokens with --grammar (#6923)
faa0e6979a11dcb731e9d778ad42ceaa0302015e 9791f402580838d7f8543ae7bc633ef265e436f0 Masaya, Kato 62578291+msy-kato@users.noreply.github.com 2024-05-25T17:42:31+09:00 GitHub noreply@github.com 2024-05-25T11:42:31+03:00 ggml: aarch64: SVE kernels for q8_0_q8_0, q4_0_q8_0 vector dot (#7433)
9791f402580838d7f8543ae7bc633ef265e436f0 902184dd3a9d6685e752b19027a48423742531db Elton Kola eltonkola@gmail.com 2024-05-25T04:11:33-04:00 GitHub noreply@github.com 2024-05-25T11:11:33+03:00 android : module (#7502)
902184dd3a9d6685e752b19027a48423742531db 57684331fc2d685f7d1f5775af0b9e47d1829833 Xuan Son Nguyen thichthat@gmail.com 2024-05-25T05:30:59+02:00 GitHub noreply@github.com 2024-05-25T13:30:59+10:00 fix missing slash in `fs_get_cache_directory()` (#7503)
57684331fc2d685f7d1f5775af0b9e47d1829833 b83bab15a5d2a1e7807d09613a9b34309d86cfaa Mikko Juola mikjuo@gmail.com 2024-05-24T18:14:42-07:00 GitHub noreply@github.com 2024-05-25T11:14:42+10:00 Make tokenize CLI tool have nicer command line arguments. (#6188)
b83bab15a5d2a1e7807d09613a9b34309d86cfaa d041d2ceaaf50e058622d92921b3e680ffa4e9e7 compilade git@compilade.net 2024-05-24T21:11:48-04:00 GitHub noreply@github.com 2024-05-25T11:11:48+10:00 gguf-py : fix and simplify quantized shape round-trip (#7483)
d041d2ceaaf50e058622d92921b3e680ffa4e9e7 27891f6db03de6e3fd5941983838c29bef253352 Georgi Gerganov ggerganov@gmail.com 2024-05-24T18:59:06+03:00 GitHub noreply@github.com 2024-05-24T08:59:06-07:00 flake.lock: Update (#7232)
27891f6db03de6e3fd5941983838c29bef253352 fbca2f27fc7fa9aa4a8ad0357478fdb908472908 Brian mofosyne@gmail.com 2024-05-24T23:47:56+10:00 GitHub noreply@github.com 2024-05-24T23:47:56+10:00 docker.yml: disable light-intel and server-intel test (#7515)
fbca2f27fc7fa9aa4a8ad0357478fdb908472908 0df0aa8e43c3378975269a51f9b876c8692e70da fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-05-24T14:31:13+02:00 GitHub noreply@github.com 2024-05-24T14:31:13+02:00 Add support for ArcticForCausalLM (#7020)
0df0aa8e43c3378975269a51f9b876c8692e70da 74f33adf5f8b20b08fc5a6aa17ce081abe86ef2f Neo Zhang 14088817+arthw@users.noreply.github.com 2024-05-24T10:06:56+08:00 GitHub noreply@github.com 2024-05-24T10:06:56+08:00 add build shared lib in win release package (#7438)
74f33adf5f8b20b08fc5a6aa17ce081abe86ef2f 1debe72737ea131cb52975da3d53ed3a835df3a6 Georgi Gerganov ggerganov@gmail.com 2024-05-23T17:43:18+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-23T17:43:18+03:00 readme : remove trailing space (#7469)
1debe72737ea131cb52975da3d53ed3a835df3a6 007489e895bad02e4e54758bf0bdf2d6a4cdb7c1 Georgi Gerganov ggerganov@gmail.com 2024-05-23T17:17:43+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-23T17:25:38+03:00 ggml : silence UB sanitizer error during iq2_xxs quantization (#0)
007489e895bad02e4e54758bf0bdf2d6a4cdb7c1 8b94e799dfa482adf63419df4905dc79b37e179f Tristan Druyen tristan@vault81.mozmail.com 2024-05-23T16:15:15+02:00 GitHub noreply@github.com 2024-05-23T16:15:15+02:00 Fix phi3 chat template confusion with zephyr (#7449)
8b94e799dfa482adf63419df4905dc79b37e179f 3015851c5ac7334fb544a23a70a284c117b87044 Raj Hammeer Singh Hada hammeerraj@gmail.com 2024-05-23T18:00:13+05:30 GitHub noreply@github.com 2024-05-23T15:30:13+03:00 readme : add Bunny in supported models [no ci] (#7469)
3015851c5ac7334fb544a23a70a284c117b87044 55ac3b7aeaf52f19786ed96e885d89521fc0f6c8 Daniel Bevenius daniel.bevenius@gmail.com 2024-05-23T14:29:26+02:00 GitHub noreply@github.com 2024-05-23T15:29:26+03:00 llama : add getters for n_threads/n_threads_batch (#7464)
55ac3b7aeaf52f19786ed96e885d89521fc0f6c8 dacfcebd6022175848e978f82811a244f1033038 Georgi Gerganov ggerganov@gmail.com 2024-05-23T15:28:14+03:00 GitHub noreply@github.com 2024-05-23T15:28:14+03:00 ci : use Pythia models instead of OpenLlama (#7470)
dacfcebd6022175848e978f82811a244f1033038 9b82476ee9e73065a759f8bcc4cf27ec7ab2ed8c Victor Nogueira felladrin@gmail.com 2024-05-23T15:12:43+03:00 GitHub noreply@github.com 2024-05-23T15:12:43+03:00 readme : add GPT-NeoX + Pythia to the list of supported models (#7491)
9b82476ee9e73065a759f8bcc4cf27ec7ab2ed8c a61a94e543e3c6877c087e80fca27a0313ce5fd5 fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-05-23T11:49:53+02:00 GitHub noreply@github.com 2024-05-23T11:49:53+02:00 Add missing inference support for GPTNeoXForCausalLM (Pythia and GPT-NeoX base models) (#7461)
a61a94e543e3c6877c087e80fca27a0313ce5fd5 152da28ae54139e3754189b9e6e1c28e11277502 Georgi Gerganov ggerganov@gmail.com 2024-05-23T12:38:18+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-23T12:38:18+03:00 llama : rename n_ctx -> cache.size, less confusing (#0)
152da28ae54139e3754189b9e6e1c28e11277502 d48c88cbd563b6cf0ce972e2f56796896e240736 Brian mofosyne@gmail.com 2024-05-23T17:40:43+10:00 GitHub noreply@github.com 2024-05-23T17:40:43+10:00 labeler.yml: add embedding label detector [no ci] (#7482)
d48c88cbd563b6cf0ce972e2f56796896e240736 e84b71c2c6da6e69c8f815168ea836f9716a325e Georgi Gerganov ggerganov@gmail.com 2024-05-23T10:00:44+03:00 GitHub noreply@github.com 2024-05-23T10:00:44+03:00 ggml : remove ggml_flash_attn and ggml_flash_ff (#7463)
e84b71c2c6da6e69c8f815168ea836f9716a325e 1b1e27cb49158123ef4902aa41eb368c9e76e6a1 Georgi Gerganov ggerganov@gmail.com 2024-05-23T10:00:21+03:00 GitHub noreply@github.com 2024-05-23T10:00:21+03:00 ggml : drop support for QK_K=64 (#7473)
1b1e27cb49158123ef4902aa41eb368c9e76e6a1 fbf777d2b9c30e7569e3d1c149501c1e31d9b5b9 0cc4m picard12@live.de 2024-05-23T08:59:59+02:00 GitHub noreply@github.com 2024-05-23T08:59:59+02:00 Update vulkan rope implementation to support frequency factors (#7475)
fbf777d2b9c30e7569e3d1c149501c1e31d9b5b9 cd93a28cb1446319af5e2f4b416174c3a8e43546 Georgi Gerganov ggerganov@gmail.com 2024-05-23T09:43:24+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-23T09:43:49+03:00 main : minor (#7462)
cd93a28cb1446319af5e2f4b416174c3a8e43546 1e374365d170b7f692fd7753c145e21bc14486c8 Johannes Gäßler johannesg@5d6.de 2024-05-23T00:31:20+02:00 GitHub noreply@github.com 2024-05-23T00:31:20+02:00 CUDA: fix FA out-of-bounds reads (#7479)
1e374365d170b7f692fd7753c145e21bc14486c8 197ff91462dd05bb9a3be03578114abf0c355536 HanishKVC hanishkvc@gmail.com 2024-05-22T23:23:21+05:30 GitHub noreply@github.com 2024-05-23T03:53:21+10:00 SimpleChat: a simple and dumb web front end for testing /chat/completions and /completions end points and try chat (#7350)
197ff91462dd05bb9a3be03578114abf0c355536 6ff13987ad1a9519bee13dd98b6a21cd98979aab Georgi Gerganov ggerganov@gmail.com 2024-05-22T20:05:38+03:00 GitHub noreply@github.com 2024-05-22T20:05:38+03:00 build : remove zig (#7471)
6ff13987ad1a9519bee13dd98b6a21cd98979aab 38c03478a37e460ecd3a21155b338a83bfed7f90 Georgi Gerganov ggerganov@gmail.com 2024-05-22T20:04:20+03:00 GitHub noreply@github.com 2024-05-22T20:04:20+03:00 common : normalize naming style (#7462)
38c03478a37e460ecd3a21155b338a83bfed7f90 b18532a4efeca8796fea8e36195c81cbfd596a4a Johannes Gäßler johannesg@5d6.de 2024-05-22T17:58:25+02:00 GitHub noreply@github.com 2024-05-22T17:58:25+02:00 CUDA: fix FA out-of-bounds writes (#7465)
b18532a4efeca8796fea8e36195c81cbfd596a4a fcda1128bc5f8eb7e1811708fe9d9867b9aec815 slaren slarengh@gmail.com 2024-05-22T16:10:46+02:00 GitHub noreply@github.com 2024-05-22T16:10:46+02:00 phi3 : duplicate rope factors in each layer (#7447)
fcda1128bc5f8eb7e1811708fe9d9867b9aec815 03d8900ebe062355e26a562379daee5f17ea099f k.h.lai adrian.k.h.lai@outlook.com 2024-05-22T20:53:21+08:00 GitHub noreply@github.com 2024-05-22T14:53:21+02:00 vulkan: add workaround for iterator boundary check to fix clang-cl debug build (#7426)
03d8900ebe062355e26a562379daee5f17ea099f 9b3d83318931aa98c487baaa977626931d059e6a Justine Tunney jtunney@mozilla.com 2024-05-22T07:08:18-04:00 GitHub noreply@github.com 2024-05-22T14:08:18+03:00 llama : add missing model type names (#7445)
9b3d83318931aa98c487baaa977626931d059e6a 95fb0aefab568348da159efdd370e064d1b35f97 Georgi Gerganov ggerganov@gmail.com 2024-05-22T12:36:37+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-22T12:36:37+03:00 cuda : fix compile warning (#7454)
95fb0aefab568348da159efdd370e064d1b35f97 3e5faa85032ec3106a2ad831bf412be9ff139f47 Johannes Gäßler johannesg@5d6.de 2024-05-22T10:24:29+02:00 GitHub noreply@github.com 2024-05-22T10:24:29+02:00 CUDA: remove incorrect precision check (#7454)
3e5faa85032ec3106a2ad831bf412be9ff139f47 201cc11afa0a1950e1f632390b2ac6c937a0d8f0 Georgi Gerganov ggerganov@gmail.com 2024-05-22T11:01:35+03:00 GitHub noreply@github.com 2024-05-22T11:01:35+03:00 cuda : fix rope + add tests (#7452)
201cc11afa0a1950e1f632390b2ac6c937a0d8f0 6369bf04336ab60e5c892dd77a3246df91015147 liuwei-git 14815172+liuwei-git@users.noreply.github.com 2024-05-22T04:28:32+08:00 GitHub noreply@github.com 2024-05-21T23:28:32+03:00 llama : add phi3 128K model support (#7225)
6369bf04336ab60e5c892dd77a3246df91015147 e402de364b643cb89ea9f43057733b5d36298670 Georgi Gerganov ggerganov@gmail.com 2024-05-21T23:03:42+03:00 GitHub noreply@github.com 2024-05-21T23:03:42+03:00 metal : handle F16 inf values, fix FA partial offload (#7434)
e402de364b643cb89ea9f43057733b5d36298670 fcf6538ba6702c55eaec70da9a75c81d04900a72 Olivier Chafik ochafik@users.noreply.github.com 2024-05-21T20:40:00+01:00 GitHub noreply@github.com 2024-05-21T20:40:00+01:00 `grammars`: fix resampling logic regression (#7424)
fcf6538ba6702c55eaec70da9a75c81d04900a72 c3f8d583560b4f261fa21c976793e538c60cd66c Johannes Gäßler johannesg@5d6.de 2024-05-21T19:27:12+02:00 GitHub noreply@github.com 2024-05-21T20:27:12+03:00 CUDA: fix unused warning in mmq.cu (#7442)
c3f8d583560b4f261fa21c976793e538c60cd66c 11474e756de3f56b760986e73086d40e787e52f8 Georgi Gerganov ggerganov@gmail.com 2024-05-21T19:53:48+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-21T19:53:48+03:00 tests : test-tokenizer-0.sh print more info (#7402)
11474e756de3f56b760986e73086d40e787e52f8 d8ee90222791afff2ab666ded4cb6195fd94cced Amir amir_zia@outlook.com 2024-05-21T17:13:12+03:00 GitHub noreply@github.com 2024-05-21T17:13:12+03:00 examples: cache hf model when --model not provided (#7353)
d8ee90222791afff2ab666ded4cb6195fd94cced d7e852c1bc8e85bf62a6f1aede08cd2de723404a Johannes Gäßler johannesg@5d6.de 2024-05-21T16:02:12+02:00 GitHub noreply@github.com 2024-05-21T16:02:12+02:00 CUDA: deduplicate mmq code (#7397)
d7e852c1bc8e85bf62a6f1aede08cd2de723404a 917dc8cfa67a72fb7c8bf7392270da3bf4833af4 jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-05-21T14:39:48+02:00 GitHub noreply@github.com 2024-05-21T14:39:48+02:00 Tokenizer SPM fixes for phi-3 and llama-spm (bugfix) (#7425)
917dc8cfa67a72fb7c8bf7392270da3bf4833af4 fabf30b4c4fca32e116009527180c252919ca922 jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-05-20T20:15:57+02:00 GitHub noreply@github.com 2024-05-20T20:15:57+02:00 Tokenizer SPM fixes for phi-3 and llama-spm (#7375)
fabf30b4c4fca32e116009527180c252919ca922 20385cebcc4bb3f6dd10f989573c11864d70d53d Georgi Gerganov ggerganov@gmail.com 2024-05-20T19:35:28+03:00 GitHub noreply@github.com 2024-05-21T02:35:28+10:00 llama : remove Persimmon (#7408)
20385cebcc4bb3f6dd10f989573c11864d70d53d db10f01310beea8a1ef7798651b9d692fd1149d0 Johannes Gäßler johannesg@5d6.de 2024-05-20T18:15:38+02:00 GitHub noreply@github.com 2024-05-20T18:15:38+02:00 perplexity: update README FP16 results [no ci] (#7413)
db10f01310beea8a1ef7798651b9d692fd1149d0 3bc10cb485dd7efa4da6c64e73ad0c9e2bfe0821 Radoslav Gerganov rgerganov@gmail.com 2024-05-20T16:36:55+03:00 GitHub noreply@github.com 2024-05-20T16:36:55+03:00 rpc : track allocated buffers (#7411)
3bc10cb485dd7efa4da6c64e73ad0c9e2bfe0821 6bf9b66fa3f263ca2175dcb5f6d0a658581e1dfb Georgi Gerganov ggerganov@gmail.com 2024-05-20T15:10:03+03:00 GitHub noreply@github.com 2024-05-20T22:10:03+10:00 server : fix temperature + disable some tests (#7409)
6bf9b66fa3f263ca2175dcb5f6d0a658581e1dfb 26cd4237bc499f7144d76f440aa775d749b170bb AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-05-20T12:08:23+01:00 GitHub noreply@github.com 2024-05-20T16:38:23+05:30 [SYCL] Update SYCL upscale operation (#7321)
26cd4237bc499f7144d76f440aa775d749b170bb 213e90ed73f8ac3cd3026dc3f086beae0d414f96 Bingan 70050083+binganao@users.noreply.github.com 2024-05-20T17:55:34+08:00 GitHub noreply@github.com 2024-05-20T11:55:34+02:00 Update README.md (#7410)
213e90ed73f8ac3cd3026dc3f086beae0d414f96 65c58207ece92ad213f4bfd0f91dcb2dfb664f5b Herman Semenov GermanAizek@yandex.ru 2024-05-20T07:33:21Z GitHub noreply@github.com 2024-05-20T10:33:21+03:00 ggml-opencl, llama: using reserve() if count already known (#7272)
65c58207ece92ad213f4bfd0f91dcb2dfb664f5b 1cc0155d04918cb3017afa472acea51b77483c4a junchao-loongson 68935141+junchao-loongson@users.noreply.github.com 2024-05-20T15:19:21+08:00 GitHub noreply@github.com 2024-05-20T10:19:21+03:00 ggml : add loongarch lsx and lasx support (#6454)
1cc0155d04918cb3017afa472acea51b77483c4a e932094d58f513d5996c3efc9f6fed8238894c57 Georgi Gerganov ggerganov@gmail.com 2024-05-20T10:16:41+03:00 GitHub noreply@github.com 2024-05-20T10:16:41+03:00 server : tuning tests (#7388)
e932094d58f513d5996c3efc9f6fed8238894c57 2789baf480ba7dc9281b65f601f0918e58920f54 Georgi Gerganov ggerganov@gmail.com 2024-05-20T08:56:05+03:00 GitHub noreply@github.com 2024-05-20T08:56:05+03:00 server : return error on too large embedding input (#7389)
2789baf480ba7dc9281b65f601f0918e58920f54 33c8d50accd6dca73c9c4af00a05e24209c160fe Georgi Gerganov ggerganov@gmail.com 2024-05-20T08:55:09+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-20T08:55:09+03:00 tests : fix --keep_split -> --keep-split (#7374)
33c8d50accd6dca73c9c4af00a05e24209c160fe d359f30921a9f62a0fd299c412ff3f270286fea6 Srihari-mcw 96763064+Srihari-mcw@users.noreply.github.com 2024-05-19T19:18:39-07:00 GitHub noreply@github.com 2024-05-20T12:18:39+10:00 Add provisions for windows support for BF16 code including CMake provision for enabling AVX512_BF16 (#7258)
d359f30921a9f62a0fd299c412ff3f270286fea6 1ea2a0036e88172d6c8bf7e1a1989a03894dc955 slaren slarengh@gmail.com 2024-05-20T01:17:03+02:00 GitHub noreply@github.com 2024-05-20T01:17:03+02:00 llama : remove MPI backend (#7395)
1ea2a0036e88172d6c8bf7e1a1989a03894dc955 f030ec1f7a72aa825b2104823946551b9ec5dfc1 Fred Douglas 43351173+fredlas@users.noreply.github.com 2024-05-19T11:37:04-05:00 GitHub noreply@github.com 2024-05-19T19:37:04+03:00 quantize : fix --keep-split check (#7374)
f030ec1f7a72aa825b2104823946551b9ec5dfc1 e4e6f67be6a8a697f5f89a28c98934e53c99c359 0cc4m picard12@live.de 2024-05-19T17:19:53+02:00 GitHub noreply@github.com 2024-05-19T17:19:53+02:00 Vulkan Embedding Fix (#7360)
e4e6f67be6a8a697f5f89a28c98934e53c99c359 5ca49cbecda27ce0a7266658fc3b640bff3ed386 slaren slarengh@gmail.com 2024-05-19T17:08:46+02:00 GitHub noreply@github.com 2024-05-19T17:08:46+02:00 ggml : fix another case of quants nans (#7387)
5ca49cbecda27ce0a7266658fc3b640bff3ed386 1b01f06db0cff5f5f600bb754fc39fde565ed56a Johannes Gäßler johannesg@5d6.de 2024-05-19T16:46:13+02:00 GitHub noreply@github.com 2024-05-19T16:46:13+02:00 ggml: implement quantized KV cache for FA (#7372)
1b01f06db0cff5f5f600bb754fc39fde565ed56a 41858392e17abead21735309bf17cb55183d8c31 Johannes Gäßler johannesg@5d6.de 2024-05-19T16:26:02+02:00 GitHub noreply@github.com 2024-05-19T16:26:02+02:00 server: add test for token probs (#7347)
41858392e17abead21735309bf17cb55183d8c31 6aade19ee74b896c59929676629340b36be3e22c Johannes Gäßler johannesg@5d6.de 2024-05-19T16:06:33+02:00 GitHub noreply@github.com 2024-05-19T17:06:33+03:00 server: fix seed being reported back (#7382)
6aade19ee74b896c59929676629340b36be3e22c ab33f7a338593f6cf1ae98b10b6f8684f63bd72c Anas Ahouzi 112881240+aahouzi@users.noreply.github.com 2024-05-19T14:46:46+02:00 GitHub noreply@github.com 2024-05-19T22:46:46+10:00 Add StableLM2 pre-tokenizer (#7349)
ab33f7a338593f6cf1ae98b10b6f8684f63bd72c e23b974f4cf9270d05062d446f406e3ff55d9451 slaren slarengh@gmail.com 2024-05-19T14:19:37+02:00 GitHub noreply@github.com 2024-05-19T14:19:37+02:00 cuda : clear error after buffer allocation failure (#7376)
e23b974f4cf9270d05062d446f406e3ff55d9451 854d365abab7194b5013a523f72da19860c3c550 Brian mofosyne@gmail.com 2024-05-19T20:51:03+10:00 GitHub noreply@github.com 2024-05-19T20:51:03+10:00 labeler.yml: Use settings from ggerganov/llama.cpp [no ci] (#7363)
854d365abab7194b5013a523f72da19860c3c550 f5bf761747988ee1832766f7d1433739aff810da Georgi Gerganov ggerganov@gmail.com 2024-05-19T11:01:01+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-19T11:01:01+03:00 cmake : update android comments (#7341)
f5bf761747988ee1832766f7d1433739aff810da 059031b8c40e1f4ba60586842c5b1ed3ddf61842 fraxy-v 65565042+fraxy-v@users.noreply.github.com 2024-05-19T01:44:42+03:00 GitHub noreply@github.com 2024-05-19T00:44:42+02:00 Capture CUDA logging output (#7298)
059031b8c40e1f4ba60586842c5b1ed3ddf61842 511182eabb36f6ec9776e2b3c4d7e16d93d0ac0d Georgi Gerganov ggerganov@gmail.com 2024-05-18T18:55:54+03:00 GitHub noreply@github.com 2024-05-18T18:55:54+03:00 ci : re-enable sanitizer runs (#7358)
511182eabb36f6ec9776e2b3c4d7e16d93d0ac0d 133d99c59980139f5bb75922c8b5fca67d7ba9b8 Georgi Gerganov ggerganov@gmail.com 2024-05-18T13:40:39+03:00 GitHub noreply@github.com 2024-05-18T20:40:39+10:00 android : use "ci-android" branch for CI (#7341)
133d99c59980139f5bb75922c8b5fca67d7ba9b8 cb42c294279bc4a0a4e926a7b5a5568049f12fa7 Johannes Gäßler johannesg@5d6.de 2024-05-18T12:36:25+02:00 GitHub noreply@github.com 2024-05-18T12:36:25+02:00 CUDA: deduplicate FlashAttention code (#7352)
cb42c294279bc4a0a4e926a7b5a5568049f12fa7 d233b507cd19fcc2d8d8963ecc6a3eb7a33f2ecc Johannes Gäßler johannesg@5d6.de 2024-05-18T11:10:47+02:00 GitHub noreply@github.com 2024-05-18T11:10:47+02:00 server: correct --threads documentation [no ci] (#7362)
d233b507cd19fcc2d8d8963ecc6a3eb7a33f2ecc 0f98acfac6cc561dc57586bfff778405e42b576b Engininja2 139037756+Engininja2@users.noreply.github.com 2024-05-18T02:05:17-06:00 GitHub noreply@github.com 2024-05-18T10:05:17+02:00 cuda : add half2 __shfl_xor() for ROCm 5.5 (#7263)
0f98acfac6cc561dc57586bfff778405e42b576b ca57e0f35e33f714b9a6c2c4482b87bfe059c819 Steffen Röcker sroecker@gmail.com 2024-05-18T10:04:55+02:00 GitHub noreply@github.com 2024-05-18T11:04:55+03:00 llama : add support for larger Granite Code Models (20B, 34B) (#7324)
ca57e0f35e33f714b9a6c2c4482b87bfe059c819 c1b295eea5c49887a066559527a74e8b94fe9db0 strawberrymelonpanda 152940198+strawberrymelonpanda@users.noreply.github.com 2024-05-18T00:57:08-07:00 GitHub noreply@github.com 2024-05-18T10:57:08+03:00 perplexity : ndot progress and show stats with < 100 tasks (#7348)
c1b295eea5c49887a066559527a74e8b94fe9db0 de731963441ff128248259e1b99573d75264d210 0cc4m picard12@live.de 2024-05-18T08:10:58+02:00 GitHub noreply@github.com 2024-05-18T08:10:58+02:00 Update and fix Vulkan soft_max and argsort implementations (#7237)
de731963441ff128248259e1b99573d75264d210 b49a13dd2fa9c94c2c19a8c248bb7fa45499f9a8 Brian mofosyne@gmail.com 2024-05-18T16:04:23+10:00 GitHub noreply@github.com 2024-05-18T16:04:23+10:00 github-actions-labeler: initial commit (#7330)
b49a13dd2fa9c94c2c19a8c248bb7fa45499f9a8 05834841dcb4f922983ea976539c70472272df9a Georgi Gerganov ggerganov@gmail.com 2024-05-18T08:46:20+03:00 GitHub noreply@github.com 2024-05-18T08:46:20+03:00 convert : fix set_vocab_sentencepiece (#6866)
05834841dcb4f922983ea976539c70472272df9a ef277de2add255a08b2b909ebfbf70364d1f4dc4 slaren slarengh@gmail.com 2024-05-18T02:39:54+02:00 GitHub noreply@github.com 2024-05-18T02:39:54+02:00 ggml : fix quants nans when all the group weights are very close to zero (#7313)
ef277de2add255a08b2b909ebfbf70364d1f4dc4 b43272afa29a64dcb8bcf26a96a05bac40792b92 Engininja2 139037756+Engininja2@users.noreply.github.com 2024-05-17T18:39:25-06:00 GitHub noreply@github.com 2024-05-18T02:39:25+02:00 cmake : fix typo in AMDGPU_TARGETS (#7356)
b43272afa29a64dcb8bcf26a96a05bac40792b92 0fc1e820a9900a3dd08ddd3c6abe6604c53b689b jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-05-18T01:09:13+02:00 GitHub noreply@github.com 2024-05-18T01:09:13+02:00 Unicode codepoint flags for custom regexs (#7245)
0fc1e820a9900a3dd08ddd3c6abe6604c53b689b 82ca83db3c8d45df559c03a4225b6eb34808a2db Johannes Gäßler johannesg@5d6.de 2024-05-17T18:54:52+02:00 GitHub noreply@github.com 2024-05-17T18:54:52+02:00 CUDA: faster large batch FA without tensor cores (#7314)
82ca83db3c8d45df559c03a4225b6eb34808a2db f4bd8b3d260bb09491ba63c77ab7012b744362ef Gavin Zhao gavinzhaojw@protonmail.com 2024-05-17T11:03:03-04:00 GitHub noreply@github.com 2024-05-17T17:03:03+02:00 ROCm: use native CMake HIP support (#5966)
f4bd8b3d260bb09491ba63c77ab7012b744362ef 51e9d02599336e62948d29f1d6c05addeb921ac2 Radoslav Gerganov rgerganov@gmail.com 2024-05-17T17:25:44+03:00 GitHub noreply@github.com 2024-05-17T17:25:44+03:00 rpc : set SO_REUSEADDR for the server socket (#7320)
51e9d02599336e62948d29f1d6c05addeb921ac2 d273c1402b25086fd91aef2467ac13f2e49fa0ea Brian mofosyne@gmail.com 2024-05-17T22:40:14+10:00 GitHub noreply@github.com 2024-05-17T22:40:14+10:00 Added a single test function script and fix debug-test.sh to be more robust (#7279)
d273c1402b25086fd91aef2467ac13f2e49fa0ea 27b040691cbe45314147c2745e891a38e9c048d4 Aarni Koskela akx@iki.fi 2024-05-17T15:11:45+03:00 GitHub noreply@github.com 2024-05-17T15:11:45+03:00 py : convert-hf-to-gguf-update improvements (#7340)
27b040691cbe45314147c2745e891a38e9c048d4 29c60d8cddcfd14fa8a6bf023a6c4eb8692c76ba fairydreaming 166155368+fairydreaming@users.noreply.github.com 2024-05-17T13:24:38+02:00 GitHub noreply@github.com 2024-05-17T14:24:38+03:00 llama : use n_embd_head_v when reshaping kqv (#7327)
29c60d8cddcfd14fa8a6bf023a6c4eb8692c76ba 359cbe3f46c90ce6f5151005e411b8fb74f8139e Johannes Gäßler johannesg@5d6.de 2024-05-17T09:59:57+02:00 GitHub noreply@github.com 2024-05-17T09:59:57+02:00 tokenization: add warning for double BOS (#7332)
359cbe3f46c90ce6f5151005e411b8fb74f8139e e18bc6aaf3b547890609ed254ee5248e720e5840 Herman Semenov GermanAizek@yandex.ru 2024-05-17T07:08:49Z GitHub noreply@github.com 2024-05-17T10:08:49+03:00 ggml-quants, llama : removed excess checks (#7274)
e18bc6aaf3b547890609ed254ee5248e720e5840 ee94172d33399d2e814ca05c8a3ff8c523ebb093 amd-lalithnc lalithnc@amd.com 2024-05-17T12:31:58+05:30 GitHub noreply@github.com 2024-05-17T10:01:58+03:00 convert : fix Qwen/Qwen-7b conversion (#7308)
ee94172d33399d2e814ca05c8a3ff8c523ebb093 934266c0e0b2aa9781fdba2deb112c161ff038a9 Radoslav Gerganov rgerganov@gmail.com 2024-05-17T10:00:17+03:00 GitHub noreply@github.com 2024-05-17T10:00:17+03:00 server : add support for the RPC backend (#7305)
934266c0e0b2aa9781fdba2deb112c161ff038a9 9c4fdcbec8c7fcc428e723b0d8a1cf1f351ba642 Justine Tunney jtunney@mozilla.com 2024-05-17T02:58:52-04:00 GitHub noreply@github.com 2024-05-17T09:58:52+03:00 ggml : rewrite silu and softmax for cpu (#7154)
9c4fdcbec8c7fcc428e723b0d8a1cf1f351ba642 24ecb58168dce81646c2ed425690a106591c8c6d Leon Knauer git@leonknauer.com 2024-05-17T02:11:03+02:00 GitHub noreply@github.com 2024-05-17T10:11:03+10:00 [Server] Added --verbose option to README [no ci] (#7335)
24ecb58168dce81646c2ed425690a106591c8c6d 9afdffe70ebf3166d429b4434783bb0b7f97bdeb Pierrick Hymbert pierrick.hymbert@gmail.com 2024-05-16T20:43:45+02:00 GitHub noreply@github.com 2024-05-16T20:43:45+02:00 Revert "server bench: fix bench not waiting for model load (#7284)" (#7334)
9afdffe70ebf3166d429b4434783bb0b7f97bdeb 3b3963c55c8332e33533c44b2aa882b0e45f8292 Radoslav Gerganov rgerganov@gmail.com 2024-05-15T16:04:40+03:00 Radoslav Gerganov rgerganov@gmail.com 2024-05-16T12:04:08+03:00 rpc : get available mem for the CPU backend
3b3963c55c8332e33533c44b2aa882b0e45f8292 dda64fc17c97820ea9489eb0cc9ae8b8fdce4926 Radoslav Gerganov rgerganov@gmail.com 2024-05-15T15:29:07+03:00 Radoslav Gerganov rgerganov@gmail.com 2024-05-16T09:58:29+03:00 rpc : add command line arg for specifying backend memory
dda64fc17c97820ea9489eb0cc9ae8b8fdce4926 0350f5815218c483fb3026a86adc44a115481625 Jared Van Bortel jared@nomic.ai 2024-05-16T02:15:23-04:00 GitHub noreply@github.com 2024-05-16T16:15:23+10:00 convert : get general.name from model dir, not its parent (#5615)
0350f5815218c483fb3026a86adc44a115481625 ad52d5c259344888b06fd5acd3344c663dd0621d Herman Semenov GermanAizek@yandex.ru 2024-05-16T06:14:24Z GitHub noreply@github.com 2024-05-16T16:14:24+10:00 grammar, json, llama: replace push on emplace if it possible (#7273)
ad52d5c259344888b06fd5acd3344c663dd0621d 172b78210aae0e54d3668c5de14200efab9fac23 Vaibhav Srivastav vaibhavs10@gmail.com 2024-05-16T07:38:43+02:00 GitHub noreply@github.com 2024-05-16T15:38:43+10:00 doc: add references to hugging face GGUF-my-repo quantisation web tool. (#7288)
172b78210aae0e54d3668c5de14200efab9fac23 13ad16af1231ab2d245d35df3295bcfa23de1305 Max Krasnyansky quic_maxk@quicinc.com 2024-05-15T22:36:43-07:00 GitHub noreply@github.com 2024-05-16T15:36:43+10:00 ci: fix bin/Release path for windows-arm64 builds (#7317)
13ad16af1231ab2d245d35df3295bcfa23de1305 8f7080bf48828b538bc9387c3d150bbd4fb4cf2d Max Krasnyansky max.krasnyansky@gmail.com 2024-05-15T19:47:36-07:00 GitHub noreply@github.com 2024-05-16T12:47:36+10:00 Add support for properly optimized Windows ARM64 builds with LLVM and MSVC (#7191)
8f7080bf48828b538bc9387c3d150bbd4fb4cf2d e1b40ac3b94824d761b5e26ea1bc5692706029d9 Daniel Bevenius daniel.bevenius@gmail.com 2024-05-15T23:41:03+02:00 GitHub noreply@github.com 2024-05-15T23:41:03+02:00 readme : remove stray double quote (#7310)
e1b40ac3b94824d761b5e26ea1bc5692706029d9 dc020985b8755dd6aa93a2f002f43c3ede808cce kunnis kunnis@users.noreply.github.com 2024-05-15T12:59:12-05:00 GitHub noreply@github.com 2024-05-15T19:59:12+02:00 ggml : use dynamic thread scheduling for matrix multiplication (#6915)
dc020985b8755dd6aa93a2f002f43c3ede808cce 344f9126cc0d15891fde9472fe40b8572628ad7d agray3 agray3@users.noreply.github.com 2024-05-15T14:44:49+01:00 GitHub noreply@github.com 2024-05-15T15:44:49+02:00 Avoid unnecessarily disabling CUDA graphs (#7302)
344f9126cc0d15891fde9472fe40b8572628ad7d 9a17ab914b0aa7353389c656a3f2a0f086726868 slaren slarengh@gmail.com 2024-05-15T15:08:48+02:00 GitHub noreply@github.com 2024-05-15T15:08:48+02:00 ggml : tag ggml_tensor::backend as deprecated (#7290)
9a17ab914b0aa7353389c656a3f2a0f086726868 ea3b0590ee33d3573eb8ef76f88cc60f36d2a38d AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-05-15T13:26:30+01:00 GitHub noreply@github.com 2024-05-15T17:56:30+05:30 Add missing " (#7303)
ea3b0590ee33d3573eb8ef76f88cc60f36d2a38d 29499bb59383c2a8c5d557a90abb08b696cef7f6 dm4 sunrisedm4@gmail.com 2024-05-15T20:01:12+08:00 GitHub noreply@github.com 2024-05-15T15:01:12+03:00 embedding : free the batch after execution (#7297)
29499bb59383c2a8c5d557a90abb08b696cef7f6 48aa8fd1f213a69b41569f809cc954f24dbc4366 Georgi Gerganov ggerganov@gmail.com 2024-05-15T13:23:41+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-15T13:23:41+03:00 sync : ggml
48aa8fd1f213a69b41569f809cc954f24dbc4366 583fd6b000ec9ad1b465b5c98524f4a0ae388077 John Balis phobossystems@gmail.com 2024-05-15T03:52:33-05:00 Georgi Gerganov ggerganov@gmail.com 2024-05-15T13:23:33+03:00 ggml : add `ggml_upscale_ext` (ggml/814)
583fd6b000ec9ad1b465b5c98524f4a0ae388077 9f773486ab78d65f5cca3f7e31c862b7043bf721 Johannes Gäßler johannesg@5d6.de 2024-05-15T08:44:16+02:00 GitHub noreply@github.com 2024-05-15T08:44:16+02:00 server bench: fix bench not waiting for model load (#7284)
9f773486ab78d65f5cca3f7e31c862b7043bf721 e8a7fd4fb06d82f663850c21fcf86c0fb98ad9b4 Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:14:38+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:14:38+03:00 script : sync ggml-rpc
e8a7fd4fb06d82f663850c21fcf86c0fb98ad9b4 a5e3fde8578d54b98d941344a4da150669af200d Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:09:30+03:00 GitHub noreply@github.com 2024-05-14T19:09:30+03:00 metal : support FA without mask + add asserts (#7278)
a5e3fde8578d54b98d941344a4da150669af200d f308ea705974dff62a1fe5367d776ad9d5109239 Georgi Gerganov ggerganov@gmail.com 2024-05-14T15:33:16+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:08:09+03:00 sync : ggml
f308ea705974dff62a1fe5367d776ad9d5109239 c3c88f296a72432edb697ac8026dbf2ec18f2b21 Georgi Gerganov ggerganov@gmail.com 2024-05-13T11:01:07+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:08:09+03:00 metal : tune soft_max number of threads (whisper/0)
c3c88f296a72432edb697ac8026dbf2ec18f2b21 182adefcf36fc5f4263082ff032c0796fda65578 Georgi Gerganov ggerganov@gmail.com 2024-05-12T20:36:31+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:08:09+03:00 ggml : try fix ppc64 (whisper/0)
182adefcf36fc5f4263082ff032c0796fda65578 0d26d8ccd8caebab75af697c0275f599075fdacf Przemysław Pawełczyk przemoc@gmail.com 2024-05-08T17:33:43+02:00 Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:08:09+03:00 ggml : expose SSE3 and SSSE3 for MSVC when AVX is available (whisper/2128)
0d26d8ccd8caebab75af697c0275f599075fdacf 4f0263633b40e94e8b69fd6e7e4395cfedfd5c12 Hong Bo PENG penghb@cn.ibm.com 2024-05-12T17:17:18+08:00 Georgi Gerganov ggerganov@gmail.com 2024-05-14T19:08:09+03:00 ggml : optimize for ppc64le using VSX intrinsics (ggml/784)
4f0263633b40e94e8b69fd6e7e4395cfedfd5c12 1265c670fd8e41e1947352c96c5179adda97fb2c Steve Grubb ausearch.1@gmail.com 2024-05-14T10:11:24-04:00 GitHub noreply@github.com 2024-05-14T16:11:24+02:00 server: free sampling contexts on exit (#7264)
1265c670fd8e41e1947352c96c5179adda97fb2c 5e31828d3e35c76ecfee665bc23771a4bec1d130 Brian mofosyne@gmail.com 2024-05-14T23:10:39+10:00 GitHub noreply@github.com 2024-05-14T16:10:39+03:00 Revert "move ndk code to a new library (#6951)" (#7282)
5e31828d3e35c76ecfee665bc23771a4bec1d130 541600201e6480f54ae09e58d16b154d4b4b331d Radoslav Gerganov rgerganov@gmail.com 2024-05-14T14:27:19+03:00 GitHub noreply@github.com 2024-05-14T14:27:19+03:00 ggml : add RPC backend (#6829)
541600201e6480f54ae09e58d16b154d4b4b331d efc8f767c8c8c749a245dd96ad4e2f37c164b54c slaren slarengh@gmail.com 2024-05-14T09:33:42+02:00 GitHub noreply@github.com 2024-05-14T17:33:42+10:00 llama : disable pipeline parallelism with nkvo (#7265)
efc8f767c8c8c749a245dd96ad4e2f37c164b54c e0f556186b6e1f2b7032a1479edf5e89e2b1bd86 Elton Kola eltonkola@gmail.com 2024-05-14T03:30:30-04:00 GitHub noreply@github.com 2024-05-14T17:30:30+10:00 move ndk code to a new library (#6951)
e0f556186b6e1f2b7032a1479edf5e89e2b1bd86 27f65d6267cf22a44c5ccefa7765d53a05bd1259 Haggai Nuchi h.nuchi@gmail.com 2024-05-13T22:25:56-07:00 GitHub noreply@github.com 2024-05-14T15:25:56+10:00 Add left recursion check: quit early instead of going into an infinite loop (#7083)
27f65d6267cf22a44c5ccefa7765d53a05bd1259 ee52225067622babc277371511b8124884e1c797 Ryuei louixs@users.noreply.github.com 2024-05-14T14:20:47+09:00 GitHub noreply@github.com 2024-05-14T15:20:47+10:00 docs: Fix typo and update description for --embeddings flag (#7026)
ee52225067622babc277371511b8124884e1c797 614d3b914e1c3e02596f869649eb4f1d3b68614d compilade git@compilade.net 2024-05-13T14:10:51-04:00 GitHub noreply@github.com 2024-05-13T14:10:51-04:00 convert-hf : support direct Q8_0 conversion (#7234)
614d3b914e1c3e02596f869649eb4f1d3b68614d 30e70334f71b3bd115024affcf98cac3d79aaa95 Georgi Gerganov ggerganov@gmail.com 2024-05-13T17:15:15+03:00 GitHub noreply@github.com 2024-05-13T17:15:15+03:00 llama : less KV padding when FA is off (#7257)
30e70334f71b3bd115024affcf98cac3d79aaa95 1c570d8beeebad95872dc738ea542a4a0022f78a k.h.lai adrian.k.h.lai@outlook.com 2024-05-13T22:02:36+08:00 GitHub noreply@github.com 2024-05-14T00:02:36+10:00 llava-cli: fix base64 prompt (#7248)
1c570d8beeebad95872dc738ea542a4a0022f78a 948f4ec7c5bff92b18e63303f2b2d1645bccd943 Johannes Gäßler johannesg@5d6.de 2024-05-13T13:03:27+02:00 GitHub noreply@github.com 2024-05-13T13:03:27+02:00 perplexity: add BF16 vs. FP16 results (#7150)
948f4ec7c5bff92b18e63303f2b2d1645bccd943 9aa672490c848e45eaa704a554e0f1f6df995fc8 Neo Zhang 14088817+arthw@users.noreply.github.com 2024-05-13T18:11:26+08:00 GitHub noreply@github.com 2024-05-13T18:11:26+08:00 [SYCL] rm wait() (#7233)
9aa672490c848e45eaa704a554e0f1f6df995fc8 b1f8af1886e8187db6bb2a9b87cfc1c0f175f629 Joan Fontanals joan.fontanals.martinez@jina.ai 2024-05-13T10:35:14+02:00 GitHub noreply@github.com 2024-05-13T11:35:14+03:00 llama : rename jina tokenizers to v2 (#7249)
b1f8af1886e8187db6bb2a9b87cfc1c0f175f629 e586ee42595500c53938e937b6b6ad5353ad76dc Brian mofosyne@gmail.com 2024-05-13T12:56:47+10:00 GitHub noreply@github.com 2024-05-13T12:56:47+10:00 convert.py: Outfile default name change and additional metadata support (#4858)
e586ee42595500c53938e937b6b6ad5353ad76dc cbf75894d256f1861f6409565db599365de3d4b8 Benjamin Findley 39356821+Kartoffelsaft@users.noreply.github.com 2024-05-12T19:40:08-07:00 GitHub noreply@github.com 2024-05-13T12:40:08+10:00 change default temperature of OAI compat API from 0 to 1 (#7226)
cbf75894d256f1861f6409565db599365de3d4b8 0d5cef78aeafae4d4e6d56e2d4bcda771af58cc9 Neo Zhang 14088817+arthw@users.noreply.github.com 2024-05-13T08:04:29+08:00 GitHub noreply@github.com 2024-05-13T08:04:29+08:00 [SYCL] Add oneapi runtime dll files to win release package (#7241)
0d5cef78aeafae4d4e6d56e2d4bcda771af58cc9 dc685be46622a8fabfd57cfa804237c8f15679b8 Neo Zhang 14088817+arthw@users.noreply.github.com 2024-05-13T08:02:55+08:00 GitHub noreply@github.com 2024-05-13T08:02:55+08:00 [SYCL] update CI with oneapi 2024.1 (#7235)
dc685be46622a8fabfd57cfa804237c8f15679b8 6f1b63606fc68a09d62d1d74dbd156c35219026d Johannes Gäßler johannesg@5d6.de 2024-05-12T19:40:45+02:00 GitHub noreply@github.com 2024-05-12T19:40:45+02:00 CUDA: add FP32 FlashAttention vector kernel (#7188)
6f1b63606fc68a09d62d1d74dbd156c35219026d b228aba91ac2cd9eb90e9d423ba1d0d20e0117e2 Georgi Gerganov ggerganov@gmail.com 2024-05-12T18:30:23+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-12T18:30:23+03:00 cmake : fix version cmp (#7227)
b228aba91ac2cd9eb90e9d423ba1d0d20e0117e2 7bd4ffb78062587e4012a1c24186223f09b1bc70 slaren slarengh@gmail.com 2024-05-12T02:29:33+02:00 GitHub noreply@github.com 2024-05-12T02:29:33+02:00 remove convert-lora-to-ggml.py (#7204)
7bd4ffb78062587e4012a1c24186223f09b1bc70 1622ac023f42e5e01c163321cd98c6596aa9402d Georgi Gerganov ggerganov@gmail.com 2024-05-11T21:36:20+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-11T21:38:13+03:00 metal : fix warnings (skipme) (#0)
1622ac023f42e5e01c163321cd98c6596aa9402d 6aeff24f8b91e145e92d17ec7ce3adc4ef60b8e9 Georgi Gerganov ggerganov@gmail.com 2024-05-11T21:35:05+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-11T21:35:05+03:00 sync : ggml
6aeff24f8b91e145e92d17ec7ce3adc4ef60b8e9 325756d28df7d018a7bac424e1b3bc8acb4ecf07 Georgi Gerganov ggerganov@gmail.com 2024-05-11T16:57:53+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-11T21:34:21+03:00 metal : fix indent (ggml/0)
325756d28df7d018a7bac424e1b3bc8acb4ecf07 fed0108491a3a3cbec6c6480dc8667ffff9d7659 Georgi Gerganov ggerganov@gmail.com 2024-05-11T16:25:50+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-11T21:33:08+03:00 ggml : resolve merge (ggml/0)
fed0108491a3a3cbec6c6480dc8667ffff9d7659 72c177c1f6c16693eee319d4ebd4eaab5e630dd2 Josh Ramer josh.ramer@icloud.com 2024-05-11T12:26:35-05:00 GitHub noreply@github.com 2024-05-12T03:26:35+10:00 Scripting & documenting debugging one test without anything else in the loop. (#7096)
72c177c1f6c16693eee319d4ebd4eaab5e630dd2 5a419926b0c4efab0531401aea91522aaea9fd07 Xuan Son Nguyen thichthat@gmail.com 2024-05-11T17:28:10+02:00 GitHub noreply@github.com 2024-05-11T17:28:10+02:00 fix system prompt handling (#7153)
5a419926b0c4efab0531401aea91522aaea9fd07 fae9d234b6606693704eca62fe4aefbb6c6abb45 compilade git@compilade.net 2024-05-11T11:06:26-04:00 GitHub noreply@github.com 2024-05-11T11:06:26-04:00 convert-hf : support bfloat16 conversion (#7158)
fae9d234b6606693704eca62fe4aefbb6c6abb45 f5ef34e428f3886544590ecb2d532e4d333c114c Georgi Gerganov ggerganov@gmail.com 2024-05-11T12:02:39+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-11T15:38:34+03:00 sync : ggml
f5ef34e428f3886544590ecb2d532e4d333c114c ef0d5e3ec9f99003af3ff326384816c02850ea3f Justina Cho justcho5@gmail.com 2024-05-01T14:44:26-07:00 Georgi Gerganov ggerganov@gmail.com 2024-05-11T15:38:34+03:00 feat: implemented sigmoid function (ggml/806)
ef0d5e3ec9f99003af3ff326384816c02850ea3f 3292733f95d4632a956890a438af5192e7031c12 Borislav Stanimirov b.stanimirov@abv.bg 2024-04-25T17:24:07+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-11T15:38:34+03:00 build: fix and ignore msvc warnings (ggml/805)
3292733f95d4632a956890a438af5192e7031c12 988631335a20d06497f58be0b8ba13adb4323a22 CrispStrobe 154636388+CrispStrobe@users.noreply.github.com 2024-05-11T10:18:35+02:00 GitHub noreply@github.com 2024-05-11T11:18:35+03:00 convert : skip unaccessible HF repos (#7210)
988631335a20d06497f58be0b8ba13adb4323a22 f99e1e456eaf69cc38c1982a2693ce41c0f897ef Steve Grubb ausearch.1@gmail.com 2024-05-11T04:13:02-04:00 GitHub noreply@github.com 2024-05-11T11:13:02+03:00 server : free llama_batch on exit (#7212)
f99e1e456eaf69cc38c1982a2693ce41c0f897ef 5ae3426b0b64672991563d4c28b2018b9f961467 Haoxiang Fei tonyfettes@tonyfettes.com 2024-05-11T16:12:06+08:00 GitHub noreply@github.com 2024-05-11T11:12:06+03:00 llama : lookup word in vocab before doing BPE merges (#7193)
5ae3426b0b64672991563d4c28b2018b9f961467 b83cc3f5b303ff30c52874b2d5864dc6385ebf9f Johannes Gäßler johannesg@5d6.de 2024-05-11T10:11:28+02:00 GitHub noreply@github.com 2024-05-11T10:11:28+02:00 server: fix reported top tokens for temperature 0 (#7203)
b83cc3f5b303ff30c52874b2d5864dc6385ebf9f 9cb317f77e53067f7a138cc89ef7657148eae8e6 Joan Fontanals jfontanalsmartinez@gmail.com 2024-05-11T09:46:09+02:00 GitHub noreply@github.com 2024-05-11T10:46:09+03:00 llama : add Jina Embeddings architecture (#6826)
9cb317f77e53067f7a138cc89ef7657148eae8e6 e849648888a11de13aaaa4cb2eda3f5a9c7b444d Georgi Gerganov ggerganov@gmail.com 2024-05-11T10:32:41+03:00 GitHub noreply@github.com 2024-05-11T10:32:41+03:00 ggml : full ALiBi support (#7192)
e849648888a11de13aaaa4cb2eda3f5a9c7b444d 18e437665ce626dddbd79119aa7498493e7cb13b slaren slarengh@gmail.com 2024-05-10T18:03:54+02:00 GitHub noreply@github.com 2024-05-10T18:03:54+02:00 llama-bench : add pp+tg test type (#7199)
18e437665ce626dddbd79119aa7498493e7cb13b 8c660242d708d3913a2adc2b6e4a9ee9cf5e4ce7 Georgi Gerganov ggerganov@gmail.com 2024-05-10T18:20:10+03:00 GitHub noreply@github.com 2024-05-10T18:20:10+03:00 metal : fix flash attention kernel requirements (#7169)
8c660242d708d3913a2adc2b6e4a9ee9cf5e4ce7 25c6e82e7a1ad25a42b0894e87d9b5c557409516 Georgi Gerganov ggerganov@gmail.com 2024-05-10T17:53:04+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-10T17:53:04+03:00 convert : print "ignore_merges" field
25c6e82e7a1ad25a42b0894e87d9b5c557409516 4e3880978f8b1bf546dd4e6f3b524d6b8739c49c slaren slarengh@gmail.com 2024-05-10T14:28:01+02:00 GitHub noreply@github.com 2024-05-10T14:28:01+02:00 llama : use n_vocab to differentiate between mistral 7B and llama3 8B (#7200)
4e3880978f8b1bf546dd4e6f3b524d6b8739c49c f89fe2732c5709f6e86d5f4aee2e6d2a561f2eb2 Justine Tunney jtunney@mozilla.com 2024-05-10T07:01:08-04:00 GitHub noreply@github.com 2024-05-10T21:01:08+10:00 Fix memory bug in grammar parser (#7194)
f89fe2732c5709f6e86d5f4aee2e6d2a561f2eb2 d11afd665241c1b3910ab5f040d0216403019d87 HanishKVC hanishkvc@gmail.com 2024-05-10T15:51:58+05:30 GitHub noreply@github.com 2024-05-10T20:21:58+10:00 Main+: optionally allow special tokens from user in interactive mode (#7097)
d11afd665241c1b3910ab5f040d0216403019d87 8c570c9496212073079476651c7517c02581101f Andrei abetlen@gmail.com 2024-05-10T02:41:10-04:00 GitHub noreply@github.com 2024-05-10T09:41:10+03:00 llava : fix moondream support (#7163)
8c570c9496212073079476651c7517c02581101f eaf4bd8b399a6ed4b834f91d22409d2a05c4d266 Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-05-10T01:32:15+01:00 GitHub noreply@github.com 2024-05-10T08:32:15+08:00 Minor arithmetic improvement to mmvq wrapper kernel (#7172)
eaf4bd8b399a6ed4b834f91d22409d2a05c4d266 befddd0f15de6efb15d7e7f5b527dfb671f4196f slaren slarengh@gmail.com 2024-05-10T01:04:12+02:00 GitHub noreply@github.com 2024-05-10T01:04:12+02:00 eval-callback : fix conversion to float (#7184)
befddd0f15de6efb15d7e7f5b527dfb671f4196f d46dbc76f8770caec0175f1e57777173c70556a0 0cc4m picard12@live.de 2024-05-09T20:39:54+02:00 GitHub noreply@github.com 2024-05-09T20:39:54+02:00 Vulkan Bugfixes and Improvements (#7084)
d46dbc76f8770caec0175f1e57777173c70556a0 0961d866044143eefec5b525e991611f73fd4f6e Georgi Gerganov ggerganov@gmail.com 2024-05-09T16:40:42+03:00 GitHub noreply@github.com 2024-05-09T16:40:42+03:00 readme : add scheduled server workflow status badge
0961d866044143eefec5b525e991611f73fd4f6e 43248e559472556f368988575d9fba906b3eb139 l3utterfly gc.pthzfoldr@gmail.com 2024-05-09T22:32:40+09:00 GitHub noreply@github.com 2024-05-09T16:32:40+03:00 readme : add app (#6371)
43248e559472556f368988575d9fba906b3eb139 a743d76a01f23038b2c85af1e9048ee836767b44 jaime-m-p 167997752+jaime-m-p@users.noreply.github.com 2024-05-09T15:30:44+02:00 GitHub noreply@github.com 2024-05-09T23:30:44+10:00 llama3 custom regex split (#6965)
a743d76a01f23038b2c85af1e9048ee836767b44 f31ec120bc36c6270e4948e6a065a7c4cfa0c404 Johannes Gäßler johannesg@5d6.de 2024-05-09T14:32:02+02:00 GitHub noreply@github.com 2024-05-09T14:32:02+02:00 CUDA: generalize FP16 fattn vec kernel (#7061)
f31ec120bc36c6270e4948e6a065a7c4cfa0c404 fd9f92b154850014146f61717cd292a59a5cee5a Galunid karolek1231456@gmail.com 2024-05-09T14:13:05+02:00 GitHub noreply@github.com 2024-05-09T14:13:05+02:00 Add warning if token is invalid (#7173)
fd9f92b154850014146f61717cd292a59a5cee5a 22842164bcae3251b81ad9e497a16ef66833cb9e Daniel Bevenius daniel.bevenius@gmail.com 2024-05-09T13:03:29+02:00 GitHub noreply@github.com 2024-05-09T14:03:29+03:00 llama : update llama_timings.n_p_eval setting (#7160)
22842164bcae3251b81ad9e497a16ef66833cb9e 47345248827f426038098d016ed11975021b4919 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-05-09T12:56:00+02:00 GitHub noreply@github.com 2024-05-09T13:56:00+03:00 gguf-py : add special token modification capability (#7166)
47345248827f426038098d016ed11975021b4919 07cd41d0965829463eff73eda3348aedbfd3a444 Albert Jin albert.jin@gmail.com 2024-05-09T17:34:37+08:00 GitHub noreply@github.com 2024-05-09T12:34:37+03:00 opencl : alignment size converted from bits to bytes (#7090)
07cd41d0965829463eff73eda3348aedbfd3a444 4426e2987b566f09c7aa96ada9706cc778637620 Ahmet Zeer ahmed.zeer@std.yildiz.edu.tr 2024-05-09T11:16:45+03:00 GitHub noreply@github.com 2024-05-09T10:16:45+02:00 TypoFix (#7162)
4426e2987b566f09c7aa96ada9706cc778637620 f98eb31c517c95960df1d0abc48002787f145f3b Jared Van Bortel jared@nomic.ai 2024-05-08T19:55:32-04:00 GitHub noreply@github.com 2024-05-08T19:55:32-04:00 cmake : fix typo (#7151)
f98eb31c517c95960df1d0abc48002787f145f3b bc4bba364fb96d908f2698e908648df5e6f55e02 compilade git@compilade.net 2024-05-08T18:16:38-04:00 GitHub noreply@github.com 2024-05-08T18:16:38-04:00 convert-hf : save memory with lazy evaluation (#7075)
bc4bba364fb96d908f2698e908648df5e6f55e02 c12452c7aec8a02264afc00196a13caa591a13ac agray3 agray3@users.noreply.github.com 2024-05-08T21:55:49+01:00 GitHub noreply@github.com 2024-05-08T22:55:49+02:00 Introduction of CUDA Graphs to LLama.cpp (#6766)
c12452c7aec8a02264afc00196a13caa591a13ac 9da243b36ac0b9d609adfaaa4c8f1cc8c592f737 Johannes Gäßler johannesg@5d6.de 2024-05-08T21:53:08+02:00 GitHub noreply@github.com 2024-05-08T21:53:08+02:00 JSON: [key] -> .at(key), assert() -> GGML_ASSERT (#7143)
9da243b36ac0b9d609adfaaa4c8f1cc8c592f737 bd1871fa2b1bf8a081b43ba9bc85f8ffd46fac46 Georgi Gerganov ggerganov@gmail.com 2024-05-08T22:14:39+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-08T22:14:39+03:00 Revert "llava : add support for moondream vision language model (#6899)"
bd1871fa2b1bf8a081b43ba9bc85f8ffd46fac46 26458af1d63c85195cd96cd1673051e332d06d30 JohnnyB jboero@users.noreply.github.com 2024-05-08T20:12:06+01:00 GitHub noreply@github.com 2024-05-08T22:12:06+03:00 server : add themes + favicon (#6848)
26458af1d63c85195cd96cd1673051e332d06d30 83330d8cd6491e53e1aca4c5dfc47e039b3c04ff Gilad S giladgd@users.noreply.github.com 2024-05-08T22:08:10+03:00 GitHub noreply@github.com 2024-05-08T22:08:10+03:00 metal : use `vm_allocate` instead of `posix_memalign` on macOS (#7078)
83330d8cd6491e53e1aca4c5dfc47e039b3c04ff 465263d0cf1e8f8bc41948332dbd009d27a68590 Dawid Potocki github@dawidpotocki.com 2024-05-09T02:32:32+12:00 GitHub noreply@github.com 2024-05-08T17:32:32+03:00 main : add --conversation / -cnv flag (#7108)
465263d0cf1e8f8bc41948332dbd009d27a68590 911b3900dded9a1cfe0f0e41b82c7a29baf3a217 Eve 139727413+netrunnereve@users.noreply.github.com 2024-05-08T14:29:23Z GitHub noreply@github.com 2024-05-08T17:29:23+03:00 sgemm : AVX Q4_0 and Q8_0 (#6891)
911b3900dded9a1cfe0f0e41b82c7a29baf3a217 ad211edef5db1f1fb955874b7ca6a67bd0c88708 Johan JohanAR@users.noreply.github.com 2024-05-08T14:27:58+02:00 GitHub noreply@github.com 2024-05-08T15:27:58+03:00 server : add_special option for tokenize endpoint (#7059)
ad211edef5db1f1fb955874b7ca6a67bd0c88708 229ffff872f8ad0d21c997d18ee7a23692ae60a0 20kdc asdd2808@gmail.com 2024-05-08T13:22:32+01:00 GitHub noreply@github.com 2024-05-08T15:22:32+03:00 convert.py : --vocab-only generates false but valid params (#7027)
229ffff872f8ad0d21c997d18ee7a23692ae60a0 1fd9c1741d864d01cd7ec6d67227b92d7bfabf22 Ren Xuancheng jklj077@users.noreply.github.com 2024-05-08T20:06:43+08:00 GitHub noreply@github.com 2024-05-08T15:06:43+03:00 llama : add BPE pre-tokenization for Qwen2 (#7114)
1fd9c1741d864d01cd7ec6d67227b92d7bfabf22 4cd621c26de2095cd7c4464bdec5fe2e696ef3f3 Xuan Son Nguyen thichthat@gmail.com 2024-05-08T13:24:14+02:00 GitHub noreply@github.com 2024-05-08T13:24:14+02:00 clean up json_value & server_log (#7142)
4cd621c26de2095cd7c4464bdec5fe2e696ef3f3 7e0b6a7b3ba94ff624dc27c1e0e735fded8819b8 DAN™ dranger003@gmail.com 2024-05-08T06:43:23-04:00 GitHub noreply@github.com 2024-05-08T13:43:23+03:00 convert : add BPE pre-tokenization for DBRX (#7132)
7e0b6a7b3ba94ff624dc27c1e0e735fded8819b8 acdce3cdef6fc2f0b7b5623231fd7762c0884d1c Georgi Gerganov ggerganov@gmail.com 2024-05-08T12:47:07+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-08T12:47:07+03:00 py : also print the normalizers
acdce3cdef6fc2f0b7b5623231fd7762c0884d1c 3855416027cb25d9a708ffa5581cf503a87856a6 Brian mofosyne@gmail.com 2024-05-08T18:54:39+10:00 GitHub noreply@github.com 2024-05-08T10:54:39+02:00 compare-llama-bench.py: add missing basicConfig (#7138)
3855416027cb25d9a708ffa5581cf503a87856a6 c0e6fbf8c380718102bd25fcb8d2e55f8f9480d1 Justine Tunney jtunney@mozilla.com 2024-05-08T02:30:09-04:00 GitHub noreply@github.com 2024-05-08T09:30:09+03:00 ggml : introduce bfloat16 support (#6412)
c0e6fbf8c380718102bd25fcb8d2e55f8f9480d1 c780e75305dba1f67691a8dc0e8bc8425838a452 Georgi Gerganov ggerganov@gmail.com 2024-05-08T09:14:50+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-08T09:14:50+03:00 metal : fix unused warning
c780e75305dba1f67691a8dc0e8bc8425838a452 48b2f9c1fc71ab7df5432be2ed9fa7cdf5e8405e Jeximo jeximo@gmail.com 2024-05-07T21:26:43-03:00 GitHub noreply@github.com 2024-05-08T02:26:43+02:00 Further tidy on Android instructions README.md (#7077)
48b2f9c1fc71ab7df5432be2ed9fa7cdf5e8405e af0a5b616359809ce886ea433acedebb39b12969 jukofyork 69222624+jukofyork@users.noreply.github.com 2024-05-08T01:24:16+01:00 GitHub noreply@github.com 2024-05-08T02:24:16+02:00 Fixed save_imatrix to match old behaviour for MoE (#7099)
af0a5b616359809ce886ea433acedebb39b12969 b6aa6702030320a3d5fbc2508307af0d7c947e40 Johannes Gäßler johannesg@5d6.de 2024-05-07T23:07:58+02:00 GitHub noreply@github.com 2024-05-07T23:07:58+02:00 server: fix incorrectly reported token probabilities (#7125)
b6aa6702030320a3d5fbc2508307af0d7c947e40 260b7c65296fba0568eeb1ff05244ea0be206b54 nopperl 54780682+nopperl@users.noreply.github.com 2024-05-07T19:39:43Z GitHub noreply@github.com 2024-05-07T21:39:43+02:00 Fix OLMo HF to GGUF conversion (#6910)
260b7c65296fba0568eeb1ff05244ea0be206b54 53d6c52e227dedef347b21e28febcfb9caeecdad Kyle Mistele kyle@mistele.com 2024-05-07T13:44:29-05:00 GitHub noreply@github.com 2024-05-07T21:44:29+03:00 server : update readme with undocumented options (#7013)
53d6c52e227dedef347b21e28febcfb9caeecdad 3af34c1d1b0da47f85b95f60922abeded1cb5d33 Georgi Gerganov ggerganov@gmail.com 2024-05-07T21:43:13+03:00 GitHub noreply@github.com 2024-05-07T21:43:13+03:00 readme : update hot topics
3af34c1d1b0da47f85b95f60922abeded1cb5d33 04976db7a819fcf8bfefbfc09a3344210b79dd27 RhinoDevel RhinoDevel@users.noreply.github.com 2024-05-07T19:51:31+02:00 GitHub noreply@github.com 2024-05-07T20:51:31+03:00 main : update log text (EOS to EOG) (#7104)
04976db7a819fcf8bfefbfc09a3344210b79dd27 947d3ad27d94f1addef76b5d64c314618f063933 omahs 73983677+omahs@users.noreply.github.com 2024-05-07T17:20:33+02:00 GitHub noreply@github.com 2024-05-07T18:20:33+03:00 docs: fix typos (#7124)
947d3ad27d94f1addef76b5d64c314618f063933 858f6b73f6e57a62523d16a955d565254be889b4 Georgi Gerganov ggerganov@gmail.com 2024-05-07T11:08:49+03:00 GitHub noreply@github.com 2024-05-07T11:08:49+03:00 ci : add GG_BUILD_EXTRA_TESTS_0 env (#7098)
858f6b73f6e57a62523d16a955d565254be889b4 b3a995b416e13ae3123a117a743e11d0ede0ca4c William Tambellini william.tambellini@gmail.com 2024-05-06T11:12:14-07:00 GitHub noreply@github.com 2024-05-06T20:12:14+02:00 Add an option to build without CUDA VMM (#7067)
b3a995b416e13ae3123a117a743e11d0ede0ca4c bcdee0daa7c5e8e086b719e5eb4073b00df70e01 Georgi Gerganov ggerganov@gmail.com 2024-05-06T18:36:06+03:00 GitHub noreply@github.com 2024-05-06T08:36:06-07:00 flake.lock: Update (#7079)
bcdee0daa7c5e8e086b719e5eb4073b00df70e01 628b299106d1e9476fdecb3cbe546bf5c60f1b89 Georgi Gerganov ggerganov@gmail.com 2024-05-06T09:31:30+03:00 Georgi Gerganov ggerganov@gmail.com 2024-05-06T09:31:30+03:00 minor : fix trailing whitespace
628b299106d1e9476fdecb3cbe546bf5c60f1b89 8f8acc8683a00ce40fa5a81161a079d2167126e6 kunnis kunnis@users.noreply.github.com 2024-05-05T07:17:47-05:00 GitHub noreply@github.com 2024-05-05T14:17:47+02:00 Adding support for the --numa argument for llama-bench. (#7080)
8f8acc8683a00ce40fa5a81161a079d2167126e6 ca3632602091e959ed2ad4c09c67a7c790b10d31 Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-05-05T13:38:55+02:00 GitHub noreply@github.com 2024-05-05T13:38:55+02:00 Disable benchmark on forked repo (#7034)
ca3632602091e959ed2ad4c09c67a7c790b10d31 889bdd76866ea31a7625ec2dcea63ff469f3e981 Lyle Dean dean@lyle.dev 2024-05-05T06:21:46+01:00 GitHub noreply@github.com 2024-05-05T08:21:46+03:00 readme : add note that LLaMA 3 is not supported with convert.py (#7065)
889bdd76866ea31a7625ec2dcea63ff469f3e981 6fbd43221167bf96112f899daf22c127b282cbcf DAN™ dranger003@gmail.com 2024-05-05T01:19:30-04:00 GitHub noreply@github.com 2024-05-05T08:19:30+03:00 command-r : add BPE pre-tokenization (#7063)
6fbd43221167bf96112f899daf22c127b282cbcf 842500144ee02c8b0a46d2cc43880f8d80998fa5 Brian mofosyne@gmail.com 2024-05-05T15:07:48+10:00 GitHub noreply@github.com 2024-05-05T08:07:48+03:00 py : logging and flake8 suppression refactoring (#7081)
842500144ee02c8b0a46d2cc43880f8d80998fa5 cf768b7e71cbcc9886c753ae963c2b68893d02e4 Xuan Son Nguyen thichthat@gmail.com 2024-05-04T18:56:22+02:00 GitHub noreply@github.com 2024-05-04T18:56:22+02:00 gguf-split: add --no-tensor-first-split (#7072)
cf768b7e71cbcc9886c753ae963c2b68893d02e4 fcd84a0f5a584ab5271745d7ffef21c8a6bc7b0c Jeximo jeximo@gmail.com 2024-05-04T13:10:15-03:00 GitHub noreply@github.com 2024-05-04T18:10:15+02:00 Tidy Android Instructions README.md (#7016)
fcd84a0f5a584ab5271745d7ffef21c8a6bc7b0c 03fb8a002df2e96104f9e06de9c78d2a8ed91e92 viric viric@viric.name 2024-05-04T15:26:53+02:00 GitHub noreply@github.com 2024-05-04T15:26:53+02:00 Fix Linux /sys cpu path to guess number of cores (#7064)
03fb8a002df2e96104f9e06de9c78d2a8ed91e92 92139b90af4841d7fd060b526bdd443b621770ff maor-ps 154728172+maor-ps@users.noreply.github.com 2024-05-04T12:06:40+03:00 GitHub noreply@github.com 2024-05-04T11:06:40+02:00 If first token generated from the server is the stop word the server will crash (#7038)
92139b90af4841d7fd060b526bdd443b621770ff a2ac89d6efb41b535778bfeaecaae8fe295b6ed3 Georgi Gerganov ggerganov@gmail.com 2024-05-04T08:32:32+03:00 GitHub noreply@github.com 2024-05-04T08:32:32+03:00 tests : add test-tokenizer-0.sh + fix some tokenizers (#7036)
a2ac89d6efb41b535778bfeaecaae8fe295b6ed3 433def286e98751bf17db75dce53847d075c0be5 Brian mofosyne@gmail.com 2024-05-04T05:36:41+10:00 GitHub noreply@github.com 2024-05-03T22:36:41+03:00 convert.py : add python logging instead of print() (#6511)
433def286e98751bf17db75dce53847d075c0be5 60325fa56f61c228464c9f065db3aa6a61f2156e Daniel Bevenius daniel.bevenius@gmail.com 2024-05-03T15:24:30+02:00 GitHub noreply@github.com 2024-05-03T15:24:30+02:00 llama : rename ctx to user_data in progress_callback (#7045)
60325fa56f61c228464c9f065db3aa6a61f2156e 6ecf3189e00a1e8e737a78b6d10e1d7006e050a2 Bartowski ckealty1182@gmail.com 2024-05-02T19:49:09-04:00 GitHub noreply@github.com 2024-05-03T01:49:09+02:00 Remove .attention from skipped tensors to match more accurately (#7051)
6ecf3189e00a1e8e737a78b6d10e1d7006e050a2 b0d943de179ad5dbd83d51f327fb566066f4ccda alwqx kenan3015@gmail.com 2024-05-02T23:56:41+08:00 GitHub noreply@github.com 2024-05-02T11:56:41-04:00 chore: fix typo in llama.cpp (#7032)
b0d943de179ad5dbd83d51f327fb566066f4ccda 8d608a81b7bd170f700648f8214e6f3279d4d715 Andrew Downing andrew2085@gmail.com 2024-05-01T17:31:30-04:00 GitHub noreply@github.com 2024-05-01T23:31:30+02:00 Update LOG_IMPL and LOG_TEE_IMPL (#7029)
8d608a81b7bd170f700648f8214e6f3279d4d715 3ea0d36000e2314baba7e9fc6a97f08670a6f7e4 l3utterfly gc.pthzfoldr@gmail.com 2024-05-02T04:27:41+09:00 GitHub noreply@github.com 2024-05-01T22:27:41+03:00 main : fix off by one error for context shift (#6921)
3ea0d36000e2314baba7e9fc6a97f08670a6f7e4 1613ef8d8eb2479ba55c4d598e08c8f3f18a0fed Johannes Gäßler johannesg@5d6.de 2024-05-01T17:52:55+02:00 GitHub noreply@github.com 2024-05-01T17:52:55+02:00 Server: add tests for batch size, different seeds (#6950)
1613ef8d8eb2479ba55c4d598e08c8f3f18a0fed c4ec9c0d3d67e6b33638e6dad86419e6fd5ffe01 Johannes Gäßler johannesg@5d6.de 2024-05-01T14:46:37+02:00 GitHub noreply@github.com 2024-05-01T14:46:37+02:00 CUDA: CUDART < 11.7 workaround for __hmax, __hmax2 (#7019)
c4ec9c0d3d67e6b33638e6dad86419e6fd5ffe01 a8f9b076316e16aadd0791015b3bfd446fe1e904 slaren slarengh@gmail.com 2024-05-01T07:13:59+02:00 GitHub noreply@github.com 2024-05-01T08:13:59+03:00 ci : exempt confirmed bugs from being tagged as stale (#7014)
a8f9b076316e16aadd0791015b3bfd446fe1e904 f364eb6fb5d46118a76fa045f487318de4c24961 Johannes Gäßler johannesg@5d6.de 2024-04-30T23:36:27+02:00 GitHub noreply@github.com 2024-04-30T23:36:27+02:00 perplexity: more statistics, added documentation (#6936)
f364eb6fb5d46118a76fa045f487318de4c24961 77e15bec6217a39be59b9cc83d6b9afb6b0d8167 Kevin Gibbons bakkot@gmail.com 2024-04-30T08:14:02-07:00 GitHub noreply@github.com 2024-04-30T17:14:02+02:00 switch to using localizedDescription (#7010)
77e15bec6217a39be59b9cc83d6b9afb6b0d8167 a68a1e7ed060ee5af2d638585d19e3510ddbf16c Georgi Gerganov ggerganov@gmail.com 2024-04-30T15:52:21+03:00 GitHub noreply@github.com 2024-04-30T15:52:21+03:00 metal : remove deprecated error code (#7008)
a68a1e7ed060ee5af2d638585d19e3510ddbf16c 9c67c2773d4b706cf71d70ecf4aa180b62501960 Kevin Gibbons bakkot@gmail.com 2024-04-30T02:34:50-07:00 GitHub noreply@github.com 2024-04-30T12:34:50+03:00 metal : log more info on error (#6987)
9c67c2773d4b706cf71d70ecf4aa180b62501960 952d03dbead16e4dbdd1d3458486340673cc2465 Georgi Gerganov ggerganov@gmail.com 2024-04-30T12:16:08+03:00 GitHub noreply@github.com 2024-04-30T12:16:08+03:00 ggml : add Flash Attention (#5021)
952d03dbead16e4dbdd1d3458486340673cc2465 8843a98c2ba97a25e93319a104f9ddfaf83ce4c4 Georgi Gerganov ggerganov@gmail.com 2024-04-30T11:05:25+03:00 GitHub noreply@github.com 2024-04-30T11:05:25+03:00 convert : use utf8 encoding (#7000)
8843a98c2ba97a25e93319a104f9ddfaf83ce4c4 b8c1476e44cc1f3a1811613f65251cf779067636 Olivier Chafik ochafik@users.noreply.github.com 2024-04-30T00:52:50+01:00 GitHub noreply@github.com 2024-04-30T00:52:50+01:00 Improve usability of --model-url & related flags (#6930)
b8c1476e44cc1f3a1811613f65251cf779067636 5539e6fdd1b97e0c37c54d34df67f334fc344a26 Clint Herron hanclinto@gmail.com 2024-04-29T14:40:14-04:00 GitHub noreply@github.com 2024-04-29T14:40:14-04:00 Extending grammar integration tests (#6644)
5539e6fdd1b97e0c37c54d34df67f334fc344a26 b8a7a5a90fd3187175d84227dad705ade395ba46 Daniel Bevenius daniel.bevenius@gmail.com 2024-04-29T19:56:59+02:00 GitHub noreply@github.com 2024-04-29T13:56:59-04:00 main : fix typo in comment in main.cpp (#6985)
b8a7a5a90fd3187175d84227dad705ade395ba46 d2c898f746a527f09effb061829e68b2e1812a28 Olivier Chafik ochafik@users.noreply.github.com 2024-04-29T17:02:45+01:00 GitHub noreply@github.com 2024-04-29T17:02:45+01:00 build(cmake): simplify instructions (`cmake -B build && cmake --build build ...`) (#6964)
d2c898f746a527f09effb061829e68b2e1812a28 544f1f10adeec3d5ed91c4d3bd3f903904ecea63 Georgi Gerganov ggerganov@gmail.com 2024-04-29T18:36:39+03:00 GitHub noreply@github.com 2024-04-29T18:36:39+03:00 ci : tmp disable gguf-split (#6983)
544f1f10adeec3d5ed91c4d3bd3f903904ecea63 ffe666572f98a686b17a2cd1dbf4c0a982e5ac0a Georgi Gerganov ggerganov@gmail.com 2024-04-29T17:55:02+03:00 GitHub noreply@github.com 2024-04-29T17:55:02+03:00 ggml : fix __MSC_VER -> _MSC_VER (#6977)
ffe666572f98a686b17a2cd1dbf4c0a982e5ac0a 24affa7db3c9db148854b0ab4fd63de8bca7d898 cpumaxx 163466046+cpumaxx@users.noreply.github.com 2024-04-29T07:34:24-07:00 GitHub noreply@github.com 2024-04-29T17:34:24+03:00 llava-cli : multiple images (#6969)
24affa7db3c9db148854b0ab4fd63de8bca7d898 f4ab2a41476600a98067a9474ea8f9e6db41bcfa Georgi Gerganov ggerganov@gmail.com 2024-04-29T17:06:19+03:00 GitHub noreply@github.com 2024-04-29T17:06:19+03:00 readme : update hot topics
f4ab2a41476600a98067a9474ea8f9e6db41bcfa 3f167476b11efa7ab08f6cacdeb8cab0935c1249 Georgi Gerganov ggerganov@gmail.com 2024-04-29T16:58:41+03:00 GitHub noreply@github.com 2024-04-29T16:58:41+03:00 llama : fix BPE pre-tokenization (#6920)
3f167476b11efa7ab08f6cacdeb8cab0935c1249 3055a4180557c6cbe29eacc8284c9e070ac10eab David Renshaw dwrenshaw@gmail.com 2024-04-29T09:35:45-04:00 GitHub noreply@github.com 2024-04-29T16:35:45+03:00 sampling : use std::random_device{}() for default random seed (#6962)
3055a4180557c6cbe29eacc8284c9e070ac10eab 577277ffd203b190c3dc2ab3e737946dc432132c Christian Zhou-Zheng 59622928+christianazinn@users.noreply.github.com 2024-04-29T09:34:41-04:00 GitHub noreply@github.com 2024-04-29T16:34:41+03:00 convert : fix conversion of some BERT embedding models (#6937)
577277ffd203b190c3dc2ab3e737946dc432132c ca7f29f568803bee4c92d1b3e41c7d721b0dc570 Przemysław Pawełczyk przemoc@gmail.com 2024-04-29T15:08:20+02:00 GitHub noreply@github.com 2024-04-29T16:08:20+03:00 make : change GNU make default CXX from g++ to c++ (#6966)
ca7f29f568803bee4c92d1b3e41c7d721b0dc570 c4f708a93f1df5e35167f9313e000d381298be7f Przemysław Pawełczyk przemoc@gmail.com 2024-04-29T14:59:47+02:00 GitHub noreply@github.com 2024-04-29T15:59:47+03:00 ci : add building in MSYS2 environments (Windows) (#6967)
c4f708a93f1df5e35167f9313e000d381298be7f e00b4a8f816ebc45b98a46e5f5231359b9a017e0 Johannes Gäßler johannesg@5d6.de 2024-04-29T14:36:22+02:00 GitHub noreply@github.com 2024-04-29T15:36:22+03:00 llama : fix typo LAMMAFILE -> LLAMAFILE (#6974)
e00b4a8f816ebc45b98a46e5f5231359b9a017e0 7bb36ccf91b8a2e92b182dd75624f1fd7cb205ac DAN™ dranger003@gmail.com 2024-04-28T18:38:44-04:00 GitHub noreply@github.com 2024-04-29T00:38:44+02:00 Fix more int overflow during quant (PPL/CUDA). (#6563)
7bb36ccf91b8a2e92b182dd75624f1fd7cb205ac ce023f6f2ff34fbe840e32e65d443d2fed7393de Xuan Son Nguyen thichthat@gmail.com 2024-04-28T17:36:18+02:00 GitHub noreply@github.com 2024-04-28T17:36:18+02:00 gguf : enforce that tensor names are unique (#6905)
ce023f6f2ff34fbe840e32e65d443d2fed7393de 6e472f58e40cd4acf6023e15c75a2700535c5f0b Neo Zhang 14088817+arthw@users.noreply.github.com 2024-04-28T22:40:31+08:00 GitHub noreply@github.com 2024-04-28T22:40:31+08:00 add device version in device list (#6959)
6e472f58e40cd4acf6023e15c75a2700535c5f0b 4dba7e8114d84241c842b986e008af8b88d1a019 github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-04-28T00:18:27Z Someone newkozlukov@gmail.com 2024-04-28T11:12:50Z flake.lock: Update
4dba7e8114d84241c842b986e008af8b88d1a019 b7368332e24c5b2c8038bf8267f43632783fcc35 mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2024-04-27T21:02:06+02:00 GitHub noreply@github.com 2024-04-27T21:02:06+02:00 Replace "alternative" boolean operator in conditional compilation directive (#6949)
b7368332e24c5b2c8038bf8267f43632783fcc35 928e0b7013c862cf10701957b3d654aa70f11bd8 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-27T17:50:48+02:00 GitHub noreply@github.com 2024-04-27T17:50:48+02:00 ci: server: tests python env on github container ubuntu latest / fix n_predict (#6935)
928e0b7013c862cf10701957b3d654aa70f11bd8 0c4d489e29e53589bf13a801fe7c94b7b546d8f6 agray3 agray3@users.noreply.github.com 2024-04-26T19:08:30+01:00 GitHub noreply@github.com 2024-04-26T20:08:30+02:00 Reset schedule earlier to allow overlap with ggml graph computation on device (#6933)
0c4d489e29e53589bf13a801fe7c94b7b546d8f6 017e6999b5184234370b22a2f868e1be911e8d88 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-26T20:06:33+02:00 GitHub noreply@github.com 2024-04-26T20:06:33+02:00 quantize: add imatrix and dataset metadata in GGUF (#6658)
017e6999b5184234370b22a2f868e1be911e8d88 e2764cd7ca1112d9303eba9e81c9935ee67352ff slaren slarengh@gmail.com 2024-04-26T18:39:58+02:00 GitHub noreply@github.com 2024-04-26T18:39:58+02:00 add basic tensor data validation function (#6884)
e2764cd7ca1112d9303eba9e81c9935ee67352ff 4b1c3c98b442a4c84a788fff6323f6fa7e678a5b slaren slarengh@gmail.com 2024-04-26T17:07:42+02:00 GitHub noreply@github.com 2024-04-26T18:07:42+03:00 gguf : fix mismatch between alloc and free functions (#6929)
4b1c3c98b442a4c84a788fff6323f6fa7e678a5b bbe3c6e76157a5d806fdc155451f0ca8936248ee Justine Tunney jtunney@mozilla.com 2024-04-26T10:05:33-04:00 GitHub noreply@github.com 2024-04-26T17:05:33+03:00 llamafile : use 64-bit integers in sgemm (#6928)
bbe3c6e76157a5d806fdc155451f0ca8936248ee 7f5ff558eed0f732af8f25c2ab0645610bdec80c Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-26T12:27:25+02:00 GitHub noreply@github.com 2024-04-26T12:27:25+02:00 ci: server: fix python installation (#6925)
7f5ff558eed0f732af8f25c2ab0645610bdec80c 9e4e077ec50fde6049b128662c72d37a3c28e34b Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-26T12:15:30+02:00 GitHub noreply@github.com 2024-04-26T12:15:30+02:00 server: stop generation at `n_ctx_train` if `n_predict` is not set (#6638)
9e4e077ec50fde6049b128662c72d37a3c28e34b 83b72cb086ce46a33dececc86bfe4648b6120aa8 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-26T11:11:51+02:00 GitHub noreply@github.com 2024-04-26T11:11:51+02:00 ci: server: fix python installation (#6922)
83b72cb086ce46a33dececc86bfe4648b6120aa8 d4a9afc1009f0da88d04b2c5f672d81d5ae94675 Georgi Gerganov ggerganov@gmail.com 2024-04-26T10:41:53+03:00 GitHub noreply@github.com 2024-04-26T10:41:53+03:00 Merge pull request from GHSA-p5mv-gjc5-mwqv
d4a9afc1009f0da88d04b2c5f672d81d5ae94675 7d641c26ac73956a54b204cabefe85c764823678 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-26T09:27:49+02:00 GitHub noreply@github.com 2024-04-26T09:27:49+02:00 ci: server: fix python installation (#6918)
7d641c26ac73956a54b204cabefe85c764823678 5790c8dac1a4f0aa80b4efee3b962d8c04c829e8 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-26T09:26:59+02:00 GitHub noreply@github.com 2024-04-26T09:26:59+02:00 ci: fix concurrency for pull_request_target (#6917)
5790c8dac1a4f0aa80b4efee3b962d8c04c829e8 46e12c4692a37bdd31a0432fc5153d7d22bc7f72 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-26T09:26:16+02:00 GitHub noreply@github.com 2024-04-26T09:26:16+02:00 bench: server add stop word for PHI-2 (#6916)
46e12c4692a37bdd31a0432fc5153d7d22bc7f72 dba497e0c1eff27cf0e85d1d73c86fa08e1b5557 vik vikhyatk@gmail.com 2024-04-25T12:38:31-07:00 GitHub noreply@github.com 2024-04-25T22:38:31+03:00 llava : add support for moondream vision language model (#6899)
dba497e0c1eff27cf0e85d1d73c86fa08e1b5557 fa0b4ad25208d5ec2df6b998ccb29b49b249e82c Georgi Gerganov ggerganov@gmail.com 2024-04-25T21:31:17+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-25T21:37:27+03:00 cmake : restore LLAMA_LLAMAFILE_DEFAULT
fa0b4ad25208d5ec2df6b998ccb29b49b249e82c d6e1d44f16e5580cf47f13325ff49960bf13ca37 Georgi Gerganov ggerganov@gmail.com 2024-04-25T18:59:51+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-25T18:59:51+03:00 cmake : remove obsolete ANDROID check
d6e1d44f16e5580cf47f13325ff49960bf13ca37 853d06ffe26621176d60909a6e3f7c4cd067b305 slaren slarengh@gmail.com 2024-04-25T17:59:03+02:00 GitHub noreply@github.com 2024-04-25T17:59:03+02:00 llama : synchronize before get/set session data (#6911)
853d06ffe26621176d60909a6e3f7c4cd067b305 3fe0596c1817a6114ffffb6dbfd6c36ca7815dc7 Georgi Gerganov ggerganov@gmail.com 2024-04-25T17:06:27+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-25T17:06:27+03:00 ci : tmp disable slow tests
3fe0596c1817a6114ffffb6dbfd6c36ca7815dc7 0ead1f1072fdc70720f92e008a294dc74a826b1d BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2024-04-25T09:52:28-04:00 GitHub noreply@github.com 2024-04-25T16:52:28+03:00 readme : update model list (#6908)
0ead1f1072fdc70720f92e008a294dc74a826b1d 51543729ff5b1361ebfb164875c93dff0850d5fe slaren slarengh@gmail.com 2024-04-25T15:23:47+02:00 GitHub noreply@github.com 2024-04-25T15:23:47+02:00 llama : check that all the tensor data is in the model file (#6885)
51543729ff5b1361ebfb164875c93dff0850d5fe 4ab99d8d4762869506bb675b36501fd7c2af00b2 Georgi Gerganov ggerganov@gmail.com 2024-04-25T15:48:25+03:00 GitHub noreply@github.com 2024-04-25T15:48:25+03:00 ggml : fix redefinition of vaddvq_f32 for 32-bit ARM (#6906)
4ab99d8d4762869506bb675b36501fd7c2af00b2 54770413c484660d021dd51b5dbacab7880b8827 Daniel Bevenius daniel.bevenius@gmail.com 2024-04-25T14:38:14+02:00 GitHub noreply@github.com 2024-04-25T15:38:14+03:00 clip : rename lerp function to avoid conflict (#6894)
54770413c484660d021dd51b5dbacab7880b8827 aa750c1ede6232c91de890a14a7731d6daa2bc8e Georgi Gerganov ggerganov@gmail.com 2024-04-25T15:12:28+03:00 GitHub noreply@github.com 2024-04-25T15:12:28+03:00 ggml : fix MIN / MAX macros (#6904)
aa750c1ede6232c91de890a14a7731d6daa2bc8e 1966eb2615242f224bf9ca939db8905ab6a174a0 Georgi Gerganov ggerganov@gmail.com 2024-04-25T14:27:20+03:00 GitHub noreply@github.com 2024-04-25T14:27:20+03:00 tests : minor bash stuff (#6902)
1966eb2615242f224bf9ca939db8905ab6a174a0 784e11dea1f5ce9638851b2b0dddb107e2a609c8 jiez 373447296@qq.com 2024-04-25T18:29:35+08:00 GitHub noreply@github.com 2024-04-25T13:29:35+03:00 quantize : add '--keep-split' to quantize model into shards (#6688)
784e11dea1f5ce9638851b2b0dddb107e2a609c8 b4e4b8a9351d918a56831c73cf9f25c1837b80d1 Johannes Gäßler johannesg@5d6.de 2024-04-24T21:29:13+02:00 GitHub noreply@github.com 2024-04-24T21:29:13+02:00 README: add graphic for matrix multiplication (#6881)
b4e4b8a9351d918a56831c73cf9f25c1837b80d1 3fe847b5747676ee1bf90371c46ed0bc66b57240 Douglas Hanley thesecretaryofwar@gmail.com 2024-04-24T08:10:07-05:00 GitHub noreply@github.com 2024-04-24T16:10:07+03:00 llama : add llama_get_pooling_type function (#6862)
3fe847b5747676ee1bf90371c46ed0bc66b57240 37246b1031b1680c0dcaf20aef736d6b446203fa mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2024-04-24T12:54:24+02:00 GitHub noreply@github.com 2024-04-24T13:54:24+03:00 server : do not apply Markdown formatting in code sections (#6850)
37246b1031b1680c0dcaf20aef736d6b446203fa 28103f4832e301a9c84d44ff0df9d75d46ab6c76 Kyle Mistele kyle@mistele.com 2024-04-24T05:15:29-05:00 GitHub noreply@github.com 2024-04-24T13:15:29+03:00 common : revert showing control tokens by default for server (#6860)
28103f4832e301a9c84d44ff0df9d75d46ab6c76 c0d1b3e03e27634ac2871761f5033cf9324d472d Johannes Gäßler johannesg@5d6.de 2024-04-24T11:08:36+02:00 GitHub noreply@github.com 2024-04-24T11:08:36+02:00 Server: fix seed for multiple slots (#6835)
c0d1b3e03e27634ac2871761f5033cf9324d472d abd3314064cd3c513f9eef34c3ba6c23a107442c Georgi Gerganov ggerganov@gmail.com 2024-04-24T12:00:07+03:00 GitHub noreply@github.com 2024-04-24T12:00:07+03:00 ggml : move 32-bit arm compat in ggml-impl.h (#6865)
abd3314064cd3c513f9eef34c3ba6c23a107442c 3fec68be4e9577fc53158366d3b3af039c17bb1f Tristan Druyen tristan@vault81.mozmail.com 2024-04-24T10:52:37+02:00 GitHub noreply@github.com 2024-04-24T11:52:37+03:00 llama : add phi 3 chat template (#6857)
3fec68be4e9577fc53158366d3b3af039c17bb1f c8297c6af5693555652c40b95974b95d49d2674d Junyang Lin justinlin930319@hotmail.com 2024-04-24T15:16:21+08:00 GitHub noreply@github.com 2024-04-24T10:16:21+03:00 convert : add support of codeqwen due to tokenizer (#6707)
c8297c6af5693555652c40b95974b95d49d2674d 4e96a812b3ce7322a29a3008db2ed73d9087b176 liuwei-git 14815172+liuwei-git@users.noreply.github.com 2024-04-24T15:00:37+08:00 GitHub noreply@github.com 2024-04-24T10:00:37+03:00 llama : add phi3 support (#6852)
4e96a812b3ce7322a29a3008db2ed73d9087b176 192090bae47960f0d38d4967abe398a5d190057e Anas Ahouzi 112881240+aahouzi@users.noreply.github.com 2024-04-23T02:53:18+02:00 GitHub noreply@github.com 2024-04-23T08:53:18+08:00 [SYCL] Windows default build instructions without -DLLAMA_SYCL_F16 flag activated (#6767)
192090bae47960f0d38d4967abe398a5d190057e e931888d5024de814ce7119a18d6a959bfff3821 Justine Tunney jtunney@mozilla.com 2024-04-22T15:00:36-04:00 GitHub noreply@github.com 2024-04-22T22:00:36+03:00 llamafile : improve sgemm.cpp (#6796)
e931888d5024de814ce7119a18d6a959bfff3821 8960fe86ae075c846c5df8848230d1904ba8877f Dave Airlie airlied@gmail.com 2024-04-23T00:05:06+10:00 GitHub noreply@github.com 2024-04-22T16:05:06+02:00 ggml : fix calloc argument ordering. (#6820)
8960fe86ae075c846c5df8848230d1904ba8877f c0956b09ba845a7cd787d5580d7c8b96e80f40f5 Georgi Gerganov ggerganov@gmail.com 2024-04-22T15:41:11+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-22T15:41:11+03:00 llama : fix typo in <|im_end|> token text (#6745)
c0956b09ba845a7cd787d5580d7c8b96e80f40f5 e9b4a1bf68c18beff4e33f23ea62c1245b296915 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-22T13:22:54+02:00 GitHub noreply@github.com 2024-04-22T13:22:54+02:00 ci: fix job are cancelling each other (#6781)
e9b4a1bf68c18beff4e33f23ea62c1245b296915 5cf5e7d490dfdd2e70bface2d35dfd14aa44b4fb github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-04-21T00:17:47Z Someone newkozlukov@gmail.com 2024-04-22T10:42:43Z flake.lock: Update
5cf5e7d490dfdd2e70bface2d35dfd14aa44b4fb 40f74e4d739e9250431cf339ae7588b28d8d0663 Olivier Chafik ochafik@users.noreply.github.com 2024-04-21T18:48:53+01:00 GitHub noreply@github.com 2024-04-21T18:48:53+01:00 `build`: generate hex dump of server assets during build (#6661)
40f74e4d739e9250431cf339ae7588b28d8d0663 b9cc76d87e3d7ae5900f19d4fe8f8976d0a35888 Georgi Gerganov ggerganov@gmail.com 2024-04-21T18:36:45+03:00 GitHub noreply@github.com 2024-04-21T18:36:45+03:00 llama : add option to render special/control tokens (#6807)
b9cc76d87e3d7ae5900f19d4fe8f8976d0a35888 7dbdba5690ca61b3ee8c92cfac8e7e251042e787 Georgi Gerganov ggerganov@gmail.com 2024-04-21T16:47:57+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-21T16:48:50+03:00 ggml : fix ggml_backend_cpu_supports_op() for CPY (#0)
7dbdba5690ca61b3ee8c92cfac8e7e251042e787 c1386c936e9fbc38eb2816c711ab28f13355708e Wouter 9594229+DifferentialityDevelopment@users.noreply.github.com 2024-04-21T15:03:39+02:00 GitHub noreply@github.com 2024-04-21T16:03:39+03:00 llama : add llama-3 chat template (#6751)
c1386c936e9fbc38eb2816c711ab28f13355708e e8d35f47cb8cb4002fca02e18aaa1cb9fa21d6f1 pmysl piotr.myslinski@outlook.com 2024-04-21T14:49:30+02:00 GitHub noreply@github.com 2024-04-21T15:49:30+03:00 gguf-py : add IQ1_M to GGML_QUANT_SIZES (#6761)
e8d35f47cb8cb4002fca02e18aaa1cb9fa21d6f1 2cca09d509c0e114acc16cb347c3cdd86e5f1b40 Jan Boon jan.boon@kaetemi.be 2024-04-21T20:35:40+08:00 GitHub noreply@github.com 2024-04-21T15:35:40+03:00 doc : add link to falcon (#6789)
2cca09d509c0e114acc16cb347c3cdd86e5f1b40 89b0bf0d5dc123cc1053708f5f84b89e52cdc80b Mohammadreza Hendiani mohammad.r.hendiani@gmail.com 2024-04-21T16:02:05+03:30 GitHub noreply@github.com 2024-04-21T15:32:05+03:00 readme : add Fedora instructions (#6783)
89b0bf0d5dc123cc1053708f5f84b89e52cdc80b b97bc3966e852adb626c90be64fd48282800f504 Justine Tunney jtunney@mozilla.com 2024-04-21T08:19:04-04:00 GitHub noreply@github.com 2024-04-21T15:19:04+03:00 llava : use logger in llava-cli (#6797)
b97bc3966e852adb626c90be64fd48282800f504 b8109bc0139f15a5b321909f47510b89dca47ffc Pedro Cuenca pedro@huggingface.co 2024-04-21T13:50:41+02:00 GitHub noreply@github.com 2024-04-21T14:50:41+03:00 llama : support Llama 3 HF conversion (#6745)
b8109bc0139f15a5b321909f47510b89dca47ffc aed82f6837a3ea515f4d50201cfc77effc7d41b4 Jan Boon jan.boon@kaetemi.be 2024-04-21T00:29:50+08:00 GitHub noreply@github.com 2024-04-20T18:29:50+02:00 doc : server tests require llama to be built with curl enabled (#6788)
aed82f6837a3ea515f4d50201cfc77effc7d41b4 0e4802b2ecbaab04b4f829fde4a3096ca19c84b5 Georgi Gerganov ggerganov@gmail.com 2024-04-20T13:27:12+03:00 GitHub noreply@github.com 2024-04-20T13:27:12+03:00 common : try to fix Android CI (#6780)
0e4802b2ecbaab04b4f829fde4a3096ca19c84b5 637e9a86c220718d008b54842dfd294aa96d3b7a loonerin 132926317+loonerin@users.noreply.github.com 2024-04-19T13:03:35-04:00 GitHub noreply@github.com 2024-04-19T19:03:35+02:00 ci: add ubuntu latest release and fix missing build number (mac & ubuntu) (#6748)
637e9a86c220718d008b54842dfd294aa96d3b7a 9958c81b798a5872087b30b360e4674871f2479e Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-19T13:19:01+02:00 GitHub noreply@github.com 2024-04-19T13:19:01+02:00 server: static: upstream upgrade (#6765)
9958c81b798a5872087b30b360e4674871f2479e 8b1b1f4982d3e9b994308d05a1c8b9e45c23edb5 nopperl 54780682+nopperl@users.noreply.github.com 2024-04-19T09:35:54Z GitHub noreply@github.com 2024-04-19T11:35:54+02:00 Implement the OLMo architecture (#6741)
8b1b1f4982d3e9b994308d05a1c8b9e45c23edb5 bca40e98149c7b673558ddd7a3ebeffef789349d Austin 77757836+teleprint-me@users.noreply.github.com 2024-04-19T03:16:45-04:00 GitHub noreply@github.com 2024-04-19T10:16:45+03:00 train : add general name (#6752)
bca40e98149c7b673558ddd7a3ebeffef789349d 0d56246f4b9764158525d894b96606f6163c53a8 Neo Zhang 14088817+arthw@users.noreply.github.com 2024-04-19T09:16:31+08:00 GitHub noreply@github.com 2024-04-19T09:16:31+08:00 fix wrong parameter in cmd in readme-sycl.md (#6755)
0d56246f4b9764158525d894b96606f6163c53a8 03c0946d73c63ea73e1d85015b7088298443d438 slaren slarengh@gmail.com 2024-04-18T15:18:48+02:00 GitHub noreply@github.com 2024-04-18T15:18:48+02:00 ggml : group all experts in a single ggml_mul_mat_id (#6505)
03c0946d73c63ea73e1d85015b7088298443d438 e11b2e6e1e18522ca7cf129600875a0f6fb9307d Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-04-18T13:49:01+02:00 GitHub noreply@github.com 2024-04-18T14:49:01+03:00 convert : support models with multiple chat templates (#6588)
e11b2e6e1e18522ca7cf129600875a0f6fb9307d c71bfd736ee99a56e697697b39240f2ee06ed26d Ren Xuancheng jklj077@users.noreply.github.com 2024-04-18T19:38:04+08:00 GitHub noreply@github.com 2024-04-18T14:38:04+03:00 Qwen2 : assume tied weights if lm_head/output weights is missing (#6738)
c71bfd736ee99a56e697697b39240f2ee06ed26d 3b8f1ec4b18770531d0b1d792f3edf08254e4f0c slaren slarengh@gmail.com 2024-04-18T09:04:47+02:00 GitHub noreply@github.com 2024-04-18T10:04:47+03:00 llama : fix compatibility with old 2 expert models (#6735)
3b8f1ec4b18770531d0b1d792f3edf08254e4f0c 8dd1ec8b3ffbfa2d26e82e672cea89f5eeb2f141 Georgi Gerganov ggerganov@gmail.com 2024-04-17T23:58:26+03:00 GitHub noreply@github.com 2024-04-17T23:58:26+03:00 llamafile : tmp disable + build sgemm.o when needed (#6716)
8dd1ec8b3ffbfa2d26e82e672cea89f5eeb2f141 facb8b56f8fd3bb10a693bf0943ae9d69d0828ef Yaroslav yaroslav.yashin@me.com 2024-04-17T14:47:50+02:00 GitHub noreply@github.com 2024-04-17T15:47:50+03:00 readme : add UI (#6724)
facb8b56f8fd3bb10a693bf0943ae9d69d0828ef 532c1737a14bb4b99747e6f460874947df37e450 Zheng.Deng 32841220+dengzheng-cloud@users.noreply.github.com 2024-04-17T04:51:07+08:00 GitHub noreply@github.com 2024-04-16T23:51:07+03:00 convert : fix autoawq gemma (#6704)
532c1737a14bb4b99747e6f460874947df37e450 666867b799ddd9da7dfdc905ece291ecf286effa Georgi Gerganov ggerganov@gmail.com 2024-04-16T23:50:38+03:00 GitHub noreply@github.com 2024-04-16T23:50:38+03:00 llama : make general.name optional (#6709)
666867b799ddd9da7dfdc905ece291ecf286effa 8cc91dc63c0df397d644a581b2cbeea74eb51ae0 Georgi Gerganov ggerganov@gmail.com 2024-04-16T23:50:22+03:00 GitHub noreply@github.com 2024-04-16T23:50:22+03:00 ggml : fix llamafile sgemm wdata offsets (#6710)
8cc91dc63c0df397d644a581b2cbeea74eb51ae0 dbceec87c0221ec952e69448df6a71f1372a7487 Justine Tunney jtunney@mozilla.com 2024-04-16T14:55:30-04:00 GitHub noreply@github.com 2024-04-16T21:55:30+03:00 ggml : add llamafile sgemm (#6414)
dbceec87c0221ec952e69448df6a71f1372a7487 f4dea7da1841a92d2788b0535063abf2f0e28461 Ashish 1856117+ashishdatta@users.noreply.github.com 2024-04-16T08:48:35-07:00 GitHub noreply@github.com 2024-04-16T18:48:35+03:00 llama : add StableLM2 12B (#6635)
f4dea7da1841a92d2788b0535063abf2f0e28461 8a56075b07a8b571bf95a912ffdce4c928c2b414 Shijie 821898965@qq.com 2024-04-16T23:40:48+08:00 GitHub noreply@github.com 2024-04-16T18:40:48+03:00 llama : add qwen2moe (#6074)
8a56075b07a8b571bf95a912ffdce4c928c2b414 58227ffdeb4fb89cacb0cffaadf76b1914324ad3 Daniel Bevenius daniel.bevenius@gmail.com 2024-04-16T08:34:06+02:00 GitHub noreply@github.com 2024-04-16T09:34:06+03:00 gritlm : add --outdir option to hf.sh script (#6699)
58227ffdeb4fb89cacb0cffaadf76b1914324ad3 4fbd8098e63670c6ae11a8adc350f5ba191cfda3 Georgi Gerganov ggerganov@gmail.com 2024-04-16T09:28:33+03:00 GitHub noreply@github.com 2024-04-16T09:28:33+03:00 perplexity : require positive --ctx-size arg (#6695)
4fbd8098e63670c6ae11a8adc350f5ba191cfda3 7593639ce335e8d7f89aa9a54d616951f273af60 Daniel Bevenius daniel.bevenius@gmail.com 2024-04-16T08:13:13+02:00 GitHub noreply@github.com 2024-04-16T09:13:13+03:00 gguf : add special tokens metadata for FIM/Infill (#6689)
7593639ce335e8d7f89aa9a54d616951f273af60 132f55795e51094954f1b1f647f97648be724a3a Olivier Chafik ochafik@users.noreply.github.com 2024-04-15T18:35:21+01:00 GitHub noreply@github.com 2024-04-15T18:35:21+01:00 `main`: add --json-schema / -j flag (#6659)
132f55795e51094954f1b1f647f97648be724a3a 3272896d79465fd2befef3425dac8e83933b7ea4 compilade git@compilade.net 2024-04-15T08:56:55-04:00 GitHub noreply@github.com 2024-04-15T15:56:55+03:00 llama : fix restoring the number of outputs from state files (#6687)
3272896d79465fd2befef3425dac8e83933b7ea4 7fc16a2c32650d46e79b382ecc6720f58c82f31b Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-15T14:18:47+02:00 GitHub noreply@github.com 2024-04-15T15:18:47+03:00 server : revert "minor layout improvements" (#6684)
7fc16a2c32650d46e79b382ecc6720f58c82f31b 17e98d4c96a583d420f12046bc92102381dbd28e Steven Prichard spprichard20@gmail.com 2024-04-15T05:14:46-05:00 GitHub noreply@github.com 2024-04-15T13:14:46+03:00 swift : linux support (#6590)
17e98d4c96a583d420f12046bc92102381dbd28e 1958f7e06ca2d2e3ab5698cc67513ba359144d8e Neo Zhang Jianyu jianyu.zhang@intel.com 2024-04-15T17:12:26+08:00 GitHub noreply@github.com 2024-04-15T17:12:26+08:00 fix mul_mat_id() for new input, make the ut pass (#6682)
1958f7e06ca2d2e3ab5698cc67513ba359144d8e 04fbc5f23e9708136ff03ebe194c7a7e965a7ca4 David Renshaw dwrenshaw@gmail.com 2024-04-14T15:24:15-04:00 GitHub noreply@github.com 2024-04-14T15:24:15-04:00 llama : add missing kv clear in llama_beam_search (#6664)
04fbc5f23e9708136ff03ebe194c7a7e965a7ca4 f184dd920852d6d372b754f871ee06cfe6f977ad Chao Jiang jc19chaoj@zoho.com 2024-04-15T00:16:34+08:00 GitHub noreply@github.com 2024-04-14T18:16:34+02:00 Add Command R chat template (#6650)
f184dd920852d6d372b754f871ee06cfe6f977ad 422c2aff1c9735853c9d8f5162104e41a364adc4 Georgi Gerganov ggerganov@gmail.com 2024-04-14T16:55:30+03:00 GitHub noreply@github.com 2024-04-14T06:55:30-07:00 flake.lock: Update (#6669)
422c2aff1c9735853c9d8f5162104e41a364adc4 8800226d65d5c98cd34eede6a6c05c78405c52da Dave dave-fl@users.noreply.github.com 2024-04-14T07:14:19-04:00 GitHub noreply@github.com 2024-04-14T13:14:19+02:00 Added support for GGML_OP_CLAMP in Metal (#6662)
8800226d65d5c98cd34eede6a6c05c78405c52da e689fc4e912feb19085be6894f475a873759cbfe Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-04-14T13:12:59+02:00 GitHub noreply@github.com 2024-04-14T13:12:59+02:00 Fix --split-max-size (#6655)
e689fc4e912feb19085be6894f475a873759cbfe a4ec34e1cd68560e7ff16e21ddc6962b6abd3d1d Jaemin Son woalsdnd@gmail.com 2024-04-14T20:12:36+09:00 GitHub noreply@github.com 2024-04-14T13:12:36+02:00 [bug fix] convert github repository_owner to lowercase (#6673)
a4ec34e1cd68560e7ff16e21ddc6962b6abd3d1d de17e3f7455dc7fd298cc61d86798533b9ca7a29 James A Capozzoli 157492257+jac-jim@users.noreply.github.com 2024-04-14T04:40:18-04:00 GitHub noreply@github.com 2024-04-14T11:40:18+03:00 convert : enable the `--use-temp-file` cli flag (#6645)
de17e3f7455dc7fd298cc61d86798533b9ca7a29 b5e7285baffb0da8a6619567b52d8e67de41291d Neo Zhang Jianyu jianyu.zhang@intel.com 2024-04-14T10:42:29+08:00 GitHub noreply@github.com 2024-04-14T10:42:29+08:00 fix memcpy() crash, add missed cmd in guide, fix softmax (#6622)
b5e7285baffb0da8a6619567b52d8e67de41291d 4bd0f93e4ab4fe6682e7d0241c1bdec1397e954a Johannes Gäßler johannesg@5d6.de 2024-04-14T00:21:55+02:00 GitHub noreply@github.com 2024-04-14T00:21:55+02:00 CUDA: fix matrix multiplication logic for tests (#6667)
4bd0f93e4ab4fe6682e7d0241c1bdec1397e954a ab9a3240a9da941fdef5cd4a25f2b97c2f5a67aa Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-13T11:33:52+02:00 GitHub noreply@github.com 2024-04-13T11:33:52+02:00 model: support arch `DbrxForCausalLM` (#6515)
ab9a3240a9da941fdef5cd4a25f2b97c2f5a67aa fbbc030ba93561fac842af994c5c6c4c1147f13b Olivier Chafik ochafik@users.noreply.github.com 2024-04-12T19:43:38+01:00 GitHub noreply@github.com 2024-04-12T19:43:38+01:00 JSON schema conversion: ⚡️ faster repetitions, min/maxLength for strings, cap number length (#6555)
fbbc030ba93561fac842af994c5c6c4c1147f13b 4cc120c7443cf9dab898736f3c3b45dc8f14672b slaren slarengh@gmail.com 2024-04-12T18:13:20+02:00 GitHub noreply@github.com 2024-04-12T18:13:20+02:00 metal : unify mul_mv_id kernels (#6556)
4cc120c7443cf9dab898736f3c3b45dc8f14672b 24ee66ed0d908d156bd0d1747b63a636a495cd7a Daniel Bevenius daniel.bevenius@gmail.com 2024-04-12T14:11:46+02:00 GitHub noreply@github.com 2024-04-12T15:11:46+03:00 infill : add download instructions for model (#6626)
24ee66ed0d908d156bd0d1747b63a636a495cd7a 91c736015b66ba1d0b82cbae6313b6d5eaa61b68 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-12T13:49:21+02:00 GitHub noreply@github.com 2024-04-12T14:49:21+03:00 server : coherent log output for KV cache full (#6637)
91c736015b66ba1d0b82cbae6313b6d5eaa61b68 5c4d767ac028c0f9c31cba3fceaf765c6097abfc jiez 373447296@qq.com 2024-04-12T18:45:06+08:00 GitHub noreply@github.com 2024-04-12T13:45:06+03:00 llama : add gguf_remove_key + remove split meta during quantize (#6591)
5c4d767ac028c0f9c31cba3fceaf765c6097abfc ef21ce4ccb41164cb52997bd2210d92bc6a6c5d1 Rene Leonhardt 65483435+reneleonhardt@users.noreply.github.com 2024-04-12T10:52:36+02:00 GitHub noreply@github.com 2024-04-12T10:52:36+02:00 chore: Fix markdown warnings (#6625)
ef21ce4ccb41164cb52997bd2210d92bc6a6c5d1 dee7f8d6928cc680cc969f7d93f98c3e24dcad41 Georgi Gerganov ggerganov@gmail.com 2024-04-12T11:49:58+03:00 GitHub noreply@github.com 2024-04-12T11:49:58+03:00 imatrix : remove invalid assert (#6632)
dee7f8d6928cc680cc969f7d93f98c3e24dcad41 81da18e71ccfc196d4516fbea5dc3a6a1f92dccb MasterYi1024 39848311+MasterYi1024@users.noreply.github.com 2024-04-12T16:28:12+08:00 GitHub noreply@github.com 2024-04-12T10:28:12+02:00 Correct free memory and total memory. (#6630)
81da18e71ccfc196d4516fbea5dc3a6a1f92dccb 9ed2737acc233716374860e6b2ea7399c4aae29e Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-12T10:26:47+02:00 GitHub noreply@github.com 2024-04-12T10:26:47+02:00 eval-callback: use ggml_op_desc to pretty print unary operator name (#6631)
9ed2737acc233716374860e6b2ea7399c4aae29e 04a5ac211ef40936295980b7cdf0ba6e97093146 Georgi Gerganov ggerganov@gmail.com 2024-04-12T11:15:05+03:00 GitHub noreply@github.com 2024-04-12T11:15:05+03:00 ci : disable Metal for macOS-latest-cmake-x64 (#6628)
04a5ac211ef40936295980b7cdf0ba6e97093146 f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7 Clint Herron hanclinto@gmail.com 2024-04-11T21:44:50-04:00 GitHub noreply@github.com 2024-04-11T21:44:50-04:00 Optimization: eliminate addition of redundant stacks when advancing grammar. (#6616)
f7001ccc5aa359fcf41bba19d1c99c3d25c9bcc7 a474f50ebb3e10be3371562f75f3f573f1a86b5f Clint Herron hanclinto@gmail.com 2024-04-11T17:44:48-04:00 GitHub noreply@github.com 2024-04-11T17:44:48-04:00 As suggested by @slaren, disabling Metal for test to fix CI build on OSX from #6576 (#6619)
a474f50ebb3e10be3371562f75f3f573f1a86b5f cbaadc92942c50aab599a9e4c163afc1f44f7c26 Nikolas 127742645+nneubacher@users.noreply.github.com 2024-04-11T21:56:29+02:00 GitHub noreply@github.com 2024-04-11T21:56:29+02:00 Refactor Error Handling for CUDA (#6575)
cbaadc92942c50aab599a9e4c163afc1f44f7c26 1bbdaf6ecda6f0a360dfb307b256fcb6838c560b Olivier Chafik ochafik@users.noreply.github.com 2024-04-11T19:47:34+01:00 GitHub noreply@github.com 2024-04-11T19:47:34+01:00 grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609)
1bbdaf6ecda6f0a360dfb307b256fcb6838c560b f4183afe6a22f356ee222a710686ae7f83dbd949 Hugo Roussel hugo.rous@gmail.com 2024-04-11T19:52:21+02:00 GitHub noreply@github.com 2024-04-11T19:52:21+02:00 ci: download artifacts to release directory (#6612)
f4183afe6a22f356ee222a710686ae7f83dbd949 b804b1ef77351d2a11be945462c6c251710476cb Daniel Bevenius daniel.bevenius@gmail.com 2024-04-11T15:22:47+02:00 GitHub noreply@github.com 2024-04-11T16:22:47+03:00 scripts : add --outdir option to hf.sh (#6600)
b804b1ef77351d2a11be945462c6c251710476cb 8228b66dbc16290c5cbd70e80ab47c068e2569d8 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-11T14:51:07+02:00 GitHub noreply@github.com 2024-04-11T14:51:07+02:00 eval-callback: Example how to use eval callback for debugging (#6576)
8228b66dbc16290c5cbd70e80ab47c068e2569d8 b3a96f27f065a828f08c5d89ff60aab5361188fe Daniel Bevenius daniel.bevenius@gmail.com 2024-04-10T20:16:48+02:00 GitHub noreply@github.com 2024-04-10T21:16:48+03:00 gguf : add option to not check tensor data (#6582)
b3a96f27f065a828f08c5d89ff60aab5361188fe 4f407a0a353dae4726c74cc33250b623a4911dd7 Ralph Soika ralph.soika@imixs.com 2024-04-10T19:18:25+02:00 GitHub noreply@github.com 2024-04-10T19:18:25+02:00 minor layout improvements (#6572)
4f407a0a353dae4726c74cc33250b623a4911dd7 65c64dc36f9bca5b3f100614cdd02bf12d6b3e49 slaren slarengh@gmail.com 2024-04-10T17:24:14+02:00 GitHub noreply@github.com 2024-04-10T17:24:14+02:00 llama : add model types for mixtral (#6589)
65c64dc36f9bca5b3f100614cdd02bf12d6b3e49 67fac4b95fcccfda8ab965e9ba4992a9ddf3a25f slaren slarengh@gmail.com 2024-04-10T15:23:12+02:00 GitHub noreply@github.com 2024-04-10T15:23:12+02:00 convert.py : add consolidated.safetensors for mixtral 8x22b (#6587)
67fac4b95fcccfda8ab965e9ba4992a9ddf3a25f 29122d32ac8075ad27b7a8be05dcad2b7e7a5f9e Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-10T08:58:48+02:00 GitHub noreply@github.com 2024-04-10T09:58:48+03:00 docs : how to add a model (#6565)
29122d32ac8075ad27b7a8be05dcad2b7e7a5f9e b231b37b095f172b26b1300ca442a147ea048b64 Artem Zinnatullin ceo@abstractny.gay 2024-04-10T00:49:12-06:00 GitHub noreply@github.com 2024-04-10T09:49:12+03:00 readme : fix ROCm link (#6579)
b231b37b095f172b26b1300ca442a147ea048b64 ba5e134e073ec6837078c874aba44a702944a676 sjxx 63994076+ylsdamxssjxxdd@users.noreply.github.com 2024-04-10T14:34:00+08:00 GitHub noreply@github.com 2024-04-10T09:34:00+03:00 readme : update UI list (#6560)
ba5e134e073ec6837078c874aba44a702944a676 1b67731e184e27a465b8c5476061294a4af668ea Jiří Sejkora Sejseloid@gmail.com 2024-04-10T00:23:02+02:00 GitHub noreply@github.com 2024-04-10T00:23:02+02:00 readme: fix typo in amdgpu target name (#6573)
1b67731e184e27a465b8c5476061294a4af668ea c4a3a4ff47d62d2503ddf9bd91b58c21f04fe3c3 Jared Van Bortel jared@nomic.ai 2024-04-09T13:44:08-04:00 GitHub noreply@github.com 2024-04-09T13:44:08-04:00 BERT tokenizer fixes (#6498)
c4a3a4ff47d62d2503ddf9bd91b58c21f04fe3c3 400d5d722d7edf7de0cf24a18c42b183c65047d2 Georgi Gerganov ggerganov@gmail.com 2024-04-09T20:29:06+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-09T20:29:06+03:00 sync : ggml
400d5d722d7edf7de0cf24a18c42b183c65047d2 5dc9dd7152dedc6046b646855585bd070c91e8c8 Ed Lee edilee@mozilla.com 2024-04-09T01:31:47-07:00 GitHub noreply@github.com 2024-04-09T10:31:47+02:00 server : detect search query to start webchat (#6554)
5dc9dd7152dedc6046b646855585bd070c91e8c8 e11a8999b5690f810c2c99c14347f0834e68c524 Carolinabanana 140120812+Carolinabanana@users.noreply.github.com 2024-04-09T09:16:13+01:00 GitHub noreply@github.com 2024-04-09T11:16:13+03:00 llama : add Command R Plus support (#6491)
e11a8999b5690f810c2c99c14347f0834e68c524 cc4a95426d17417d3c83f12bdb514fbe8abe2a88 Georgi Gerganov ggerganov@gmail.com 2024-04-09T09:23:19+03:00 GitHub noreply@github.com 2024-04-09T09:23:19+03:00 license : update copyright notice + add AUTHORS (#6405)
cc4a95426d17417d3c83f12bdb514fbe8abe2a88 cecd8d3c98b48f51aaa1d4c729e55bd319f6799c Georgi Gerganov ggerganov@gmail.com 2024-04-08T22:25:49+03:00 GitHub noreply@github.com 2024-04-08T22:25:49+03:00 llama : fix attention layer count sanity check (#6550)
cecd8d3c98b48f51aaa1d4c729e55bd319f6799c b73e564b16086845a8b4fffd26e22685d3e0c3db kunnis kunnis@users.noreply.github.com 2024-04-08T10:44:19-05:00 GitHub noreply@github.com 2024-04-08T17:44:19+02:00 Comment explaining a decision (#6531)
b73e564b16086845a8b4fffd26e22685d3e0c3db e3c337d87ca650972105a51c6ce302dd236c07ad Georgi Gerganov ggerganov@gmail.com 2024-04-08T16:23:01+03:00 GitHub noreply@github.com 2024-04-08T16:23:01+03:00 quantize : fix precedence of cli args (#6541)
e3c337d87ca650972105a51c6ce302dd236c07ad beea6e1b16e783a0886e78dec01002a8c00db24d Rick G 26732651+TheFlipbook@users.noreply.github.com 2024-04-08T06:02:30-07:00 GitHub noreply@github.com 2024-04-08T16:02:30+03:00 llama : support negative ith in llama_get_ API (#6519)
beea6e1b16e783a0886e78dec01002a8c00db24d 87fb5b4234d4b9c56ac94cf7aa229c8fd7defdb0 Jan Boon jan.boon@kaetemi.be 2024-04-08T20:43:30+08:00 GitHub noreply@github.com 2024-04-08T15:43:30+03:00 llama : save and restore kv cache for single seq id (#6341)
87fb5b4234d4b9c56ac94cf7aa229c8fd7defdb0 d752327c3338d5b9634121d651c0105f2c933f9b Abhilash Majumder 30946547+abhilash1910@users.noreply.github.com 2024-04-08T13:56:01+05:30 GitHub noreply@github.com 2024-04-08T16:26:01+08:00 remove row=1 cond (#6532)
d752327c3338d5b9634121d651c0105f2c933f9b 855f54402e866ed19d8d675b56a81c844c64b325 Firat firatkiral@gmail.com 2024-04-08T00:48:29-07:00 GitHub noreply@github.com 2024-04-08T09:48:29+02:00 Adding KodiBot to UI list (#6535)
855f54402e866ed19d8d675b56a81c844c64b325 b909236c0bf0b6e872af95df9490492ecec310ac Mark Fairbairn thebaron88@gmail.com 2024-04-07T19:52:19+01:00 GitHub noreply@github.com 2024-04-07T20:52:19+02:00 Change Windows AMD example to release build to make inference much faster. (#6525)
b909236c0bf0b6e872af95df9490492ecec310ac e0717e751e12af13f4eedaae8bbbd608e40d7e54 Georgi Gerganov ggerganov@gmail.com 2024-04-07T21:25:30+03:00 GitHub noreply@github.com 2024-04-07T11:25:30-07:00 flake.lock: Update (#6517)
e0717e751e12af13f4eedaae8bbbd608e40d7e54 c37247796b4d45bdbbc8259afffb80208ad8fe55 DAN™ dranger003@gmail.com 2024-04-07T13:33:59-04:00 GitHub noreply@github.com 2024-04-07T19:33:59+02:00 Add GritLM as supported models. (#6513)
c37247796b4d45bdbbc8259afffb80208ad8fe55 f77261a7c525fa1fa47b18a3d78cd308ae41cafc Georgi Gerganov ggerganov@gmail.com 2024-04-07T17:05:51+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-07T17:05:51+03:00 sync : ggml
f77261a7c525fa1fa47b18a3d78cd308ae41cafc 43e8995e754b8e04642e92822055d193a3272b37 Slava Primenko primenko.s@gmail.com 2024-04-04T14:49:24+02:00 Georgi Gerganov ggerganov@gmail.com 2024-04-07T17:05:40+03:00 ggml: bypass code incompatible with CUDA < 11.1 (whisper/2020)
43e8995e754b8e04642e92822055d193a3272b37 9472bce30800a581071478a839bf93abf404c893 Georgi Gerganov ggerganov@gmail.com 2024-04-07T16:08:12+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-07T16:08:12+03:00 scripts : sync ggml-cuda folder
9472bce30800a581071478a839bf93abf404c893 d4f220a5ccdc6308173c1a31fad21d7c3fbc96c1 limitedAtonement limitedAtonement@users.noreply.github.com 2024-04-07T07:05:40-04:00 GitHub noreply@github.com 2024-04-07T13:05:40+02:00 Run make to build the project (#6457)
d4f220a5ccdc6308173c1a31fad21d7c3fbc96c1 54ea0698fbf87e36a5d68a98c95f6bdd0fb91557 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-04-07T10:55:59+08:00 GitHub noreply@github.com 2024-04-07T10:55:59+08:00 support/fix OPs GGML_TYPE_IQ4_NL, GGML_TYPE_IQ4_XS, GGML_TYPE_IQ3_XXS, GGML_TYPE_IQ3_S, GGML_TYPE_IQ2_XXS, GGML_TYPE_IQ2_XS, GGML_TYPE_IQ2_S, GGML_TYPE_IQ1_S, GGML_TYPE_IQ1_M (#6521)
54ea0698fbf87e36a5d68a98c95f6bdd0fb91557 b66aec675c1571a06b3570b858ae711246f96f84 Georgi Gerganov ggerganov@gmail.com 2024-04-06T17:43:15+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-06T18:27:46+03:00 sync : ggml
b66aec675c1571a06b3570b858ae711246f96f84 57dd02c44b2a0eb79e28f6c5eb8242a5d2d3174d Daniel Bevenius daniel.bevenius@gmail.com 2024-04-03T22:57:20+02:00 Georgi Gerganov ggerganov@gmail.com 2024-04-06T17:42:26+03:00 backend : fix typo in scheduler documentation (ggml/781)
57dd02c44b2a0eb79e28f6c5eb8242a5d2d3174d 75cd4c77292034ecec587ecb401366f57338f7c0 Clint Herron hanclinto@gmail.com 2024-04-06T10:31:33-04:00 GitHub noreply@github.com 2024-04-06T10:31:33-04:00 Tests: Added integration tests for GBNF parser (#6472)
75cd4c77292034ecec587ecb401366f57338f7c0 a8bd14d55717754a1f48313a846a2b16fa998ad2 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-06T05:40:47+02:00 GitHub noreply@github.com 2024-04-06T05:40:47+02:00 ci: bench: support sse and fix prompt processing time / server: add tokens usage in stream OAI response (#6495)
a8bd14d55717754a1f48313a846a2b16fa998ad2 d0f5deebf898f8186a10148a03a56909ba05fc0b Brian mofosyne@gmail.com 2024-04-06T05:41:38+11:00 GitHub noreply@github.com 2024-04-05T21:41:38+03:00 gguf.py : add licence and version to gguf writer (#6504)
d0f5deebf898f8186a10148a03a56909ba05fc0b 87e21bbacd830437ab653cf03b6f26d45c15395d Hoang Nguyen hugo53@users.noreply.github.com 2024-04-05T11:39:43-07:00 GitHub noreply@github.com 2024-04-05T21:39:43+03:00 readme : update UI list (#6503)
87e21bbacd830437ab653cf03b6f26d45c15395d 1b496a745c315022df2d919374052e6004ced8d3 Ting Sun suntcrick@gmail.com 2024-04-06T01:34:53+07:00 GitHub noreply@github.com 2024-04-05T21:34:53+03:00 bench : make n_batch and n_ubatch configurable in Batched bench (#6500)
1b496a745c315022df2d919374052e6004ced8d3 a307375c02cac45cff53cf2520330b43fecc7718 Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-04-05T14:35:06+01:00 GitHub noreply@github.com 2024-04-05T19:05:06+05:30 [SYCL] Fixed minor bug when enabling FP16 for non intel targets (#6464)
a307375c02cac45cff53cf2520330b43fecc7718 b660a5729e1e7508671d3d0515fd7efaeaeb85b9 alexpinel 93524949+alexpinel@users.noreply.github.com 2024-04-04T18:22:50+01:00 GitHub noreply@github.com 2024-04-04T13:22:50-04:00 readme : add Dot to UI list (#6487)
b660a5729e1e7508671d3d0515fd7efaeaeb85b9 0a1d889e27d6aaa3293dd2c692b849a9bcf4b474 Jun Jie 71215065+junnjiee16@users.noreply.github.com 2024-04-05T01:16:37+08:00 GitHub noreply@github.com 2024-04-04T13:16:37-04:00 readme : fix typo (#6481)
0a1d889e27d6aaa3293dd2c692b849a9bcf4b474 7dda1b727ef1730783a6077136d28b83e70dd397 Ed Lepedus ed.lepedus@googlemail.com 2024-04-04T17:31:22+01:00 GitHub noreply@github.com 2024-04-04T18:31:22+02:00 server: add cURL support to server Dockerfiles (#6474)
7dda1b727ef1730783a6077136d28b83e70dd397 c666ba26c39d5c07e07b4e1e411332f408e309ad Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-04-05T01:30:53+09:00 GitHub noreply@github.com 2024-04-04T18:30:53+02:00 ci: exempt master branch workflows from getting cancelled (#6486)
c666ba26c39d5c07e07b4e1e411332f408e309ad 2e66913e5f56209f4c949f98e431925b78e7e84d Ewout ter Hoeven E.M.terHoeven@student.tudelft.nl 2024-04-04T17:08:55+02:00 GitHub noreply@github.com 2024-04-04T17:08:55+02:00 build CI: Name artifacts (#6482)
2e66913e5f56209f4c949f98e431925b78e7e84d 8120efee1d9931b514aeb5a047209d576f23286c Shakhar Dasgupta shakhardasgupta@gmail.com 2024-04-04T11:03:00-04:00 GitHub noreply@github.com 2024-04-04T17:03:00+02:00 server: allow penalizing repetition of newlines on server webpage (#6431)
8120efee1d9931b514aeb5a047209d576f23286c a74401f0e5ebb15fa4d8b6619d1baa6ea9179123 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-04T16:59:04+02:00 GitHub noreply@github.com 2024-04-04T16:59:04+02:00 ci: bench fix concurrency for workflow trigger dispatch with sha1 (#6478)
a74401f0e5ebb15fa4d8b6619d1baa6ea9179123 7a2c92637ae265654a68f62e6a7610b358255d3f limitedAtonement limitedAtonement@users.noreply.github.com 2024-04-04T10:30:02-04:00 GitHub noreply@github.com 2024-04-04T16:30:02+02:00 Correct README link (#6458)
7a2c92637ae265654a68f62e6a7610b358255d3f 4bcd6b959ca3991084ad1d8464caf2a734e29b1d Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-04T11:57:58+02:00 GitHub noreply@github.com 2024-04-04T12:57:58+03:00 ci: bench: add more ftype, fix triggers and bot comment (#6466)
4bcd6b959ca3991084ad1d8464caf2a734e29b1d 9b84ae1806cded4d6683c7b810925da5ead40607 Daniel Bevenius daniel.bevenius@gmail.com 2024-04-04T09:49:21+02:00 GitHub noreply@github.com 2024-04-04T09:49:21+02:00 common: remove duplicate check for curl (#6471)
9b84ae1806cded4d6683c7b810925da5ead40607 4399f13fb9462cd06f3f154d0aee738425000fea Clint Herron hanclinto@gmail.com 2024-04-04T03:44:28-04:00 GitHub noreply@github.com 2024-04-04T10:44:28+03:00 examples : add GBNF validator program (#5948)
4399f13fb9462cd06f3f154d0aee738425000fea 1a43c7254ed5de91d09274b853db843c518f1b64 Georgi Gerganov ggerganov@gmail.com 2024-04-04T09:34:58+03:00 Georgi Gerganov ggerganov@gmail.com 2024-04-04T09:34:58+03:00 server : remove obsolete --memory-f32 option
1a43c7254ed5de91d09274b853db843c518f1b64 72d73af65132792a52433952ca4729b01c36cde2 Xiao-Yong Jin jinxiaoyong@gmail.com 2024-04-04T01:33:48-05:00 GitHub noreply@github.com 2024-04-04T09:33:48+03:00 server : add option to disable KV offload (#6468)
72d73af65132792a52433952ca4729b01c36cde2 5fb1574c8112c757fc202fdae279da883f34e610 Clint Herron hanclinto@gmail.com 2024-04-04T02:32:53-04:00 GitHub noreply@github.com 2024-04-04T09:32:53+03:00 convert : fix for lint error complaining of bare except (#6470)
5fb1574c8112c757fc202fdae279da883f34e610 60cdf40cc32f0ad4cb11e0ca8fd38f3b93d8d640 Fattire 528174+fat-tire@users.noreply.github.com 2024-04-03T13:22:57-07:00 GitHub noreply@github.com 2024-04-03T22:22:57+02:00 A few small fixes to server's README docs (#6428)
60cdf40cc32f0ad4cb11e0ca8fd38f3b93d8d640 bb43cf7e9d86d69ffd9c7f008f75db890a35b45a JH23X 165871467+JH23X@users.noreply.github.com 2024-04-03T20:09:52+02:00 GitHub noreply@github.com 2024-04-03T21:09:52+03:00 server : handle exception on wrong type in request (#6452)
bb43cf7e9d86d69ffd9c7f008f75db890a35b45a 9f62c0173d964972849251c8ad12fc356f5b7896 bryanSwk 93190252+bryanSwk@users.noreply.github.com 2024-04-04T02:05:10+08:00 GitHub noreply@github.com 2024-04-03T21:05:10+03:00 llama : add SEA-LION support (#6448)
9f62c0173d964972849251c8ad12fc356f5b7896 5d4f12e4624bf4435ba26753814bedd4e9b62803 Ewout ter Hoeven E.M.terHoeven@student.tudelft.nl 2024-04-03T20:01:13+02:00 GitHub noreply@github.com 2024-04-03T21:01:13+03:00 ci : update checkout, setup-python and upload-artifact to latest (#6456)
5d4f12e4624bf4435ba26753814bedd4e9b62803 154d4ee39c38e231fb5c3910df14d2fc920fdf1b Ed Lepedus ed.lepedus@googlemail.com 2024-04-03T18:56:37+01:00 GitHub noreply@github.com 2024-04-03T19:56:37+02:00 server: add cURL support to `server.Dockerfile` (#6461)
154d4ee39c38e231fb5c3910df14d2fc920fdf1b e69945d953b651674d6e6978022edf00c3a34d03 Francisco Melo 43780565+francis2tm@users.noreply.github.com 2024-04-03T18:53:37+01:00 GitHub noreply@github.com 2024-04-03T20:53:37+03:00 readme : add feature-rich rust bindings (#6465)
e69945d953b651674d6e6978022edf00c3a34d03 db214fa578e00b01e0884fc2725c9349608bdab5 Joyce joycebrum@google.com 2024-04-03T14:48:07-03:00 GitHub noreply@github.com 2024-04-03T20:48:07+03:00 security : create policy (#6354)
db214fa578e00b01e0884fc2725c9349608bdab5 1ff4d9f3d683f02ef8a12e04bfac84300c44bd3a Abhishek Gopinath K 31348521+overtunned@users.noreply.github.com 2024-04-03T21:12:52+05:30 GitHub noreply@github.com 2024-04-03T11:42:52-04:00 Missing tokenizer.model error during gguf conversion (#6443)
1ff4d9f3d683f02ef8a12e04bfac84300c44bd3a 076b08649ecc3b0e1c0709c2a086a63eddd1bf32 kaizau kaizau@users.noreply.github.com 2024-04-03T23:24:31+08:00 GitHub noreply@github.com 2024-04-03T17:24:31+02:00 Add OpenChat, Alpaca, Vicuna chat templates (#6397)
076b08649ecc3b0e1c0709c2a086a63eddd1bf32 08a0c0206075556e82aca0feafad530dcc5f1426 Georgi Gerganov ggerganov@gmail.com 2024-04-03T16:11:15+03:00 GitHub noreply@github.com 2024-04-03T16:11:15+03:00 readme : update hot topics
08a0c0206075556e82aca0feafad530dcc5f1426 52604860f93063ef98863921da697576af1c7665 slaren slarengh@gmail.com 2024-04-03T15:07:05+02:00 GitHub noreply@github.com 2024-04-03T16:07:05+03:00 ggml : mul_mat_id use the same tensor for all the experts (#6387)
52604860f93063ef98863921da697576af1c7665 f87f7b898651339fe173ddf016ca826163e899d8 Meng, Hengyu hengyu.meng@intel.com 2024-04-03T10:34:40+08:00 GitHub noreply@github.com 2024-04-03T10:34:40+08:00 [SYCL] Disable iqx on windows as WA (#6435)
f87f7b898651339fe173ddf016ca826163e899d8 33a52448061cfd2ea44da9e6cb30b2ec22e2f6d0 Georgi Gerganov ggerganov@gmail.com 2024-04-01T19:05:57+03:00 GitHub noreply@github.com 2024-04-01T09:05:57-07:00 flake.lock: Update (#6402)
33a52448061cfd2ea44da9e6cb30b2ec22e2f6d0 226e819371eec0f298d9075198394a07b23ecfa9 Johannes Gäßler johannesg@5d6.de 2024-04-01T13:30:43+02:00 GitHub noreply@github.com 2024-04-01T13:30:43+02:00 compare-llama-bench.py: fix long hexsha args (#6424)
226e819371eec0f298d9075198394a07b23ecfa9 c50a82ce0f71558cbb8e555146ba124251504b38 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-04-01T12:36:40+02:00 GitHub noreply@github.com 2024-04-01T12:36:40+02:00 ci: server: verify deps are coherent with the commit (#6409)
c50a82ce0f71558cbb8e555146ba124251504b38 37e7854c104301c5b5323ccc40e07699f3a62c3e Georgi Gerganov ggerganov@gmail.com 2024-03-31T11:56:30+03:00 GitHub noreply@github.com 2024-03-31T11:56:30+03:00 readme : update hot topics
37e7854c104301c5b5323ccc40e07699f3a62c3e c342d070c64a1ffe35d22c1b16b672e684a30297 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-30T11:36:07+01:00 GitHub noreply@github.com 2024-03-30T12:36:07+02:00 ci: bench: fix Resource not accessible by integration on PR event (#6393)
c342d070c64a1ffe35d22c1b16b672e684a30297 f7fc5f6c6f3700da311de7fe93977c81798f02d3 Mohammadreza Hendiani hendiani.mohammadreza@gmail.com 2024-03-30T01:29:56+03:30 GitHub noreply@github.com 2024-03-29T22:59:56+01:00 Fedora build update (#6388)
f7fc5f6c6f3700da311de7fe93977c81798f02d3 ba0c7c70ab5b15f1f2be7fb0dfbe0366dda30d6c Xuan Son Nguyen thichthat@gmail.com 2024-03-29T22:34:44+01:00 GitHub noreply@github.com 2024-03-29T22:34:44+01:00 split: allow --split-max-size option (#6343)
ba0c7c70ab5b15f1f2be7fb0dfbe0366dda30d6c d48ccf3ad4fea5b9ede209c7f40be65371987bfe 0cc4m picard12@live.de 2024-03-29T17:29:21+01:00 GitHub noreply@github.com 2024-03-29T17:29:21+01:00 Vulkan k-quant mmq and ggml-backend offload functionality (#6155)
d48ccf3ad4fea5b9ede209c7f40be65371987bfe 069574775cea67d8a977904e4d534aff47a671f4 Georgi Gerganov ggerganov@gmail.com 2024-03-29T17:45:46+02:00 GitHub noreply@github.com 2024-03-29T17:45:46+02:00 sync : ggml (#6351)
069574775cea67d8a977904e4d534aff47a671f4 cfde806eb95de06d84162bdee593dad33a1d2693 hxer7963 hxer7963@gmail.com 2024-03-29T21:37:03+08:00 GitHub noreply@github.com 2024-03-29T14:37:03+01:00 [Model] Add support for xverse (#6301)
cfde806eb95de06d84162bdee593dad33a1d2693 b910287954d4462fd3d48938336770e258459677 Georgi Gerganov ggerganov@gmail.com 2024-03-29T14:34:28+02:00 GitHub noreply@github.com 2024-03-29T14:34:28+02:00 ci : fix BGE wget (#6383)
b910287954d4462fd3d48938336770e258459677 809398709041ee854fbbad9b344bcfdcd3712d59 zhouwg 6889919+zhouwg@users.noreply.github.com 2024-03-29T15:33:46+08:00 GitHub noreply@github.com 2024-03-29T09:33:46+02:00 readme : add project (#6356)
809398709041ee854fbbad9b344bcfdcd3712d59 057400a3fd457f4f214684eeb171444663b47a23 Matt Clayton 156335168+mattjcly@users.noreply.github.com 2024-03-29T03:27:42-04:00 GitHub noreply@github.com 2024-03-29T09:27:42+02:00 cmake : add explicit metal version options (#6370)
057400a3fd457f4f214684eeb171444663b47a23 b75c38166cf0c66793a32d5c3d6cb69929dd083d Daniel Bevenius daniel.bevenius@gmail.com 2024-03-29T08:23:22+01:00 GitHub noreply@github.com 2024-03-29T09:23:22+02:00 llama : remove redundant reshape in build_kv_store (#6369)
b75c38166cf0c66793a32d5c3d6cb69929dd083d bfe7dafc9cf96b9a09ead347fed9a547930fc631 Pedro Cuenca pedro@huggingface.co 2024-03-29T08:15:00+01:00 GitHub noreply@github.com 2024-03-29T09:15:00+02:00 convert : allow conversion of Mistral HF models (#6144)
bfe7dafc9cf96b9a09ead347fed9a547930fc631 5106ef482c65ac60ac14da9a68c7b37bca4c6993 Georgi Gerganov ggerganov@gmail.com 2024-03-28T22:56:03+02:00 GitHub noreply@github.com 2024-03-28T22:56:03+02:00 readme : add notice for UI list
5106ef482c65ac60ac14da9a68c7b37bca4c6993 be55134a535f7218c53f39211755b1c7550851b2 Ouadie EL FAROUKI ouadie.elfarouki@codeplay.com 2024-03-28T16:01:47Z GitHub noreply@github.com 2024-03-28T16:01:47Z [SYCL] Revisited & updated SYCL build documentation (#6141)
be55134a535f7218c53f39211755b1c7550851b2 66ba56025602270152f5ba5234f3a80be3dee1c9 Jared Van Bortel jared@nomic.ai 2024-03-28T11:44:36-04:00 GitHub noreply@github.com 2024-03-28T11:44:36-04:00 convert : refactor vocab selection logic (#6355)
66ba56025602270152f5ba5234f3a80be3dee1c9 0308f5e3d7bf9879f818b1a4ae589ff36b242af5 Ziang Wu 97337387+ZiangWu-77@users.noreply.github.com 2024-03-28T22:33:10+08:00 GitHub noreply@github.com 2024-03-28T16:33:10+02:00 llava : fix MobileVLM (#6364)
0308f5e3d7bf9879f818b1a4ae589ff36b242af5 28cb9a09c4d10a489be1238abe7a858dcd4d65f2 compilade 113953597+compilade@users.noreply.github.com 2024-03-28T08:05:54-04:00 GitHub noreply@github.com 2024-03-28T14:05:54+02:00 llama : fix command-r inference when omitting outputs (#6367)
28cb9a09c4d10a489be1238abe7a858dcd4d65f2 cfc4d75df6399b36153ef739f2c1abee4c114bb8 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-28T11:27:56+01:00 GitHub noreply@github.com 2024-03-28T11:27:56+01:00 ci: bench: fix master not schedule, fix commit status failed on external repo (#6365)
cfc4d75df6399b36153ef739f2c1abee4c114bb8 6902cb7f2e3479f364ee177118200fb7e4e9fc92 Ting Sun suntcrick@gmail.com 2024-03-28T16:51:06+08:00 GitHub noreply@github.com 2024-03-28T09:51:06+01:00 doc: fix outdated default value of batch size (#6336)
6902cb7f2e3479f364ee177118200fb7e4e9fc92 d2d8f389960a106b66313ff1621bdb1aaaaaa285 Eric Zhang 34133756+EZForever@users.noreply.github.com 2024-03-28T16:50:48+08:00 GitHub noreply@github.com 2024-03-28T09:50:48+01:00 server : stop gracefully on SIGTERM (#6348)
d2d8f389960a106b66313ff1621bdb1aaaaaa285 d39b308eaf0ac91c2e1f432bf66751193a470a56 hutli hutli@hutli.hu 2024-03-27T19:17:30+01:00 Someone newkozlukov@gmail.com 2024-03-28T07:48:27Z nix: removed unnessesary indentation
d39b308eaf0ac91c2e1f432bf66751193a470a56 c87397664964e5a2a21de1877d504b23a2a35332 hutli hutli@hutli.hu 2024-03-27T19:14:28+01:00 Someone newkozlukov@gmail.com 2024-03-28T07:48:27Z nix: moved blas availability check to package inputs so it is still overridable
c87397664964e5a2a21de1877d504b23a2a35332 dbb03e2b9c4fead73062926b6a134f28d3a3b46d hutli hutli@hutli.hu 2024-03-27T18:10:08+01:00 Someone newkozlukov@gmail.com 2024-03-28T07:48:27Z using blas.meta.available to check host platform
dbb03e2b9c4fead73062926b6a134f28d3a3b46d e9f17dc3bf0da76c8b35130f9ca2fda5246c418e hutli jensstaermose@hotmail.com 2024-03-27T17:25:05+01:00 Someone newkozlukov@gmail.com 2024-03-28T07:48:27Z only using explicit blas if hostPlatform is allowed
e9f17dc3bf0da76c8b35130f9ca2fda5246c418e 22a462cc1f69873f7d4c6d0201bd93478afa2ecb Someone Serge sergei.kozlukov@aalto.fi 2024-03-26T16:22:42Z Someone newkozlukov@gmail.com 2024-03-28T07:48:27Z nix: .#windows: proper cross-compilation set-up
22a462cc1f69873f7d4c6d0201bd93478afa2ecb f6a0f5c6422200764b7929064c39dcf4bb0e9cd6 Someone Serge sergei.kozlukov@aalto.fi 2024-03-26T16:22:07Z Someone newkozlukov@gmail.com 2024-03-28T07:48:27Z nix: package: don't introduce the dependency on python
f6a0f5c6422200764b7929064c39dcf4bb0e9cd6 d0e2f6416bd43eddb70137f6b96c7bc3d0246102 hutli jensstaermose@hotmail.com 2024-02-15T14:25:04+01:00 Someone newkozlukov@gmail.com 2024-03-28T07:48:27Z nix: .#widnows: init
d0e2f6416bd43eddb70137f6b96c7bc3d0246102 25f4a613c4ed6451162a87cb90be10d610b49f0f Ziang Wu 97337387+ZiangWu-77@users.noreply.github.com 2024-03-28T12:03:30+08:00 GitHub noreply@github.com 2024-03-28T13:03:30+09:00 doc: fix typo in MobileVLM-README.md (#6181)
25f4a613c4ed6451162a87cb90be10d610b49f0f a016026a3ac16d8c9b993a3573f19b9556d67de4 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-28T08:55:24+08:00 GitHub noreply@github.com 2024-03-28T08:55:24+08:00 [SYCL] fix set main gpu crash (#6339)
a016026a3ac16d8c9b993a3573f19b9556d67de4 53c7ec53d5eca26b2c0c648605543a5fa6c12817 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-27T20:26:49+01:00 GitHub noreply@github.com 2024-03-27T20:26:49+01:00 server: continuous performance monitoring and PR comment (#6283)
53c7ec53d5eca26b2c0c648605543a5fa6c12817 e5b89a441af23a74b861b0bf8db3239139041876 Someone Serge sergei.kozlukov@aalto.fi 2024-03-27T16:17:46Z Someone newkozlukov@gmail.com 2024-03-27T19:18:55Z nix: ci: dont test cuda and rocm (for now)
e5b89a441af23a74b861b0bf8db3239139041876 3a0345970ed0353fa857df3c8a62a2b3318b1364 slaren slarengh@gmail.com 2024-03-27T15:07:50+01:00 GitHub noreply@github.com 2024-03-27T15:07:50+01:00 ggml : fix bounds checking of zero size views (#6347)
3a0345970ed0353fa857df3c8a62a2b3318b1364 1e13987fba5a536965ef942f2c86549d62cef50b Georgi Gerganov ggerganov@gmail.com 2024-03-27T15:02:49+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-27T15:02:49+02:00 make : whitespace
1e13987fba5a536965ef942f2c86549d62cef50b e82f9e2b833d88cd2b30123ef57346c2cb8abd99 howlger eclipse@voormann.de 2024-03-27T12:15:44+01:00 GitHub noreply@github.com 2024-03-27T13:15:44+02:00 embedding : show full embedding for single prompt (#6342)
e82f9e2b833d88cd2b30123ef57346c2cb8abd99 cbc83436197cde617cad696e665879c20df77daa AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-03-27T08:16:40Z GitHub noreply@github.com 2024-03-27T13:46:40+05:30 [SYCL] Fix batched impl for NVidia GPU (#6164)
cbc83436197cde617cad696e665879c20df77daa e562b9714b9b3e242361a7f74bbbeb00f6bd99ac Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-27T08:44:27+01:00 GitHub noreply@github.com 2024-03-27T08:44:27+01:00 Make IQ1_M work for QK_K = 64 (#6327)
e562b9714b9b3e242361a7f74bbbeb00f6bd99ac 2ab4f00d25c0682a74472412d66454df211e4b0e Sigbjørn Skjæret sigbjorn.skjaeret@scala.com 2024-03-27T08:23:10+01:00 GitHub noreply@github.com 2024-03-27T09:23:10+02:00 common : change --no-penalize-nl to --penalize-nl (#6334)
2ab4f00d25c0682a74472412d66454df211e4b0e 1740d6dd4e00dedda87d6820b0e0d8e70dade340 Georgi Gerganov ggerganov@gmail.com 2024-03-27T09:16:02+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-27T09:16:02+02:00 llama2c : open file as binary (#6332)
1740d6dd4e00dedda87d6820b0e0d8e70dade340 0642b22cd12af278d6e7e459b73411947c169381 Mateusz Charytoniuk mateusz.charytoniuk@protonmail.com 2024-03-27T08:08:59+01:00 GitHub noreply@github.com 2024-03-27T09:08:59+02:00 readme : add php api bindings (#6326)
0642b22cd12af278d6e7e459b73411947c169381 a4f569e8a316cbd33d2b4de94b694d111507475a Eric Zhang 34133756+EZForever@users.noreply.github.com 2024-03-27T13:55:29+08:00 GitHub noreply@github.com 2024-03-27T06:55:29+01:00 server: public: use relative routes for static files (#6325)
a4f569e8a316cbd33d2b4de94b694d111507475a 32c8486e1f0297393cb22ac0a0d26a6b17ad4d54 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-27T09:47:06+08:00 GitHub noreply@github.com 2024-03-27T09:47:06+08:00 [SYCL] fix no file in win rel (#6314)
32c8486e1f0297393cb22ac0a0d26a6b17ad4d54 557410b8f06380560155ac7fcb8316d71ddc9837 Jared Van Bortel jared@nomic.ai 2024-03-26T17:46:21-04:00 GitHub noreply@github.com 2024-03-26T17:46:21-04:00 wpm : portable unicode tolower (#6305)
557410b8f06380560155ac7fcb8316d71ddc9837 55c1b2a3bbd470e9e2a3a0618b92cf64a885f806 compilade 113953597+compilade@users.noreply.github.com 2024-03-26T10:46:41-04:00 GitHub noreply@github.com 2024-03-26T16:46:41+02:00 llama : greatly reduce output buffer memory usage (#6122)
55c1b2a3bbd470e9e2a3a0618b92cf64a885f806 e097633f63fdd26d492844f7eff056e4083fd9eb Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-26T15:21:27+01:00 GitHub noreply@github.com 2024-03-26T15:21:27+01:00 IQ1_M: 1.75 bpw quantization (#6302)
e097633f63fdd26d492844f7eff056e4083fd9eb d25b1c31b07c3675443a55a828dd58cfef5a241c Pedro Cuenca pedro@huggingface.co 2024-03-26T13:32:19+01:00 GitHub noreply@github.com 2024-03-26T14:32:19+02:00 convert-hf : fix exception in sentencepiece with added tokens (#6320)
d25b1c31b07c3675443a55a828dd58cfef5a241c deb7240100da99555b9ab9dc635021e591fceaf5 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-26T13:09:30+01:00 GitHub noreply@github.com 2024-03-26T14:09:30+02:00 quantize : be able to override metadata by key (#6321)
deb7240100da99555b9ab9dc635021e591fceaf5 3d032ece8e6973441273601ca2981130608e287d Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-26T18:11:46+09:00 GitHub noreply@github.com 2024-03-26T11:11:46+02:00 embedding : adjust `n_ubatch` value (#6296)
3d032ece8e6973441273601ca2981130608e287d e190f1fca6f60d80944f9e8709d343a025c4d245 Jan Boon jan.boon@kaetemi.be 2024-03-26T16:47:43+08:00 GitHub noreply@github.com 2024-03-26T10:47:43+02:00 server : add `n_discard` parameter (#6300)
e190f1fca6f60d80944f9e8709d343a025c4d245 280345968dabc00d212d43e31145f5c9961a7604 Joseph Stahl 1269177+josephst@users.noreply.github.com 2024-03-25T20:51:46-04:00 GitHub noreply@github.com 2024-03-25T17:51:46-07:00 nix: make `xcrun` visible in Nix sandbox for precompiling Metal shaders (#6118)
280345968dabc00d212d43e31145f5c9961a7604 b06c16ef9f81d84da520232c125d4d8a1d273736 slaren slarengh@gmail.com 2024-03-26T01:16:01+01:00 GitHub noreply@github.com 2024-03-26T01:16:01+01:00 cuda : rename build flag to LLAMA_CUDA (#6299)
b06c16ef9f81d84da520232c125d4d8a1d273736 1f2fd4e727a707f85d58e0b56075c3e6334d18d8 Christian Kögler ck3d@gmx.de 2024-03-25T18:52:45+01:00 GitHub noreply@github.com 2024-03-25T10:52:45-07:00 nix: fix blas support (#6281)
1f2fd4e727a707f85d58e0b56075c3e6334d18d8 43139cc528909c3de8c144ed174e09a1f7912a80 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-25T18:33:15+01:00 GitHub noreply@github.com 2024-03-25T19:33:15+02:00 tests : include IQ2_XXS and IQ2_XS in test-quantize-fns (#6303)
43139cc528909c3de8c144ed174e09a1f7912a80 2f34b865b62b1d2b5eb8a27885e4de220deeacbd Georgi Gerganov ggerganov@gmail.com 2024-03-25T17:22:27+02:00 GitHub noreply@github.com 2024-03-25T08:22:27-07:00 flake.lock: Update (#6266)
2f34b865b62b1d2b5eb8a27885e4de220deeacbd ae1f211ce2138448b47ebb148e25c58406845278 slaren slarengh@gmail.com 2024-03-25T15:43:22+01:00 GitHub noreply@github.com 2024-03-25T16:43:22+02:00 cuda : fix LLAMA_CUDA_F16 build (#6298)
ae1f211ce2138448b47ebb148e25c58406845278 ad3a0505e3b6cd777259ee35e61d428357ffc565 slaren slarengh@gmail.com 2024-03-25T13:50:23+01:00 GitHub noreply@github.com 2024-03-25T13:50:23+01:00 cuda : refactor into multiple files (#6269)
ad3a0505e3b6cd777259ee35e61d428357ffc565 95ad616cddda50273e955bfe192328acd9aa4896 Xuan Son Nguyen thichthat@gmail.com 2024-03-25T09:42:17+01:00 GitHub noreply@github.com 2024-03-25T09:42:17+01:00 Server: clean up OAI params parsing function (#6284)
95ad616cddda50273e955bfe192328acd9aa4896 64e7b47c6986221f2ff5c57c89dfc018bb0e9e6d Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-25T15:52:41+08:00 GitHub noreply@github.com 2024-03-25T15:52:41+08:00 [SYCL] fix SYCL backend build on windows is break by LOG() error (#6290)
64e7b47c6986221f2ff5c57c89dfc018bb0e9e6d 7733f0c76081b2a69b5f8b192db2db7c43629d58 Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-25T16:38:22+09:00 GitHub noreply@github.com 2024-03-25T09:38:22+02:00 examples : add "retrieval" (#6193)
7733f0c76081b2a69b5f8b192db2db7c43629d58 a32b77c4b2c1808654d0b952f26c37d73d2e746b Justine Tunney jtunney@gmail.com 2024-03-25T01:39:56-04:00 GitHub noreply@github.com 2024-03-25T07:39:56+02:00 ggml : support AVX512VNNI (#6280)
a32b77c4b2c1808654d0b952f26c37d73d2e746b a0e584defd8c16e7a51ab895f595df0448d710d0 Rick G 26732651+TheFlipbook@users.noreply.github.com 2024-03-24T14:45:56-07:00 GitHub noreply@github.com 2024-03-24T22:45:56+01:00 Fix heap corruption from wmode out-of-bound writes on windows (#6272)
a0e584defd8c16e7a51ab895f595df0448d710d0 7aed0ffe6855e9cadcf413f288753af4566bfeb8 Georgi Gerganov ggerganov@gmail.com 2024-03-24T16:18:45+02:00 GitHub noreply@github.com 2024-03-24T16:18:45+02:00 imatrix : fix wname for mul_mat_id ops (#6271)
7aed0ffe6855e9cadcf413f288753af4566bfeb8 ea279d56091b90fbbe063b598f5229d95f58ef68 Johannes Gäßler johannesg@5d6.de 2024-03-24T14:21:17+01:00 GitHub noreply@github.com 2024-03-24T14:21:17+01:00 Fixed lookup compilation issues on Windows (#6273)
ea279d56091b90fbbe063b598f5229d95f58ef68 586e7bc561be88e929a9afca7e67d8ead00c53bd Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-24T09:57:06+01:00 GitHub noreply@github.com 2024-03-24T10:57:06+02:00 ci : close inactive issue, increase operations per run (#6270)
586e7bc561be88e929a9afca7e67d8ead00c53bd ddf65685105a39a57b1e7f80c3aa502a6313af24 Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-24T17:54:07+09:00 GitHub noreply@github.com 2024-03-24T10:54:07+02:00 sampling : deduplicated code for probability distribution access (#6240)
ddf65685105a39a57b1e7f80c3aa502a6313af24 d03224ac9840351023ff8abcf4aa0542258a53df Meng, Hengyu hengyu.meng@intel.com 2024-03-24T12:04:25+08:00 GitHub noreply@github.com 2024-03-24T12:04:25+08:00 [SYCL] offload op (#6217)
d03224ac9840351023ff8abcf4aa0542258a53df 94d1b3b4119209efcdd08df0dceaecbd1fe7f85c Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-24T09:44:01+08:00 GitHub noreply@github.com 2024-03-24T09:44:01+08:00 Support build win release for SYCL (#6241)
94d1b3b4119209efcdd08df0dceaecbd1fe7f85c 95562175f83a49755ff6fd3bad09409417c8e6f9 Jared Van Bortel jared@nomic.ai 2024-03-23T18:48:02-04:00 GitHub noreply@github.com 2024-03-23T18:48:02-04:00 use _wfopen instead of fopen on Windows (#6248)
95562175f83a49755ff6fd3bad09409417c8e6f9 f482bb2e4920e544651fb832f2e0bcb4d2ff69ab Georgi Gerganov ggerganov@gmail.com 2024-03-23T21:35:23+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-23T21:35:23+02:00 gitignore : gguf-split
f482bb2e4920e544651fb832f2e0bcb4d2ff69ab 1997577d5e121568ae39f538021733ccd4278c23 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-23T18:07:00+01:00 GitHub noreply@github.com 2024-03-23T18:07:00+01:00 common: llama_load_model_from_url split support (#6192)
1997577d5e121568ae39f538021733ccd4278c23 476b0251b27fb64c575507024a671e639d675594 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-23T18:00:38+01:00 GitHub noreply@github.com 2024-03-23T18:00:38+01:00 server: docs: `--threads` and `--threads`, `--ubatch-size`, `--log-disable` (#6254)
476b0251b27fb64c575507024a671e639d675594 21cad01b6e6e1a96f99391f95e8ea8ae25c8288e Julius Arkenberg arki05@users.noreply.github.com 2024-03-23T17:41:53+01:00 GitHub noreply@github.com 2024-03-23T18:41:53+02:00 llama : add grok-1 support (#6204)
21cad01b6e6e1a96f99391f95e8ea8ae25c8288e 1b26aebe4de4f048ac99996efd8a2c9af150904d Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-23T17:18:13+01:00 GitHub noreply@github.com 2024-03-23T17:18:13+01:00 split: add gguf-split in the make build target (#6262)
1b26aebe4de4f048ac99996efd8a2c9af150904d 50ccaf5eacb50a2ca378a4ef0dc7aeb45fead652 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-23T13:18:45+01:00 GitHub noreply@github.com 2024-03-23T13:18:45+01:00 server: flush stdout after logging in both text and json layout (#6253)
50ccaf5eacb50a2ca378a4ef0dc7aeb45fead652 56a00f0a2f48a85376f48b5ce77699df781631ae Johannes Gäßler johannesg@5d6.de 2024-03-23T01:24:36+01:00 GitHub noreply@github.com 2024-03-23T01:24:36+01:00 lookup: complement data from context with general text statistics (#5479)
56a00f0a2f48a85376f48b5ce77699df781631ae 92397d87a45a09b5449d845a64856f177cd7a920 Georgi Gerganov ggerganov@gmail.com 2024-03-22T21:10:39+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-22T21:10:39+02:00 common : default --hf-file to --model (#6234)
92397d87a45a09b5449d845a64856f177cd7a920 1d0331c12a2f2a6296b471232bd4e66fbf06e6a1 fraxy-v 65565042+fraxy-v@users.noreply.github.com 2024-03-22T20:49:06+02:00 GitHub noreply@github.com 2024-03-22T20:49:06+02:00 convert-llama2c-to-ggml : enable conversion of GQA models (#6237)
1d0331c12a2f2a6296b471232bd4e66fbf06e6a1 dba1af612926cbd4ebe2d876277af1e3305177e0 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-22T19:47:14+01:00 GitHub noreply@github.com 2024-03-22T20:47:14+02:00 quantize: options for output and token embedding tensors qtype (#6239)
dba1af612926cbd4ebe2d876277af1e3305177e0 ee804f6223777019cf921e0d99cc24669313ab98 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-22T19:00:01+01:00 GitHub noreply@github.com 2024-03-22T19:00:01+01:00 llama_model_loader: support multiple split/shard GGUFs (#6187)
ee804f6223777019cf921e0d99cc24669313ab98 80bd33bc2c4be352697dc8473339f25e1085d117 Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-23T02:15:06+09:00 GitHub noreply@github.com 2024-03-22T19:15:06+02:00 ci: apply concurrency limit for github workflows (#6243)
80bd33bc2c4be352697dc8473339f25e1085d117 e80f06d2a194be62ab5b1cd7ef7c7a5b241dd4fb Georgi Gerganov ggerganov@gmail.com 2024-03-22T15:33:38+02:00 GitHub noreply@github.com 2024-03-22T15:33:38+02:00 common : add HF arg helpers (#6234)
e80f06d2a194be62ab5b1cd7ef7c7a5b241dd4fb f77a8ffd3bbde77b7819823b0c006fd8c2d5cae4 Nexesenex 124105151+Nexesenex@users.noreply.github.com 2024-03-22T14:32:02+01:00 GitHub noreply@github.com 2024-03-22T15:32:02+02:00 llama : correction of the attn.v.weight quantization for IQ3_XS (#6209)
f77a8ffd3bbde77b7819823b0c006fd8c2d5cae4 72114edf068a9b2f54d3b09e9a198f611be397e8 Olivier Chafik ochafik@users.noreply.github.com 2024-03-22T13:09:07Z GitHub noreply@github.com 2024-03-22T15:09:07+02:00 tests : conditional python & node json schema tests (#6207)
72114edf068a9b2f54d3b09e9a198f611be397e8 2f0e81e053b41ca28e73a841e7bdbf9820baaa57 Olivier Chafik ochafik@users.noreply.github.com 2024-03-22T13:07:44Z GitHub noreply@github.com 2024-03-22T15:07:44+02:00 json-schema-to-grammar : fix order of props + non-str const/enum (#6232)
2f0e81e053b41ca28e73a841e7bdbf9820baaa57 29ab270e65975785cdca3243a3de71ccebc1252a slaren slarengh@gmail.com 2024-03-22T14:05:31+01:00 GitHub noreply@github.com 2024-03-22T14:05:31+01:00 cuda : add LLAMA_CUDA_NO_PEER_COPY to workaround broken ROCm p2p copy (#6208)
29ab270e65975785cdca3243a3de71ccebc1252a 6b8bb3a31d260a01020497c5f01025c34222fcb9 Xiaoyi Chen cxychina@gmail.com 2024-03-22T04:29:49-07:00 GitHub noreply@github.com 2024-03-22T13:29:49+02:00 readme : add RecurseChat to the list of UIs (#6219)
6b8bb3a31d260a01020497c5f01025c34222fcb9 68e210b3543e0cc71268bee0920441747679ee13 Jan Boon kaetemi@gmail.com 2024-03-22T19:12:05+08:00 GitHub noreply@github.com 2024-03-22T13:12:05+02:00 server : fix n_keep always showing as 0 in response (#6211)
68e210b3543e0cc71268bee0920441747679ee13 b3e94f26bab8e3b5338b5902e5af5bb01894cb4a Georgi Gerganov ggerganov@gmail.com 2024-03-22T13:08:28+02:00 GitHub noreply@github.com 2024-03-22T13:08:28+02:00 server : enable continuous batching by default (#6231)
b3e94f26bab8e3b5338b5902e5af5bb01894cb4a b2075fd6a578f5685060df4baa90ae9e48e98c70 Georgi Gerganov ggerganov@gmail.com 2024-03-22T11:35:53+02:00 GitHub noreply@github.com 2024-03-22T11:35:53+02:00 metal : proper assert for mat-mat memory alignment (#6225)
b2075fd6a578f5685060df4baa90ae9e48e98c70 95d576b48ebf582b112d1c9cf4eed7142fa4e464 Vaibhav Srivastav vaibhavs10@gmail.com 2024-03-22T08:53:43+01:00 GitHub noreply@github.com 2024-03-22T09:53:43+02:00 ci : add CURL flag for the mac builds (#6214)
95d576b48ebf582b112d1c9cf4eed7142fa4e464 59c17f02de8fdf7b084d6100b875b7e2bc07a83b Georgi Gerganov ggerganov@gmail.com 2024-03-22T09:36:03+02:00 GitHub noreply@github.com 2024-03-22T09:36:03+02:00 metal : pad n_ctx by 32 (#6177)
59c17f02de8fdf7b084d6100b875b7e2bc07a83b fa046eafbc70bf97dcf39843af0323f19a8c9ac3 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-22T15:19:37+08:00 GitHub noreply@github.com 2024-03-22T15:19:37+08:00 add blog link (#6222)
fa046eafbc70bf97dcf39843af0323f19a8c9ac3 be07a03217376c53b1d95e5f658adbbbb2831555 DAN™ dranger003@gmail.com 2024-03-21T21:32:42-04:00 GitHub noreply@github.com 2024-03-22T02:32:42+01:00 Fix params underscore convert to dash. (#6203)
be07a03217376c53b1d95e5f658adbbbb2831555 d0a71233fbf8ade8ef06ad8e6b81d1d7b254895f Jan Boon kaetemi@gmail.com 2024-03-22T06:41:24+08:00 GitHub noreply@github.com 2024-03-21T23:41:24+01:00 server : update readme doc from `slot_id` to `id_slot` (#6213)
d0a71233fbf8ade8ef06ad8e6b81d1d7b254895f f372c49ccdc561ab96fb3c7d2b7cbc0f89a4b359 slaren slarengh@gmail.com 2024-03-21T19:54:28+01:00 GitHub noreply@github.com 2024-03-21T20:54:28+02:00 cuda : disable host register by default (#6206)
f372c49ccdc561ab96fb3c7d2b7cbc0f89a4b359 924ce1dce7fdf54894b462d03e7b608a6e574c32 semidark me@semidark.net 2024-03-21T11:52:35-06:00 GitHub noreply@github.com 2024-03-21T18:52:35+01:00 Corrected typo to wrong file (#6199)
924ce1dce7fdf54894b462d03e7b608a6e574c32 03a8f8fafec2e21009be9fe7297d92e5d4538964 Georgi Gerganov ggerganov@gmail.com 2024-03-21T16:20:05+02:00 GitHub noreply@github.com 2024-03-21T16:20:05+02:00 tests : disable system() calls (#6198)
03a8f8fafec2e21009be9fe7297d92e5d4538964 cfd3be76e37dab92c846d75a2421178f20db4a11 slaren slarengh@gmail.com 2024-03-21T13:59:53+01:00 GitHub noreply@github.com 2024-03-21T14:59:53+02:00 cuda : fix LLAMA_CUDA_F16 build (#6197)
cfd3be76e37dab92c846d75a2421178f20db4a11 5b7b0ac8dfdd800c0fd0dc69b69991e8cb19fb46 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-21T13:59:38+01:00 GitHub noreply@github.com 2024-03-21T14:59:38+02:00 ggml : same IQ4_NL quantization for CPU/CUDA/Metal (#6196)
5b7b0ac8dfdd800c0fd0dc69b69991e8cb19fb46 1943c0198125a0da1a200390e82cf461f9080d99 Olivier Chafik ochafik@users.noreply.github.com 2024-03-21T11:50:43Z GitHub noreply@github.com 2024-03-21T11:50:43Z json-schema-to-grammar improvements (+ added to server) (#5978)
1943c0198125a0da1a200390e82cf461f9080d99 5e43ba87429d85acbde97d16b2f48d9de992cc80 Vaibhav Srivastav vaibhavs10@gmail.com 2024-03-21T10:30:40+01:00 GitHub noreply@github.com 2024-03-21T11:30:40+02:00 ci : fix indentation error (#6195)
5e43ba87429d85acbde97d16b2f48d9de992cc80 76aa30a26353f597e4fbe3cf776772ae812af89a Vaibhav Srivastav vaibhavs10@gmail.com 2024-03-21T10:13:12+01:00 GitHub noreply@github.com 2024-03-21T11:13:12+02:00 build : add mac pre-build binaries (#6182)
76aa30a26353f597e4fbe3cf776772ae812af89a c5b8595e3f4f4ed319ef71c9c9d868d1b7a27626 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-21T08:27:57+01:00 GitHub noreply@github.com 2024-03-21T08:27:57+01:00 Add ability to use Q5_0, Q5_1, and IQ4_NL for quantized K cache (#6183)
c5b8595e3f4f4ed319ef71c9c9d868d1b7a27626 42e21c68826f2e56b9592dccd9f3c43895b6890d AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-03-21T06:10:52Z GitHub noreply@github.com 2024-03-21T11:40:52+05:30 Add nvidia and amd backends (#6157)
42e21c68826f2e56b9592dccd9f3c43895b6890d 1c51f98adcbad40e3c41f0a6ffadeb723190b417 slaren slarengh@gmail.com 2024-03-21T01:47:46+01:00 GitHub noreply@github.com 2024-03-21T01:47:46+01:00 cuda : fix conflict with std::swap (#6186)
1c51f98adcbad40e3c41f0a6ffadeb723190b417 f9c7ba34476ffc4f13ae2cdb1aec493a16eb8d47 slaren slarengh@gmail.com 2024-03-20T21:03:26+01:00 GitHub noreply@github.com 2024-03-20T21:03:26+01:00 cuda : print the returned error when CUDA initialization fails (#6185)
f9c7ba34476ffc4f13ae2cdb1aec493a16eb8d47 272935b281fee5c683e3d6d1eb580b84553cf503 Ziang Wu 97337387+ZiangWu-77@users.noreply.github.com 2024-03-20T23:29:51+08:00 GitHub noreply@github.com 2024-03-20T17:29:51+02:00 llava : update MobileVLM-README.md (#6180)
272935b281fee5c683e3d6d1eb580b84553cf503 ccf58aa3ec4d20b10162ba40898dc038ad4c3fad Ziang Wu 97337387+ZiangWu-77@users.noreply.github.com 2024-03-20T23:02:32+08:00 GitHub noreply@github.com 2024-03-20T17:02:32+02:00 llava : add MobileVLM_V2 backup (#6175)
ccf58aa3ec4d20b10162ba40898dc038ad4c3fad 91f8ad167dcd24b54615b468d9dd764ebe1d37ad slaren slarengh@gmail.com 2024-03-20T14:42:59+01:00 GitHub noreply@github.com 2024-03-20T14:42:59+01:00 cuda : refactor to remove global resources (#6170)
91f8ad167dcd24b54615b468d9dd764ebe1d37ad 6b7e76d28cdf4361740054140708c71828453522 Xuan Son Nguyen thichthat@gmail.com 2024-03-20T13:30:36+01:00 GitHub noreply@github.com 2024-03-20T13:30:36+01:00 Server: version bump for httplib and json (#6169)
6b7e76d28cdf4361740054140708c71828453522 bc0baab2ea8f806961dd3fb9f534cfeb59a2e1fc Georgi Gerganov ggerganov@gmail.com 2024-03-20T14:17:34+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-20T14:17:34+02:00 gitignore : ignore curl-related files
bc0baab2ea8f806961dd3fb9f534cfeb59a2e1fc d795988d9e09f75412efe2134512914c42780ad5 Georgi Gerganov ggerganov@gmail.com 2024-03-20T14:14:32+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-20T14:14:32+02:00 server : allow to override -ngl in tests (#6170)
d795988d9e09f75412efe2134512914c42780ad5 f8c4e745e1e728204ab26dbadf52853545e6789c Georgi Gerganov ggerganov@gmail.com 2024-03-20T13:29:49+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-20T13:29:49+02:00 Revert "llava : add a MobileVLM_V2-1.7B backup (#6152)"
f8c4e745e1e728204ab26dbadf52853545e6789c 47cc7a7bf9793f81a6dfe7c2096c499403f64dd6 Ziang Wu 97337387+ZiangWu-77@users.noreply.github.com 2024-03-20T19:20:37+08:00 GitHub noreply@github.com 2024-03-20T13:20:37+02:00 llava : add a MobileVLM_V2-1.7B backup (#6152)
47cc7a7bf9793f81a6dfe7c2096c499403f64dd6 bd60d82d0cc8b6852ec535495a5042dbdf05de24 Karthick j.karthic2004@gmail.com 2024-03-20T16:32:34+05:30 GitHub noreply@github.com 2024-03-20T12:02:34+01:00 Server: Handle n_keep parameter in the request (#6174)
bd60d82d0cc8b6852ec535495a5042dbdf05de24 6c0b287748327741b113d7d6018b68c63039b1c5 Jared Van Bortel jared@nomic.ai 2024-03-20T01:33:49-04:00 GitHub noreply@github.com 2024-03-20T06:33:49+01:00 server tests : more pythonic process management; fix bare `except:` (#6146)
6c0b287748327741b113d7d6018b68c63039b1c5 d26e8b669dbf1f5f5a0afe4d2d885e86cf566302 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-20T11:21:41+08:00 GitHub noreply@github.com 2024-03-20T11:21:41+08:00 update readme sycl for new update (#6151)
d26e8b669dbf1f5f5a0afe4d2d885e86cf566302 d8b009a9456bf5284376149f3deb09300a37701a Abhilash Majumder 30946547+abhilash1910@users.noreply.github.com 2024-03-20T08:28:49+05:30 GitHub noreply@github.com 2024-03-20T08:28:49+05:30 increase igpu cluster limit (#6159)
d8b009a9456bf5284376149f3deb09300a37701a d0d5de42e5a65865b5fddb6f5c785083539b74c3 DAN™ dranger003@gmail.com 2024-03-19T12:16:09-04:00 GitHub noreply@github.com 2024-03-19T17:16:09+01:00 Remove undeed header file. (#6158)
d0d5de42e5a65865b5fddb6f5c785083539b74c3 b80cf3b2d1dee0ad325f7a794fecc66befce7336 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-19T12:05:44+01:00 GitHub noreply@github.com 2024-03-19T12:05:44+01:00 gguf-split: split and merge gguf per batch of tensors (#6135)
b80cf3b2d1dee0ad325f7a794fecc66befce7336 970a48060ab9a6cc67aa063870323781c2a7bd7d Georgi Gerganov ggerganov@gmail.com 2024-03-19T10:21:54+02:00 GitHub noreply@github.com 2024-03-19T10:21:54+02:00 common : disable repeat penalties by default (#6127)
970a48060ab9a6cc67aa063870323781c2a7bd7d 4c28b8252907561165827125d2d1a4bad6926ac6 slaren slarengh@gmail.com 2024-03-19T09:06:54+01:00 GitHub noreply@github.com 2024-03-19T10:06:54+02:00 ci : exempt some labels from being tagged as stale (#6140)
4c28b8252907561165827125d2d1a4bad6926ac6 2d15886bb092c3b780c676b5cc57ff3337af9c83 DAN™ dranger003@gmail.com 2024-03-19T01:59:36-04:00 GitHub noreply@github.com 2024-03-19T07:59:36+02:00 common : print usage on '-h' and '--help' (#6145)
2d15886bb092c3b780c676b5cc57ff3337af9c83 d199ca79f279e84ebe27caafe0aa59c461d88969 github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-03-17T06:37:44Z Someone newkozlukov@gmail.com 2024-03-18T18:51:30Z flake.lock: Update
d199ca79f279e84ebe27caafe0aa59c461d88969 104f5e0fc156d48476258295457cafeec2a2af10 Jared Van Bortel jared@nomic.ai 2024-03-18T12:49:02-04:00 GitHub noreply@github.com 2024-03-18T12:49:02-04:00 mpt : implement backwards compatiblity with duped output tensor (#6139)
104f5e0fc156d48476258295457cafeec2a2af10 5e1b7f94a03e0b3b8e4578625bbdadc7bbd2b93c Felix stenbackfelix@gmail.com 2024-03-18T16:40:22+01:00 GitHub noreply@github.com 2024-03-18T17:40:22+02:00 clip : fix memory leak (#6138)
5e1b7f94a03e0b3b8e4578625bbdadc7bbd2b93c ac9ee6a4ad740bc1ee484ede43e9f92b5af244c1 slaren slarengh@gmail.com 2024-03-18T16:33:44+01:00 GitHub noreply@github.com 2024-03-18T16:33:44+01:00 backend : set max split inputs to GGML_MAX_SRC (#6137)
ac9ee6a4ad740bc1ee484ede43e9f92b5af244c1 4f6d1337ca5a409dc74aca8c479b7c34408a69c0 Georgi Gerganov ggerganov@gmail.com 2024-03-18T13:45:38+02:00 GitHub noreply@github.com 2024-03-18T13:45:38+02:00 ci : disable stale issue messages (#6126)
4f6d1337ca5a409dc74aca8c479b7c34408a69c0 2bf8d0f7c4cc1235755ad06961ca761e458c5e55 Georgi Gerganov ggerganov@gmail.com 2024-03-18T13:45:27+02:00 GitHub noreply@github.com 2024-03-18T13:45:27+02:00 ci : temporary disable sanitizer builds (#6128)
2bf8d0f7c4cc1235755ad06961ca761e458c5e55 496bc79bc2b79bfd6124b8687a8dbd6a646e9b06 slaren slarengh@gmail.com 2024-03-18T11:03:04+01:00 GitHub noreply@github.com 2024-03-18T11:03:04+01:00 backend : offload large batches to GPU (#6083)
496bc79bc2b79bfd6124b8687a8dbd6a646e9b06 9b03719ad712e2dc36c5c0c20f352bf3e4bda332 DAN™ dranger003@gmail.com 2024-03-18T04:27:44-04:00 GitHub noreply@github.com 2024-03-18T10:27:44+02:00 common : tidy-up argument parsing (#6105)
9b03719ad712e2dc36c5c0c20f352bf3e4bda332 3a6efdd03c46c5ba08e43880d34260c02dd9999b Thérence 13496987+Royalphax@users.noreply.github.com 2024-03-18T09:17:00+01:00 GitHub noreply@github.com 2024-03-18T10:17:00+02:00 convert : add support for CamembertModel architecture (#6119)
3a6efdd03c46c5ba08e43880d34260c02dd9999b d01b3c4c32357567f3531d4e6ceffc5d23e87583 Romain D 90720+Artefact2@users.noreply.github.com 2024-03-18T09:04:41+01:00 GitHub noreply@github.com 2024-03-18T10:04:41+02:00 convert : use f32 outtype for bf16 tensors (#6106)
d01b3c4c32357567f3531d4e6ceffc5d23e87583 cd776c37c945bf58efc8fe44b370456680cb1b59 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-17T19:12:37+01:00 GitHub noreply@github.com 2024-03-17T19:12:37+01:00 common: llama_load_model_from_url using --model-url (#6098)
cd776c37c945bf58efc8fe44b370456680cb1b59 dc0f6125487dcfbff913360f9d877bc0ccf6aa57 Georgi Gerganov ggerganov@gmail.com 2024-03-17T19:51:57+02:00 GitHub noreply@github.com 2024-03-17T18:51:57+01:00 ci : close all stale issues at once (#6115)
dc0f6125487dcfbff913360f9d877bc0ccf6aa57 c47cf414efafb8f60596edc7edb5a2d68065e992 GainLee perfecter.gen@gmail.com 2024-03-18T01:12:22+08:00 GitHub noreply@github.com 2024-03-17T18:12:22+01:00 ggml:fix finding transfer queue family index error (#6094)
c47cf414efafb8f60596edc7edb5a2d68065e992 b5f4ae09c3244ae1644b67c03ed9f4227ab25ad2 AmirAli Mirian 37371367+amiralimi@users.noreply.github.com 2024-03-16T11:52:02-04:00 GitHub noreply@github.com 2024-03-16T17:52:02+02:00 ggml : add AVX512F SIMD (#6088)
b5f4ae09c3244ae1644b67c03ed9f4227ab25ad2 dfbfdd60f90207404039c6578d709231496831d9 Daniel Bevenius daniel.bevenius@gmail.com 2024-03-16T16:46:29+01:00 GitHub noreply@github.com 2024-03-16T17:46:29+02:00 gritlm : add initial README.md (#6086)
dfbfdd60f90207404039c6578d709231496831d9 15961ec04dbd59d21d8984d42e4c0f7e7e7d320a Xuan Son Nguyen thichthat@gmail.com 2024-03-16T16:42:08+01:00 GitHub noreply@github.com 2024-03-16T17:42:08+02:00 readme : add wllama as a wasm binding (#6100)
15961ec04dbd59d21d8984d42e4c0f7e7e7d320a a56d09a4407f29c21e149b44fd5308f83aa1cb09 DAN™ dranger003@gmail.com 2024-03-16T11:39:15-04:00 GitHub noreply@github.com 2024-03-16T17:39:15+02:00 common : refactor nested if causing error C1061 on MSVC (#6101)
a56d09a4407f29c21e149b44fd5308f83aa1cb09 d84c48505f60bcd358b82a751d40418c4d235643 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-16T13:20:53+01:00 GitHub noreply@github.com 2024-03-16T14:20:53+02:00 ci : close inactive issue with workflow (#6053)
d84c48505f60bcd358b82a751d40418c4d235643 877b4d0c628cc70dddb5df72ed8fc14d126ca7e8 slaren slarengh@gmail.com 2024-03-15T22:14:16+01:00 GitHub noreply@github.com 2024-03-15T23:14:16+02:00 llama : fix Baichuan2 13B (#6092)
877b4d0c628cc70dddb5df72ed8fc14d126ca7e8 12247f4c69a173b9482f68aaa174ec37fc909ccf Theia Vogel theia@vgel.me 2024-03-15T13:43:02-07:00 GitHub noreply@github.com 2024-03-15T22:43:02+02:00 llama : add support for control vectors (#5970)
12247f4c69a173b9482f68aaa174ec37fc909ccf 4e9a7f7f7fb6acbddd1462909c8d696e38edbfcc Andrew Canis andrew.canis@gmail.com 2024-03-15T16:41:22-04:00 GitHub noreply@github.com 2024-03-15T22:41:22+02:00 llama : add Command-R support (#6033)
4e9a7f7f7fb6acbddd1462909c8d696e38edbfcc 3020327f6cd6d2ce50528dd65f4b199d2ea8b1ae Ting Lou ting.lou@gmail.com 2024-03-15T22:31:05+08:00 GitHub noreply@github.com 2024-03-15T16:31:05+02:00 llava : change API to pure C style for Rust FFI bindgen (#6079)
3020327f6cd6d2ce50528dd65f4b199d2ea8b1ae 46acb3676718b983157058aecf729a2064fc7d34 slaren slarengh@gmail.com 2024-03-15T13:24:03+01:00 GitHub noreply@github.com 2024-03-15T14:24:03+02:00 cuda : disable unused cudaLaunchHostFunc code (#6078)
46acb3676718b983157058aecf729a2064fc7d34 131b0584096ee9df4d07cb28759dfea6efe6475f Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-15T18:53:53+08:00 GitHub noreply@github.com 2024-03-15T18:53:53+08:00 fix set main gpu error (#6073)
131b0584096ee9df4d07cb28759dfea6efe6475f 753e36f650fa2a5869f89188d9ee745dc74cf14b Georgi Gerganov ggerganov@gmail.com 2024-03-15T11:36:50+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-15T11:38:40+02:00 make : ggml-metal.o depends on ggml.h
753e36f650fa2a5869f89188d9ee745dc74cf14b 7ce2c77f88e1ca66ec48417e56f91746bac018c2 AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-03-15T09:26:20Z GitHub noreply@github.com 2024-03-15T14:56:20+05:30 [SYCL] Fix non-intel device selection (#6042)
7ce2c77f88e1ca66ec48417e56f91746bac018c2 aab606a11fc0a9740a7f297521c3eef851dfb351 Ondřej Čertík ondrej@certik.us 2024-03-15T02:46:51-06:00 GitHub noreply@github.com 2024-03-15T10:46:51+02:00 gguf : add support for I64 and F64 arrays (#6062)
aab606a11fc0a9740a7f297521c3eef851dfb351 b0bc9f4a9da7c19f4779106ea83b23feca747566 Xuan Son Nguyen thichthat@gmail.com 2024-03-15T09:44:57+01:00 GitHub noreply@github.com 2024-03-15T10:44:57+02:00 llama : add Orion chat template (#6066)
b0bc9f4a9da7c19f4779106ea83b23feca747566 4755afd1cbd40d93c017e5b98c39796f52345314 slaren slarengh@gmail.com 2024-03-15T09:22:24+01:00 GitHub noreply@github.com 2024-03-15T10:22:24+02:00 llama-bench : use random tokens to improve accuracy with mixtral (#6069)
4755afd1cbd40d93c017e5b98c39796f52345314 6e0438da3cc95b89cdbf55f45fa4e324d9076792 Georgi Gerganov ggerganov@gmail.com 2024-03-14T22:58:41+02:00 GitHub noreply@github.com 2024-03-14T22:58:41+02:00 llama : fix integer overflow during quantization (#6063)
6e0438da3cc95b89cdbf55f45fa4e324d9076792 727107707a73b3dc8a497cf9fc9405722c16dd2b Steve Grubb ausearch.1@gmail.com 2024-03-14T14:29:32-04:00 GitHub noreply@github.com 2024-03-14T20:29:32+02:00 gguf : fix resource leaks (#6061)
727107707a73b3dc8a497cf9fc9405722c16dd2b 69ff61397d2b7b550dcdda4a35b35128892408b0 Ondřej Čertík ondrej@certik.us 2024-03-14T11:57:31-06:00 GitHub noreply@github.com 2024-03-14T19:57:31+02:00 gguf-py : bump version to 0.8.0 (#6060)
69ff61397d2b7b550dcdda4a35b35128892408b0 044ec4b2a567f649459ccd20af2f387c784faa51 Michael Podvitskiy podvitskiymichael@gmail.com 2024-03-14T17:21:56+01:00 GitHub noreply@github.com 2024-03-14T18:21:56+02:00 llama : support models without vocabulary (#5798)
044ec4b2a567f649459ccd20af2f387c784faa51 77178eedc83d49f31bf757d8e12315d76460be78 Georgi Gerganov ggerganov@gmail.com 2024-03-14T15:14:14+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-14T15:14:14+02:00 embedding : add EOS token if not present (#899)
77178eedc83d49f31bf757d8e12315d76460be78 15a333260ab637a040ed0864c206a2ceaf806bb8 Georgi Gerganov ggerganov@gmail.com 2024-03-14T13:32:14+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-14T13:32:14+02:00 gguf-py : fix dtype check (#6045)
15a333260ab637a040ed0864c206a2ceaf806bb8 43241adf22e8231ffaf3827d2c9310cc0ffd5ac5 Jian Liao jianliao@users.noreply.github.com 2024-03-14T04:18:23-07:00 GitHub noreply@github.com 2024-03-14T13:18:23+02:00 readme : improve readme for Llava-1.6 example (#6044)
43241adf22e8231ffaf3827d2c9310cc0ffd5ac5 a44bc969e4cd62ca9f4332e17fe3c51f2093e7c6 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-14T12:15:39+01:00 GitHub noreply@github.com 2024-03-14T13:15:39+02:00 server: disable debug release type sanitizer, simplify trigger (#6047)
a44bc969e4cd62ca9f4332e17fe3c51f2093e7c6 2c4fb69246834503db7b78bcbedcef506bbc60c4 Georgi Gerganov ggerganov@gmail.com 2024-03-14T13:13:06+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-14T13:13:06+02:00 llama : fix typo
2c4fb69246834503db7b78bcbedcef506bbc60c4 3ca23481dd309bd51cc31c73a4cc34f922cc372f Michael Podvitskiy podvitskiymichael@gmail.com 2024-03-14T11:56:48+01:00 GitHub noreply@github.com 2024-03-14T12:56:48+02:00 llama : optimize defrag moves + fix fragmentation calculation (#6037)
3ca23481dd309bd51cc31c73a4cc34f922cc372f 3fe8d7a17f84bd721cd4d8db35365da44b69f68b Ondřej Čertík ondrej@certik.us 2024-03-14T04:40:14-06:00 GitHub noreply@github.com 2024-03-14T12:40:14+02:00 gguf-py : add support for I8, I16 and I32 (#6045)
3fe8d7a17f84bd721cd4d8db35365da44b69f68b 68265ebfc6a1bed022973ea0c3145be1450b7e70 Georgi Gerganov ggerganov@gmail.com 2024-03-14T12:38:37+02:00 GitHub noreply@github.com 2024-03-14T12:38:37+02:00 ggml : designate enum vals for integer types (#6050)
68265ebfc6a1bed022973ea0c3145be1450b7e70 381da2d9f0940d7009e3e918bed36338c8ff2fbb Georgi Gerganov ggerganov@gmail.com 2024-03-14T12:37:20+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-14T12:37:20+02:00 embedding : print all resulting embeddings (#899)
381da2d9f0940d7009e3e918bed36338c8ff2fbb 0fd6c1f015f6cccf3b527f7dbd8386a434728126 Georgi Gerganov ggerganov@gmail.com 2024-03-14T11:55:23+02:00 GitHub noreply@github.com 2024-03-14T11:55:23+02:00 metal : build metallib + fix embed path (#6015)
0fd6c1f015f6cccf3b527f7dbd8386a434728126 19885d205e768579ab090d1e99281cae58c21b54 Georgi Gerganov ggerganov@gmail.com 2024-03-14T10:12:29+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-14T10:12:29+02:00 embedding : print cosine similarity (#899)
19885d205e768579ab090d1e99281cae58c21b54 76a936c8939c249a7c3e8e66dfefbab13eae194f Linwei Wang wanix1988@gmail.com 2024-03-14T02:34:40+08:00 GitHub noreply@github.com 2024-03-13T20:34:40+02:00 readme : update details about running llama in Termux on Android (#6039)
76a936c8939c249a7c3e8e66dfefbab13eae194f 463628372d5fe3a0c1e5864aa5fc57deb7387039 Georgi Gerganov ggerganov@gmail.com 2024-03-13T20:33:56+02:00 GitHub noreply@github.com 2024-03-13T20:33:56+02:00 readme : update API changes and hot topics
463628372d5fe3a0c1e5864aa5fc57deb7387039 f30ea47a87ed4446ad55adb265755dc9102956a2 Clint Herron hanclinto@gmail.com 2024-03-13T14:10:40-04:00 GitHub noreply@github.com 2024-03-13T20:10:40+02:00 grammar : handle missing "root" node (#6004)
f30ea47a87ed4446ad55adb265755dc9102956a2 d8fd0ccf6ac8b07791ffd1575eed436930854ae3 slaren slarengh@gmail.com 2024-03-13T18:54:21+01:00 GitHub noreply@github.com 2024-03-13T18:54:21+01:00 llama : add pipeline parallelism support (#6017)
d8fd0ccf6ac8b07791ffd1575eed436930854ae3 b3d978600f07f22e94f2e797f18a8b5f6df23c89 slaren slarengh@gmail.com 2024-03-13T14:58:30+01:00 GitHub noreply@github.com 2024-03-13T15:58:30+02:00 test-backend-ops : skip CPU backend by default (#6028)
b3d978600f07f22e94f2e797f18a8b5f6df23c89 99b71c068f624521ad977e08e41589e2971fa1c7 AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-03-13T13:17:54Z GitHub noreply@github.com 2024-03-13T18:47:54+05:30 Update get version (#6025)
99b71c068f624521ad977e08e41589e2971fa1c7 306d34be7ad19e768975409fc80791a274ea0230 Xuan Son Nguyen thichthat@gmail.com 2024-03-13T11:39:11+01:00 GitHub noreply@github.com 2024-03-13T11:39:11+01:00 Server: Use multi-task for embeddings endpoint (#6001)
306d34be7ad19e768975409fc80791a274ea0230 8030da7afea2d89f997aeadbd14183d399a017b9 slaren slarengh@gmail.com 2024-03-12T16:55:19+01:00 GitHub noreply@github.com 2024-03-12T17:55:19+02:00 ci : remove tidy-review (#6021)
8030da7afea2d89f997aeadbd14183d399a017b9 184215e783dfad76aded2c68244c327a5c507df5 Georgi Gerganov ggerganov@gmail.com 2024-03-12T14:27:20+02:00 GitHub noreply@github.com 2024-03-12T14:27:20+02:00 ggml : reuse quantum structs across backends (#5943)
184215e783dfad76aded2c68244c327a5c507df5 48358b2e5b3983c41ba7e61a493e84d3901dc7b9 Georgi Gerganov ggerganov@gmail.com 2024-03-12T13:49:55+02:00 GitHub noreply@github.com 2024-03-12T13:49:55+02:00 ggml : fix UB in IQ2_S and IQ3_S (#6012)
48358b2e5b3983c41ba7e61a493e84d3901dc7b9 5cdb371731caa2c41fcca42d4d2d43f94f6883b4 Georgi Gerganov ggerganov@gmail.com 2024-03-12T11:15:05+02:00 GitHub noreply@github.com 2024-03-12T11:15:05+02:00 sycl : update IQ1_S kernels (WIP - not working!) (#5995)
5cdb371731caa2c41fcca42d4d2d43f94f6883b4 44ca159faf4fbe1a7ace13a962845ba7cdfd95ec gliptic gliptic@users.noreply.github.com 2024-03-11T20:59:03+01:00 GitHub noreply@github.com 2024-03-11T21:59:03+02:00 grammar : fix unnecessarily retained pointer to rules (#6003)
44ca159faf4fbe1a7ace13a962845ba7cdfd95ec 05b06210c954491cf0f12034b0a62bd4d69ce78b Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-11T16:53:15+01:00 GitHub noreply@github.com 2024-03-11T17:53:15+02:00 1.5 bit: we can do even better (#5999)
05b06210c954491cf0f12034b0a62bd4d69ce78b 83796e62bc9f6caae6228168e359890f51e60fee Georgi Gerganov ggerganov@gmail.com 2024-03-11T17:49:47+02:00 GitHub noreply@github.com 2024-03-11T17:49:47+02:00 llama : more consistent names of count variables (#5994)
83796e62bc9f6caae6228168e359890f51e60fee 828defefb66fc8a25404f5de845897145bf34061 Georgi Gerganov ggerganov@gmail.com 2024-03-11T17:47:47+02:00 GitHub noreply@github.com 2024-03-11T17:47:47+02:00 llama : refactor unicode stuff (#5992)
828defefb66fc8a25404f5de845897145bf34061 caa106d4e05a0ab94225c220b81f9e2cd522339b Jakub N jakubniemczyk97@gmail.com 2024-03-11T14:40:42+01:00 GitHub noreply@github.com 2024-03-11T14:40:42+01:00 Update server docker image URLs (#5997)
caa106d4e05a0ab94225c220b81f9e2cd522339b 3202361c5b1ba15e695b31209567ef42c22c5c32 Xuan Son Nguyen thichthat@gmail.com 2024-03-11T10:56:41+01:00 GitHub noreply@github.com 2024-03-11T10:56:41+01:00 Server: format error to json (#5961)
3202361c5b1ba15e695b31209567ef42c22c5c32 332bdfd7980718abf664bfa5460f2288a3314984 Michael Podvitskiy podvitskiymichael@gmail.com 2024-03-11T10:28:51+01:00 GitHub noreply@github.com 2024-03-11T11:28:51+02:00 ggml, ci : Windows ARM runner and build fixes (#5979)
332bdfd7980718abf664bfa5460f2288a3314984 ecab1c75de68de7c41c254e2ae170d3b07bee6d4 Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-11T17:09:32+09:00 GitHub noreply@github.com 2024-03-11T10:09:32+02:00 server : maintain chat completion id for streaming responses (#5988)
ecab1c75de68de7c41c254e2ae170d3b07bee6d4 ee35600b9061b1ea0c4ea87fce6844297632b2a8 Gilad S giladgd@users.noreply.github.com 2024-03-11T10:00:08+02:00 GitHub noreply@github.com 2024-03-11T10:00:08+02:00 cmake : fix subdir for `LLAMA_METAL_EMBED_LIBRARY` (#5985)
ee35600b9061b1ea0c4ea87fce6844297632b2a8 be858f620508385ad12d0e5e862010e666ca729c Georgi Gerganov ggerganov@gmail.com 2024-03-11T09:56:47+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-11T09:56:47+02:00 llama : fix F16/F32 downcast + improve names (#5980)
be858f620508385ad12d0e5e862010e666ca729c ef3ced26a3817d92890b97b83acaeb018ade02d0 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-11T07:51:49+01:00 GitHub noreply@github.com 2024-03-11T07:51:49+01:00 Better 1.5 bit quantization (#5971)
ef3ced26a3817d92890b97b83acaeb018ade02d0 3814a07392d2bdc22911652bc7c2f9bdb0ce042e Abhilash Majumder 30946547+abhilash1910@users.noreply.github.com 2024-03-11T10:27:56+05:30 GitHub noreply@github.com 2024-03-11T10:27:56+05:30 [SYCL] Add q3_s and q1_s (#5886)
3814a07392d2bdc22911652bc7c2f9bdb0ce042e bb6d00bbf98476215596b9df3870b5504ef5a29a AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-03-11T01:13:57Z GitHub noreply@github.com 2024-03-11T09:13:57+08:00 [SYCL] Add support for SYCL Nvidia target (#5738)
bb6d00bbf98476215596b9df3870b5504ef5a29a 7ab7b733bb48250b2df26c12b00256ef42c76932 Georgi Gerganov ggerganov@gmail.com 2024-03-10T23:12:48+02:00 GitHub noreply@github.com 2024-03-10T23:12:48+02:00 metal : move mm_id indices to shared mem (#5982)
7ab7b733bb48250b2df26c12b00256ef42c76932 d9f65c97c3dc3aa6fa27470b8c6e69b437ec1a27 Dean Dean.Sinaean@gmail.com 2024-03-11T04:03:17+08:00 GitHub noreply@github.com 2024-03-10T22:03:17+02:00 android : fix utf8 decoding error (#5935)
d9f65c97c3dc3aa6fa27470b8c6e69b437ec1a27 b838b53ad6de2e53f23ddf8f3ad5e6891cc3dd05 Georgi Gerganov ggerganov@gmail.com 2024-03-10T20:58:26+02:00 GitHub noreply@github.com 2024-03-10T20:58:26+02:00 readme : update hot topics
b838b53ad6de2e53f23ddf8f3ad5e6891cc3dd05 df4dc3e7cb43162862f339525638ba4febcb6158 Georgi Gerganov ggerganov@gmail.com 2024-03-10T20:10:46+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-10T20:10:46+02:00 sync : ggml
df4dc3e7cb43162862f339525638ba4febcb6158 bf47a5eefc669fdba71af096942af999bc1167d4 Georgi Gerganov ggerganov@gmail.com 2024-03-08T23:45:07+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-10T20:10:39+02:00 ggml : try fix 32-bit arm compat (whisper/1938)
bf47a5eefc669fdba71af096942af999bc1167d4 fa8a809a9119411bc9c3f00026550d0343f4d9b7 Georgi Gerganov ggerganov@gmail.com 2024-03-10T20:09:24+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-10T20:09:24+02:00 ggml : remove __constant__ specifier for CUDA tables (#5940)
fa8a809a9119411bc9c3f00026550d0343f4d9b7 bcebd7dbf62fd7b293d5ed089023e4e733269c71 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-10T18:17:47+01:00 GitHub noreply@github.com 2024-03-10T18:17:47+01:00 server: ci: windows build and tests (#5968)
bcebd7dbf62fd7b293d5ed089023e4e733269c71 2960eae847f8dbde23be6d170a61bcf44ebf32de DAN™ dranger003@gmail.com 2024-03-10T11:56:30-04:00 GitHub noreply@github.com 2024-03-10T17:56:30+02:00 llama : add support for GritLM (#5959)
2960eae847f8dbde23be6d170a61bcf44ebf32de c78541479cf835dd9eb568ccd9a2083198a7203d Clint Herron hanclinto@gmail.com 2024-03-10T11:17:43-04:00 GitHub noreply@github.com 2024-03-10T17:17:43+02:00 grammar : verify parsed state (#5950)
c78541479cf835dd9eb568ccd9a2083198a7203d 621e86b331f8b0e71f79fd82a4ae1cd54c3e4396 Georgi Gerganov ggerganov@gmail.com 2024-03-10T16:43:08+02:00 GitHub noreply@github.com 2024-03-10T16:43:08+02:00 nix: update flake.lock (#5969)
621e86b331f8b0e71f79fd82a4ae1cd54c3e4396 77d1ac7e00bf049b9f2bba1b5a310a78318c49c4 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-09T23:41:49+01:00 GitHub noreply@github.com 2024-03-09T23:41:49+01:00 server: benchmark: chat/completions scenario and other llm servers comparison (#5941)
77d1ac7e00bf049b9f2bba1b5a310a78318c49c4 d894f352bf433157232dc8dc54eacd50014e898e Georgi Gerganov ggerganov@gmail.com 2024-03-09T22:04:00+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-09T22:04:00+02:00 server : print chat template info
d894f352bf433157232dc8dc54eacd50014e898e 098dbaab449f5309a54871ba7e5acef72ae696de slaren slarengh@gmail.com 2024-03-09T19:55:54+01:00 GitHub noreply@github.com 2024-03-09T19:55:54+01:00 perplexity : support using multiple sequences to allow larger batch sizes (#5946)
098dbaab449f5309a54871ba7e5acef72ae696de 8380ecfb219c2e73cc706fcc83935b7c806cc7c3 Georgi Gerganov ggerganov@gmail.com 2024-03-09T18:14:13+02:00 GitHub noreply@github.com 2024-03-09T18:14:13+02:00 readme : update hot topics
8380ecfb219c2e73cc706fcc83935b7c806cc7c3 58308a0ecce7cc261b802f4803c38d420063db21 Georgi Gerganov ggerganov@gmail.com 2024-03-09T17:36:20+02:00 GitHub noreply@github.com 2024-03-09T17:36:20+02:00 ggml : fix unnecessary f32 -> f16 -> f32 casts (mmla) (#5951)
58308a0ecce7cc261b802f4803c38d420063db21 5b09797321430f08caf0473143a962916ab2ea89 Georgi Gerganov ggerganov@gmail.com 2024-03-09T17:34:15+02:00 GitHub noreply@github.com 2024-03-09T17:34:15+02:00 server : fix metrics init (#5964)
5b09797321430f08caf0473143a962916ab2ea89 97c09585d65a95864773b4d25d66d0f708baf38d Georgi Gerganov ggerganov@gmail.com 2024-03-09T15:53:59+02:00 GitHub noreply@github.com 2024-03-09T15:53:59+02:00 ggml : remove old quantization functions (#5942)
97c09585d65a95864773b4d25d66d0f708baf38d fb215c3832236fec7380c4fb618bd7154cb196ef Georgi Gerganov ggerganov@gmail.com 2024-03-09T15:47:47+02:00 GitHub noreply@github.com 2024-03-09T15:47:47+02:00 server : clarify some items in the readme (#5957)
fb215c3832236fec7380c4fb618bd7154cb196ef 2c4f566c88322ebf2f9bd11b01b5ebdaa0130b89 SeungWon Jeong 65549245+redlion0929@users.noreply.github.com 2024-03-09T21:27:58+09:00 GitHub noreply@github.com 2024-03-09T14:27:58+02:00 server : normalize embeddings (#5956)
2c4f566c88322ebf2f9bd11b01b5ebdaa0130b89 0db32beaf09d90b8959d3d0cc493ed1e45685353 Georgi Gerganov ggerganov@gmail.com 2024-03-09T14:17:11+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-09T14:17:11+02:00 tests : gitignore ggml-common.h
0db32beaf09d90b8959d3d0cc493ed1e45685353 8a3012a4ad08112bb3dc3f1399afec4e93780c44 Alexey Parfenov zxed@alkatrazstudio.net 2024-03-09T11:16:53Z GitHub noreply@github.com 2024-03-09T13:16:53+02:00 server : fix passing prompt as tokens (#5955)
8a3012a4ad08112bb3dc3f1399afec4e93780c44 9674aaf35cb81478eb38c3f3ebde713ec72fbb79 Georgi Gerganov ggerganov@gmail.com 2024-03-09T12:47:57+02:00 GitHub noreply@github.com 2024-03-09T12:47:57+02:00 ggml : add ggml-common.h to deduplicate shared code (#5940)
9674aaf35cb81478eb38c3f3ebde713ec72fbb79 950ba1ab84db199f0bbdecdb2bb911f35261b321 Georgi Gerganov ggerganov@gmail.com 2024-03-09T12:34:18+02:00 GitHub noreply@github.com 2024-03-09T12:34:18+02:00 server : simplify logic for empty prompts (#5953)
950ba1ab84db199f0bbdecdb2bb911f35261b321 e1fa9569ba8ce276bc7801a3cebdcf8b1aa116ea Xuan Son Nguyen thichthat@gmail.com 2024-03-09T11:27:53+01:00 GitHub noreply@github.com 2024-03-09T11:27:53+01:00 Server: reorganize some http logic (#5939)
e1fa9569ba8ce276bc7801a3cebdcf8b1aa116ea fd72d2d2a5e79d61ccef6af3d15f16e5e5cbc352 Gabe Goodhart gabe.l.hart@gmail.com 2024-03-09T02:57:09-07:00 GitHub noreply@github.com 2024-03-09T11:57:09+02:00 server : add SSL support (#5926)
fd72d2d2a5e79d61ccef6af3d15f16e5e5cbc352 c2101a2e909ac7c08976d414e64e96c90ee5fa9e Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-09T10:30:04+01:00 GitHub noreply@github.com 2024-03-09T11:30:04+02:00 server: tests: add truncated prompt tests, better kv cache size (#5933)
c2101a2e909ac7c08976d414e64e96c90ee5fa9e 515f7d0d4fce41c752fc253acf30707c3be2531e compilade 113953597+compilade@users.noreply.github.com 2024-03-08T17:31:00-05:00 GitHub noreply@github.com 2024-03-08T17:31:00-05:00 llama : support Mamba Selective State Space Models (#5328)
515f7d0d4fce41c752fc253acf30707c3be2531e 76e868821a94072fbc87cb1fcca291694319eae8 compilade 113953597+compilade@users.noreply.github.com 2024-03-08T10:53:37-05:00 GitHub noreply@github.com 2024-03-08T17:53:37+02:00 llama : fix quantization of shared token_embd (#5944)
76e868821a94072fbc87cb1fcca291694319eae8 e457fb3540e0aaec47cfde0abf784c213f9216ee Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-08T12:25:04+01:00 GitHub noreply@github.com 2024-03-08T12:25:04+01:00 server: metrics: add llamacpp:prompt_seconds_total and llamacpp:tokens_predicted_seconds_total, reset bucket only on /metrics. Fix values cast to int. Add Process-Start-Time-Unix header. (#5937)
e457fb3540e0aaec47cfde0abf784c213f9216ee af37fd8b30e37ccbffdd82e6f48559e2fb7ce7dd Don Mahurin dmahurin@users.noreply.github.com 2024-03-08T02:41:50-08:00 GitHub noreply@github.com 2024-03-08T12:41:50+02:00 llama : assume tied weights if lm_head/output weights is missing (#5824)
af37fd8b30e37ccbffdd82e6f48559e2fb7ce7dd 581ed5c4fe3a8909aaa8313633ac443f471ba755 Georgi Gerganov ggerganov@gmail.com 2024-03-08T12:40:02+02:00 GitHub noreply@github.com 2024-03-08T12:40:02+02:00 server : fix EOS token detection with disabled cache (#5938)
581ed5c4fe3a8909aaa8313633ac443f471ba755 6cdabe652695167263c8b447520987b11856f7ca UEXTM.com 84163508+uextm@users.noreply.github.com 2024-03-08T04:35:04-05:00 GitHub noreply@github.com 2024-03-08T11:35:04+02:00 log : fix MSVC compile errors (#5643)
6cdabe652695167263c8b447520987b11856f7ca 89fb735fcfd21781a8194b211cf32824beb3f71f Georgi Gerganov ggerganov@gmail.com 2024-03-07T16:32:38+02:00 GitHub noreply@github.com 2024-03-07T16:32:38+02:00 llama-bench : add embeddings option (#5924)
89fb735fcfd21781a8194b211cf32824beb3f71f 55a2a900ff4a02fc33708ac7858d595d289a3f2a Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-07T19:14:49+08:00 GitHub noreply@github.com 2024-03-07T12:14:49+01:00 Revert "[SYCL] fix error when set main gpu to non-zero (#5901)" (#5918)
55a2a900ff4a02fc33708ac7858d595d289a3f2a 2002bc96bf2cbf5ab981a17d7e994d817c9801f5 Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-07T19:42:39+09:00 GitHub noreply@github.com 2024-03-07T12:42:39+02:00 server : add `/v1/completions` endpoint (#5914)
2002bc96bf2cbf5ab981a17d7e994d817c9801f5 ceca1aef0738b57951cd12c603c3477e75312dec Georgi Gerganov ggerganov@gmail.com 2024-03-07T11:41:53+02:00 GitHub noreply@github.com 2024-03-07T11:41:53+02:00 server : refactor (#5882)
ceca1aef0738b57951cd12c603c3477e75312dec e04e04f8fad549bb0b3ec1c91f0413aeb08baf29 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-07T16:34:31+08:00 GitHub noreply@github.com 2024-03-07T16:34:31+08:00 [SYCL] fix error when set main gpu to non-zero (#5901)
e04e04f8fad549bb0b3ec1c91f0413aeb08baf29 e25fb4b18fcedb9bed6be4585cf842e9a669b28b Jared Van Bortel jared@nomic.ai 2024-03-06T15:42:23-05:00 GitHub noreply@github.com 2024-03-06T15:42:23-05:00 ggml : use SYS_get_cpu if SYS_getcpu is not defined (#5906)
e25fb4b18fcedb9bed6be4585cf842e9a669b28b 1e35d619a6fb0b9c5e3dc955345980ff056ddbaf bobqianic 129547291+bobqianic@users.noreply.github.com 2024-03-06T07:35:07Z GitHub noreply@github.com 2024-03-06T09:35:07+02:00 ggml : use `uint8x16_t` return type for `ggml_vqtbl1q_u8` (#5894)
1e35d619a6fb0b9c5e3dc955345980ff056ddbaf 8ced9f7e3225adb8501e9821ed1bbd92e3a5c7ae Georgi Gerganov ggerganov@gmail.com 2024-03-06T09:12:25+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-06T09:13:42+02:00 convert : remove AWQ remnants (#5768)
8ced9f7e3225adb8501e9821ed1bbd92e3a5c7ae 652ca2bded3c818320d92c70d2b67f64bdbff5e5 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-06T12:08:32+08:00 GitHub noreply@github.com 2024-03-06T12:08:32+08:00 add wait() to make code stable (#5895)
652ca2bded3c818320d92c70d2b67f64bdbff5e5 bd836944f826f07e19b7edcf994a78728da49c1c slaren slarengh@gmail.com 2024-03-05T22:27:29+01:00 GitHub noreply@github.com 2024-03-05T22:27:29+01:00 compare-llama-bench.py : remove mul_mat_q (#5892)
bd836944f826f07e19b7edcf994a78728da49c1c 3de31677d36aa4f82d4d99898902d7bcf398e666 Jared Van Bortel jared@nomic.ai 2024-03-05T11:56:37-05:00 GitHub noreply@github.com 2024-03-05T11:56:37-05:00 quants : use MM256_SET_M128I consistently to fix gcc 7 build (#5889)
3de31677d36aa4f82d4d99898902d7bcf398e666 82cb31eb93fd19b74115e0f0133225d1dfdbfdbc ExtReMLapin 3909752+ExtReMLapin@users.noreply.github.com 2024-03-05T17:33:08+01:00 GitHub noreply@github.com 2024-03-05T18:33:08+02:00 grammars : blacklists character control set (#5888)
82cb31eb93fd19b74115e0f0133225d1dfdbfdbc b1a4e994fde929300d4aeb1deb8320c59cb6edec Georgi Gerganov ggerganov@gmail.com 2024-03-05T15:56:24+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-05T15:56:24+02:00 Revert "grammars : don't allow to output unescaped new line in string (#5885)"
b1a4e994fde929300d4aeb1deb8320c59cb6edec 61d1c88e155515dd03940913a5707ea84a8b119b ExtReMLapin 3909752+ExtReMLapin@users.noreply.github.com 2024-03-05T14:44:29+01:00 GitHub noreply@github.com 2024-03-05T15:44:29+02:00 grammars : don't allow to output unescaped new line in string (#5885)
61d1c88e155515dd03940913a5707ea84a8b119b 21b08674331e1ea1b599f17c5ca91f0ed173be31 0cc4m picard12@live.de 2024-03-05T13:33:42+01:00 GitHub noreply@github.com 2024-03-05T13:33:42+01:00 Vulkan Improvements (#5835)
21b08674331e1ea1b599f17c5ca91f0ed173be31 6a87ac3a52668e117d97bcea07b529c93188b303 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-05T16:08:35+08:00 GitHub noreply@github.com 2024-03-05T13:38:35+05:30 [SYCL] fix mul_mat fault in CI/unit-test (#5862)
6a87ac3a52668e117d97bcea07b529c93188b303 29eee404746e4696143a4f3a642660a4793a15d8 Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-05T15:12:23+09:00 GitHub noreply@github.com 2024-03-05T11:42:23+05:30 fix editorconfig check break (#5879)
29eee404746e4696143a4f3a642660a4793a15d8 1d41d6f7c2a666eb9c18a686a4684c4b03289bf3 Jeffrey Quesnelle emozilla@nousresearch.com 2024-03-04T19:23:06-08:00 GitHub noreply@github.com 2024-03-04T22:23:06-05:00 fix speculative decoding build on windows (#5874)
1d41d6f7c2a666eb9c18a686a4684c4b03289bf3 29ae62d2ae163e2b68aa0ad3bf2ab4636de0c957 hutli 6594598+hutli@users.noreply.github.com 2024-03-05T02:33:08+01:00 GitHub noreply@github.com 2024-03-04T17:33:08-08:00 nix: static build (#5814)
29ae62d2ae163e2b68aa0ad3bf2ab4636de0c957 e0843afe1b37890b631bc7d3d2da2ed36c862b91 Georgi Gerganov ggerganov@gmail.com 2024-03-04T22:31:20+02:00 GitHub noreply@github.com 2024-03-04T22:31:20+02:00 llama : fix embeddings (#5796)
e0843afe1b37890b631bc7d3d2da2ed36c862b91 a1c6d96ed8f906aa1cda439f7386b1171a22bf9f Georgi Gerganov ggerganov@gmail.com 2024-03-04T21:50:50+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-04T21:50:50+02:00 flake : fix
a1c6d96ed8f906aa1cda439f7386b1171a22bf9f efd8533ef8d0752cef7119eb5dbee412c4dba270 Georgi Gerganov ggerganov@gmail.com 2024-03-04T20:53:27+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-04T20:54:23+02:00 ggml : fix unknown status (#0)
efd8533ef8d0752cef7119eb5dbee412c4dba270 9fa262734733573fa629ffc97dfcb971fe3f4832 Georgi Gerganov ggerganov@gmail.com 2024-03-04T11:06:39+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-04T20:54:23+02:00 sync : ggml
9fa262734733573fa629ffc97dfcb971fe3f4832 fe52be11e35358d2fd249f19d7ef5b6f9c08b16b Michael Podvitskiy podvitskiymichael@gmail.com 2024-03-04T10:05:42+01:00 Georgi Gerganov ggerganov@gmail.com 2024-03-04T20:54:23+02:00 ggml : introduce ggml_status (ggml/750)
fe52be11e35358d2fd249f19d7ef5b6f9c08b16b 6d341ab6c53cd51f2921d986d0090cc8b049b39a Dane Madsen dane_madsen@hotmail.com 2024-03-05T05:26:55+11:00 GitHub noreply@github.com 2024-03-04T20:26:55+02:00 cmake : handle cases where git index is not found in .git (#5844)
6d341ab6c53cd51f2921d986d0090cc8b049b39a 4ffcdce2ff877ebb683cd217ea38faf20faa5ffe Minsoo Cheong 54794500+mscheong01@users.noreply.github.com 2024-03-05T03:24:00+09:00 GitHub noreply@github.com 2024-03-04T20:24:00+02:00 speculative : implement stochastic speculative sampling (#5625)
4ffcdce2ff877ebb683cd217ea38faf20faa5ffe a0fc62661f0fd2a9edd10ae5617345bbbf972f42 Xuan Son Nguyen thichthat@gmail.com 2024-03-04T12:22:08+01:00 GitHub noreply@github.com 2024-03-04T12:22:08+01:00 add alias for chat template (#5858)
a0fc62661f0fd2a9edd10ae5617345bbbf972f42 7d43c585dc174bb586775c22c15e5db9242b5b4b Georgi Gerganov ggerganov@gmail.com 2024-03-04T10:40:04+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-04T10:40:04+02:00 sync : ggml
7d43c585dc174bb586775c22c15e5db9242b5b4b 82f3e668adafba647de703f835991e91a96b5ac4 leejet leejet714@gmail.com 2024-03-03T20:23:52+08:00 Georgi Gerganov ggerganov@gmail.com 2024-03-04T10:39:10+02:00 add some new ops, fix some operators and add batch operations to certain operators. (ggml/747)
82f3e668adafba647de703f835991e91a96b5ac4 5a51cc1bb4592f0d71f9af89cd08b11a066ba447 DAN™ dranger003@gmail.com 2024-03-04T03:08:19-05:00 GitHub noreply@github.com 2024-03-04T10:08:19+02:00 common : use LLAMA_DEFAULT_SEED (#5855)
5a51cc1bb4592f0d71f9af89cd08b11a066ba447 67be2ce1015d070b3b2cd488bcb041eefb61de72 DAN™ dranger003@gmail.com 2024-03-04T02:57:20-05:00 GitHub noreply@github.com 2024-03-04T09:57:20+02:00 main : support special tokens as reverse/anti prompt (#5847)
67be2ce1015d070b3b2cd488bcb041eefb61de72 231ae28f078c3148d097b301f2145f1e3e816cc1 slaren slarengh@gmail.com 2024-03-03T14:26:18+01:00 GitHub noreply@github.com 2024-03-03T14:26:18+01:00 cuda : fix data race in soft max (#5853)
231ae28f078c3148d097b301f2145f1e3e816cc1 475df1d6cf817060028d3ff763cb8097d4ec40d6 Georgi Gerganov ggerganov@gmail.com 2024-03-03T12:44:03+02:00 GitHub noreply@github.com 2024-03-03T12:44:03+02:00 readme : add API changes section
475df1d6cf817060028d3ff763cb8097d4ec40d6 87c2e8b2797860a06af3d6c06b8488a8ff1a09ab Douglas Hanley thesecretaryofwar@gmail.com 2024-03-03T04:40:27-06:00 GitHub noreply@github.com 2024-03-03T12:40:27+02:00 llama : allow for user specified embedding pooling type (#5849)
87c2e8b2797860a06af3d6c06b8488a8ff1a09ab de9692a7d2db66e29e5cb373c6551acc49145ccd Nindaleth Nindaleth@users.noreply.github.com 2024-03-03T09:43:42+01:00 GitHub noreply@github.com 2024-03-03T10:43:42+02:00 gguf-dump : support i-quants (#5841)
de9692a7d2db66e29e5cb373c6551acc49145ccd e6029348e86c3810d4435faee54ba822cb43e2ef compilade 113953597+compilade@users.noreply.github.com 2024-03-03T03:41:55-05:00 GitHub noreply@github.com 2024-03-03T10:41:55+02:00 llama : fix llama_copy_state_data with fragmented KV cache (#5840)
e6029348e86c3810d4435faee54ba822cb43e2ef 8ef969afcec1645d2d9c3ab1fc82263bba968989 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-03T09:35:23+01:00 GitHub noreply@github.com 2024-03-03T10:35:23+02:00 ci : schedule slow server tests only on Release or on demand (#5839)
8ef969afcec1645d2d9c3ab1fc82263bba968989 fa974646e1a2024fc7dc9e6f27cf1f2f5d4a3763 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-03T08:48:36+01:00 GitHub noreply@github.com 2024-03-03T09:48:36+02:00 server : init http requests thread pool with --parallel if set (#5836)
fa974646e1a2024fc7dc9e6f27cf1f2f5d4a3763 9731134296af3a6839cd682e51d9c2109a871de5 Georgi Gerganov ggerganov@gmail.com 2024-03-03T06:11:31+02:00 GitHub noreply@github.com 2024-03-02T20:11:31-08:00 flake.lock: Update (#5842)
9731134296af3a6839cd682e51d9c2109a871de5 4a6e2d6142ab815c964924896891e9ab3e050632 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-02T22:00:14+01:00 GitHub noreply@github.com 2024-03-02T22:00:14+01:00 server: tests: passkey challenge / self-extend with context shift demo (#5832)
4a6e2d6142ab815c964924896891e9ab3e050632 494c87032613e31c0be99b2735e732871f2c4e4d Michael Podvitskiy podvitskiymichael@gmail.com 2024-03-02T20:52:25+01:00 GitHub noreply@github.com 2024-03-02T21:52:25+02:00 llama : add abort_callback to interrupt computation (#5409)
494c87032613e31c0be99b2735e732871f2c4e4d 4d4d2366fc9c54d4a275065cfe9299c6cf7c5b78 Georgi Gerganov ggerganov@gmail.com 2024-03-02T20:00:49+02:00 GitHub noreply@github.com 2024-03-02T20:00:49+02:00 ggml : fix IQ3_S AVX implementation (#5834)
4d4d2366fc9c54d4a275065cfe9299c6cf7c5b78 c7a0ad8ec9ebb5ddb1c1c80c82f2ee041c525d47 Jared Van Bortel jared@nomic.ai 2024-03-02T12:27:26-05:00 GitHub noreply@github.com 2024-03-02T12:27:26-05:00 convert : automatically fall back to HfVocab if tokenizer.model doesn't exist (#5821)
c7a0ad8ec9ebb5ddb1c1c80c82f2ee041c525d47 bbde6eb2561153aabbdfac5001c690fe00cad639 Jared Van Bortel jared@nomic.ai 2024-03-02T12:21:47-05:00 GitHub noreply@github.com 2024-03-02T12:21:47-05:00 convert-hf : make model class definitions self-contained (#5825)
bbde6eb2561153aabbdfac5001c690fe00cad639 ef2cd694c4155fbf25bae61c5178c47eb3676dba Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-03-02T17:00:51+02:00 GitHub noreply@github.com 2024-03-02T17:00:51+02:00 ggml : IQ3_S improvements (#5829)
ef2cd694c4155fbf25bae61c5178c47eb3676dba 6c32d8c7ad8ba7b6ad2a162e929a21dd04fcdca0 Georgi Gerganov ggerganov@gmail.com 2024-03-02T16:54:08+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-02T16:54:20+02:00 scripts : add pod-llama.sh
6c32d8c7ad8ba7b6ad2a162e929a21dd04fcdca0 802da0091ba646ecf02e1a8fae2da0b8e76409bd Xuan Son Nguyen thichthat@gmail.com 2024-03-02T15:19:09+01:00 GitHub noreply@github.com 2024-03-02T16:19:09+02:00 llama : refactor internal quantization functions (#5830)
802da0091ba646ecf02e1a8fae2da0b8e76409bd 715641391dda1ff9762dc5d99d9a30acce99f2c6 compilade 113953597+compilade@users.noreply.github.com 2024-03-02T08:42:56-05:00 GitHub noreply@github.com 2024-03-02T15:42:56+02:00 llama : fix segfault from unknown model arch name (#5820)
715641391dda1ff9762dc5d99d9a30acce99f2c6 9bf297a02bfbd474e51912409a470dd797e2fe13 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-03-02T19:49:30+08:00 GitHub noreply@github.com 2024-03-02T19:49:30+08:00 Support multiple GPUs (split mode) on SYCL backend (#5806)
9bf297a02bfbd474e51912409a470dd797e2fe13 cb5e8f7fc4ee57d4bcccafbe04a82cededd35486 crasm crasm@git.vczf.net 2024-03-02T00:11:06-05:00 GitHub noreply@github.com 2024-03-02T00:11:06-05:00 workflows : remove nocleanup arg for check-requirements.sh (#5826)
cb5e8f7fc4ee57d4bcccafbe04a82cededd35486 da3b9ba2b710c0f8b44398a0eb9e5a7ae2ad967a Tushar ditsuke@protonmail.com 2024-03-02T04:48:26+05:30 GitHub noreply@github.com 2024-03-01T15:18:26-08:00 build(nix): Introduce flake.formatter for `nix fmt` (#5687)
da3b9ba2b710c0f8b44398a0eb9e5a7ae2ad967a c29af7e2252d288f2ea58a7d437c1cb7c0abf160 nold Nold360@users.noreply.github.com 2024-03-01T22:51:12+01:00 GitHub noreply@github.com 2024-03-01T16:51:12-05:00 convert-hf-to-gguf : require einops for InternLM2ForCausalLM (#5792)
c29af7e2252d288f2ea58a7d437c1cb7c0abf160 38d16b142624bdd7c41d9955752b7f7b59c5e048 Sourab Mangrulkar 13534540+pacman100@users.noreply.github.com 2024-03-02T01:00:46+05:30 GitHub noreply@github.com 2024-03-01T21:30:46+02:00 llama : add StarCoder2 support (#5795)
38d16b142624bdd7c41d9955752b7f7b59c5e048 c2224f003bf9cf558b1a3c57033563e11a4de9a5 Georgi Gerganov ggerganov@gmail.com 2024-03-01T20:00:58+02:00 GitHub noreply@github.com 2024-03-01T20:00:58+02:00 server : remove api_like_OAI.py proxy script (#5808)
c2224f003bf9cf558b1a3c57033563e11a4de9a5 e7433867288d2f142cffe596f3751bda5d7ee2c7 ddpasa 112642920+ddpasa@users.noreply.github.com 2024-03-01T18:00:00+01:00 GitHub noreply@github.com 2024-03-01T18:00:00+01:00 ggml-vulkan: fix VULKAN_CHECK_RESULTS flag, which was previously broken (#5813)
e7433867288d2f142cffe596f3751bda5d7ee2c7 f49a5356865ced0eca1df9f9d84631dfef71b9dc kunal-vaishnavi 115581922+kunal-vaishnavi@users.noreply.github.com 2024-03-01T06:08:08-08:00 GitHub noreply@github.com 2024-03-01T16:08:08+02:00 gemma : fix bfloat16 -> float16 conversion issue (#5810)
f49a5356865ced0eca1df9f9d84631dfef71b9dc 3ab8b3a92ede46df88bc5a2dfca3777de4a2b2b6 Miwa / Ensan 63481257+ensan-hcl@users.noreply.github.com 2024-03-01T22:48:56+09:00 GitHub noreply@github.com 2024-03-01T15:48:56+02:00 common : fix flag `--logits-all` to `--all-logits` (#5805)
3ab8b3a92ede46df88bc5a2dfca3777de4a2b2b6 9600d59e010c18f5872580a21734ea1bf1968d04 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-01T12:39:06+01:00 GitHub noreply@github.com 2024-03-01T13:39:06+02:00 llama : cleanup unused mmq flags (#5772)
9600d59e010c18f5872580a21734ea1bf1968d04 5cb02b4a012bb16c6c699c0c62c05ffa653eee0f Douglas Hanley thesecretaryofwar@gmail.com 2024-03-01T03:15:36-06:00 GitHub noreply@github.com 2024-03-01T11:15:36+02:00 unicode : switch to multimap based nfd_map (#5799)
5cb02b4a012bb16c6c699c0c62c05ffa653eee0f 6ea0f010ff6967034528d9e0b8330b9b0f0b7c13 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-03-01T10:08:08+01:00 GitHub noreply@github.com 2024-03-01T10:08:08+01:00 server: allow to override threads server pool with --threads-http (#5794)
6ea0f010ff6967034528d9e0b8330b9b0f0b7c13 f105471ef6aa4727afac8240da398590d7277f45 Eve 139727413+netrunnereve@users.noreply.github.com 2024-03-01T08:54:53Z GitHub noreply@github.com 2024-03-01T10:54:53+02:00 ci : add Ubuntu 22 Vulkan CI run (#5789)
f105471ef6aa4727afac8240da398590d7277f45 38d152160898b0173ffe4dc7df5daadcbd2eceb0 Georgi Gerganov ggerganov@gmail.com 2024-03-01T09:59:43+02:00 Georgi Gerganov ggerganov@gmail.com 2024-03-01T09:59:43+02:00 server : fix newlines in help (#5785)
38d152160898b0173ffe4dc7df5daadcbd2eceb0 052051d8ae4639a1c3c61e7da3237bcc572469d4 AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-03-01T07:36:47Z GitHub noreply@github.com 2024-03-01T13:06:47+05:30 [SYCL] Use batched mul_mat pathway (#5591)
052051d8ae4639a1c3c61e7da3237bcc572469d4 d5ab29757ebc59a30f03e408294ec20628a6374e Xuan Son Nguyen thichthat@gmail.com 2024-02-29T21:42:11+01:00 GitHub noreply@github.com 2024-02-29T21:42:11+01:00 Server: normalize naming (#5779)
d5ab29757ebc59a30f03e408294ec20628a6374e 87c91c07663b707e831c59ec373b5e665ff9d64a Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2024-02-29T00:17:23-08:00 GitHub noreply@github.com 2024-02-29T10:17:23+02:00 llama : constified `llama_set_state_data`'s `src` (#5774)
87c91c07663b707e831c59ec373b5e665ff9d64a 317709b2a81dbaf87850202686ec5bb2602a504e Georgi Gerganov ggerganov@gmail.com 2024-02-28T21:44:21+02:00 GitHub noreply@github.com 2024-02-28T21:44:21+02:00 ci : reduce 3b ppl chunks to 1 to avoid timeout (#5771)
317709b2a81dbaf87850202686ec5bb2602a504e 08c5ee87e4cceb603ecceac90734fcdade57311b Eve 139727413+netrunnereve@users.noreply.github.com 2024-02-28T19:33:37Z GitHub noreply@github.com 2024-02-28T20:33:37+01:00 make portability_enumeration_ext apple only (#5757)
08c5ee87e4cceb603ecceac90734fcdade57311b 78aacf36344df724cdca9f1e1af849b2d2519cb8 Georgi Gerganov ggerganov@gmail.com 2024-02-28T18:43:38+02:00 GitHub noreply@github.com 2024-02-28T18:43:38+02:00 llama : remove deprecated API (#5770)
78aacf36344df724cdca9f1e1af849b2d2519cb8 8c0e8f4e73e275756ad69f9c99b26ead085ca9f0 Georgi Gerganov ggerganov@gmail.com 2024-02-28T17:36:53+02:00 GitHub noreply@github.com 2024-02-28T17:36:53+02:00 awq-py : remove (#5768)
8c0e8f4e73e275756ad69f9c99b26ead085ca9f0 2774b0c97427ee3ad3e2ee121354d078794e89d9 Georgi Gerganov ggerganov@gmail.com 2024-02-28T11:17:32+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-28T11:17:32+02:00 sync : ggml
2774b0c97427ee3ad3e2ee121354d078794e89d9 5f706718566e3a5147916dc381f3b99de0ffad47 slaren slarengh@gmail.com 2024-02-25T20:41:35+01:00 Georgi Gerganov ggerganov@gmail.com 2024-02-28T11:17:06+02:00 add google magika inference example (ggml/748)
5f706718566e3a5147916dc381f3b99de0ffad47 a693bea1e6762a17b78b6ddf4611e54136941ea2 UEXTM.com 84163508+uextm@users.noreply.github.com 2024-02-24T11:27:36-05:00 Georgi Gerganov ggerganov@gmail.com 2024-02-28T11:17:05+02:00 Introduce backend GUIDs (ggml/743)
a693bea1e6762a17b78b6ddf4611e54136941ea2 adcb12a9bad87bc96f2f158c95892b3d04aa7ffb Xuan Son Nguyen thichthat@gmail.com 2024-02-28T09:55:37+01:00 GitHub noreply@github.com 2024-02-28T10:55:37+02:00 server : hit Ctrl+C twice to exit (#5734)
adcb12a9bad87bc96f2f158c95892b3d04aa7ffb 177628bfd85565070916ad66a5ac4071ee0527d8 compilade 113953597+compilade@users.noreply.github.com 2024-02-28T03:52:56-05:00 GitHub noreply@github.com 2024-02-28T10:52:56+02:00 llama : fix non-quantization of expert gating tensors (#5754)
177628bfd85565070916ad66a5ac4071ee0527d8 6c4416868df2e5455da7d20547f62bcf9735ba8e Douglas Hanley thesecretaryofwar@gmail.com 2024-02-28T02:51:11-06:00 GitHub noreply@github.com 2024-02-28T10:51:11+02:00 llama : improve BERT tokenization (#5740)
6c4416868df2e5455da7d20547f62bcf9735ba8e efc72253f7987ed7bdc8bde9d9fa5c7cac2f6292 Daniel Bevenius daniel.bevenius@gmail.com 2024-02-28T09:39:39+01:00 GitHub noreply@github.com 2024-02-28T10:39:39+02:00 readme : add link to LLaVA 1.6 models (#5758)
efc72253f7987ed7bdc8bde9d9fa5c7cac2f6292 7c4263d4261d6ee6f0539d53eb9e1b4d120ba8af Jorge A 161275481+jorgealias@users.noreply.github.com 2024-02-28T01:39:15-07:00 GitHub noreply@github.com 2024-02-28T10:39:15+02:00 server : add "/chat/completions" alias for "/v1/...` (#5722)
7c4263d4261d6ee6f0539d53eb9e1b4d120ba8af cb49e0f8c906e5da49e9f6d64a57742a9a241c6a Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-28T10:37:02+02:00 GitHub noreply@github.com 2024-02-28T10:37:02+02:00 ggml : make i-quants work with super-blocks of 64 (CPU,Metal) (#5760)
cb49e0f8c906e5da49e9f6d64a57742a9a241c6a 0becb22ac05b6542bd9d5f2235691aa1d3d4d307 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-27T19:16:49+02:00 GitHub noreply@github.com 2024-02-27T19:16:49+02:00 Attempt to fix android build (#5752)
0becb22ac05b6542bd9d5f2235691aa1d3d4d307 c24a2a6e6005e5d424301525a42ba45a4a362d30 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-27T16:34:24+02:00 GitHub noreply@github.com 2024-02-27T16:34:24+02:00 IQ4_XS: a 4.25 bpw quantization (#5747)
c24a2a6e6005e5d424301525a42ba45a4a362d30 1f30b7a9f1b86baa455072d3182b9ebeee0cd845 Engininja2 139037756+Engininja2@users.noreply.github.com 2024-02-27T07:22:45-06:00 GitHub noreply@github.com 2024-02-27T14:22:45+01:00 cuda : replace remaining shfl_xor with calls to warp_reduce functions (#5744)
1f30b7a9f1b86baa455072d3182b9ebeee0cd845 9d533a77d0c3850ce09d736bc1baa67fd6ad27b3 Engininja2 139037756+Engininja2@users.noreply.github.com 2024-02-27T06:50:18-06:00 GitHub noreply@github.com 2024-02-27T14:50:18+02:00 ggml-quants : fix avx2 iq1_s vec_dot when compiled with gcc (#5742)
9d533a77d0c3850ce09d736bc1baa67fd6ad27b3 cbbd1efa06f8c09f9dff58ff9d9af509cc4c152b Georgi Gerganov ggerganov@gmail.com 2024-02-27T14:35:51+02:00 GitHub noreply@github.com 2024-02-27T14:35:51+02:00 llama : fix defrag bugs + add parameter (#5735)
cbbd1efa06f8c09f9dff58ff9d9af509cc4c152b b11a93df41921846a10628a7c306d5c82a549939 le.chang cljs118@126.com 2024-02-27T10:03:06+08:00 GitHub noreply@github.com 2024-02-27T03:03:06+01:00 Makefile: use variables for cublas (#5689)
b11a93df41921846a10628a7c306d5c82a549939 a33e6a0d2a66104ea9a906bdbf8a94d050189d91 Xuan Son Nguyen thichthat@gmail.com 2024-02-26T23:15:48+01:00 GitHub noreply@github.com 2024-02-26T23:15:48+01:00 fix server hangs on empty prompt (#5733)
a33e6a0d2a66104ea9a906bdbf8a94d050189d91 47bb7b48c7cec9d8f57d56812ce811ec130b89a3 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-26T18:28:38+02:00 GitHub noreply@github.com 2024-02-26T18:28:38+02:00 Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721)
47bb7b48c7cec9d8f57d56812ce811ec130b89a3 c4d7f8178608440506e5489bae0109e4ca12e44a Johannes Gäßler johannesg@5d6.de 2024-02-26T15:36:38+01:00 GitHub noreply@github.com 2024-02-26T15:36:38+01:00 CUDA: fix DEBUG_CUDA_MALLOC (#5729)
c4d7f8178608440506e5489bae0109e4ca12e44a e849078c6e09e72fdd2c95ba61f5fba9a7b2d9ef Artem guinmoon@gmail.com 2024-02-26T17:15:28+03:00 GitHub noreply@github.com 2024-02-26T16:15:28+02:00 readme : update ui list (#5731)
e849078c6e09e72fdd2c95ba61f5fba9a7b2d9ef 67fd33132fab93e6c2087bd6fa656a8a57419efa AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-02-26T14:02:11Z GitHub noreply@github.com 2024-02-26T19:32:11+05:30 [SYCL] Add support for soft_max ALiBi (#5639)
67fd33132fab93e6c2087bd6fa656a8a57419efa 4804215cb833841ffb15a710a16b77ca0a29eb4b Georgi Gerganov ggerganov@gmail.com 2024-02-26T14:02:12+02:00 GitHub noreply@github.com 2024-02-26T14:02:12+02:00 unicode : reuse iterator (#5726)
4804215cb833841ffb15a710a16b77ca0a29eb4b 8a533f0d9078396ebaee9ba213038a1322976dee Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-26T11:41:34+01:00 GitHub noreply@github.com 2024-02-26T12:41:34+02:00 server: CI fix trailing space (#5728)
8a533f0d9078396ebaee9ba213038a1322976dee 269de86ba073b5dc9ce687c11a3bc4d7d873b962 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-26T09:56:10+01:00 GitHub noreply@github.com 2024-02-26T09:56:10+01:00 server: CI tests reduce build matrix (#5725)
269de86ba073b5dc9ce687c11a3bc4d7d873b962 c39373398803c669056304090050fe3f44b41bf9 Georgi Gerganov ggerganov@gmail.com 2024-02-26T08:30:17+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-26T08:30:17+02:00 llama : fix Gemma rope type (#5691)
c39373398803c669056304090050fe3f44b41bf9 e3965cf35aac00d4e24998c8a3d0093ae1d98bd3 github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-02-25T00:17:11Z Someone newkozlukov@gmail.com 2024-02-25T22:24:22Z flake.lock: Update
e3965cf35aac00d4e24998c8a3d0093ae1d98bd3 8b350356b28f782deab63d8b0e9ae103ceb25fcd Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-25T22:48:33+01:00 GitHub noreply@github.com 2024-02-25T22:48:33+01:00 server: tests - slow inference causes timeout on the CI (#5715)
8b350356b28f782deab63d8b0e9ae103ceb25fcd bf08e00643fd529f748f0a858fd79f3061e3fa18 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-25T21:46:29+01:00 GitHub noreply@github.com 2024-02-25T21:46:29+01:00 server: docs - refresh and tease a little bit more the http server (#5718)
bf08e00643fd529f748f0a858fd79f3061e3fa18 f7625019c51ca437a5840576d92362cfa710e4a2 Georgi Gerganov ggerganov@gmail.com 2024-02-25T22:12:24+02:00 GitHub noreply@github.com 2024-02-25T22:12:24+02:00 llama : refactor k-shift implementation + KV defragmentation (#5691)
f7625019c51ca437a5840576d92362cfa710e4a2 abbabc5e51d0d4656b438aec10b7fae9479ef37d compilade 113953597+compilade@users.noreply.github.com 2024-02-25T13:43:50-05:00 GitHub noreply@github.com 2024-02-25T20:43:50+02:00 server : fix crash when system prompt is bigger than batch size (#5714)
abbabc5e51d0d4656b438aec10b7fae9479ef37d f1a98c52546d009f742bdec2154c2a314ea950a6 Radosław Gryta radek.gryta@gmail.com 2024-02-25T19:43:00+01:00 GitHub noreply@github.com 2024-02-25T20:43:00+02:00 ggml-quants : provide ggml_vqtbl1q_u8 for 64bit compatibility (#5711)
f1a98c52546d009f742bdec2154c2a314ea950a6 7d548a1827f6fc6aece6db74c9d112da42c40d68 kwin1412 42286931+kwin1412@users.noreply.github.com 2024-02-26T00:46:49+08:00 GitHub noreply@github.com 2024-02-25T18:46:49+02:00 make : fix nvcc version is empty (#5713)
7d548a1827f6fc6aece6db74c9d112da42c40d68 930b1780269a69948d106e2d1b838ab7661f679a Ashok Gelal 401055+ashokgelal@users.noreply.github.com 2024-02-25T10:57:34-05:00 GitHub noreply@github.com 2024-02-25T17:57:34+02:00 readme : add Msty to UI list (#5618)
930b1780269a69948d106e2d1b838ab7661f679a d52d7819b8ced70c642a88a59da8c78208dc58ec Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-25T13:50:32+01:00 GitHub noreply@github.com 2024-02-25T13:50:32+01:00 server: logs - unified format and --log-format option (#5700)
d52d7819b8ced70c642a88a59da8c78208dc58ec 12894088170f62e4cad4f8d6a3043c185b414bab Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-25T13:49:43+01:00 GitHub noreply@github.com 2024-02-25T13:49:43+01:00 server: concurrency fix + monitoring - add /metrics prometheus compatible endpoint (#5708)
12894088170f62e4cad4f8d6a3043c185b414bab ab336a9d5e5352ecdcdf4c12d2d54cf4ef82ce31 Radosław Gryta radek.gryta@gmail.com 2024-02-25T11:53:11+01:00 GitHub noreply@github.com 2024-02-25T12:53:11+02:00 cmake : fix compilation for Android armeabi-v7a (#5702)
ab336a9d5e5352ecdcdf4c12d2d54cf4ef82ce31 69917dfa55674c608360638bb4d6a12a315e2810 Georgi Gerganov ggerganov@gmail.com 2024-02-25T12:09:09+02:00 GitHub noreply@github.com 2024-02-25T12:09:09+02:00 code : normalize enum names (#5697)
69917dfa55674c608360638bb4d6a12a315e2810 9e359a4f47c1b2dceb99e29706c9f7403d32ab5e Anas Ahouzi 112881240+aahouzi@users.noreply.github.com 2024-02-25T10:54:04+01:00 GitHub noreply@github.com 2024-02-25T11:54:04+02:00 py : fix StableLM conversion after config.json changes (#5703)
9e359a4f47c1b2dceb99e29706c9f7403d32ab5e 4c4cb30736582cacb1a164a9d4bc8e17b1014be7 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-24T19:16:04+01:00 GitHub noreply@github.com 2024-02-24T19:16:04+01:00 server: continue to update other slots on embedding concurrent request (#5699)
4c4cb30736582cacb1a164a9d4bc8e17b1014be7 525213d2f5da1eaf4b922b6b792cb52b2c613368 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-24T16:23:52+02:00 GitHub noreply@github.com 2024-02-24T16:23:52+02:00 IQ3_S: a much better alternative to Q3_K (#5676)
525213d2f5da1eaf4b922b6b792cb52b2c613368 fd43d66f46ee3b5345fb8a74a252d86ccd34a409 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-24T12:28:55+01:00 GitHub noreply@github.com 2024-02-24T12:28:55+01:00 server: init functional tests (#5566)
fd43d66f46ee3b5345fb8a74a252d86ccd34a409 54fbcd2ce6c48c9e22eca6fbf9e53fb68c3e72ea AlpinDale 52078762+AlpinDale@users.noreply.github.com 2024-02-23T19:31:54Z GitHub noreply@github.com 2024-02-23T21:31:54+02:00 server : add KV cache quantization options (#5684)
54fbcd2ce6c48c9e22eca6fbf9e53fb68c3e72ea 15499eb94227401bdc8875da6eb85c15d37068f7 Jared Van Bortel jared@nomic.ai 2024-02-23T13:39:14-05:00 GitHub noreply@github.com 2024-02-23T20:39:14+02:00 convert : fix missing ftype for gemma (#5690)
15499eb94227401bdc8875da6eb85c15d37068f7 96633eeca1265ed03e57230de54032041c58f9cd Jared Van Bortel jared@nomic.ai 2024-02-22T17:05:23-05:00 GitHub noreply@github.com 2024-02-22T17:05:23-05:00 mpt : do not duplicate token_embd.weight on disk (#5670)
96633eeca1265ed03e57230de54032041c58f9cd 847eedbdb2d1ebf14ef56eb507d4b4b975510908 Georgi Gerganov ggerganov@gmail.com 2024-02-22T23:23:46+02:00 GitHub noreply@github.com 2024-02-22T23:23:46+02:00 gemma : use more bits for the token_embd.weight tensor (#5650)
847eedbdb2d1ebf14ef56eb507d4b4b975510908 7e4f339c404dbe029d4a117c03b37a9bf646cf0e Georgi Gerganov ggerganov@gmail.com 2024-02-22T23:22:48+02:00 GitHub noreply@github.com 2024-02-22T23:22:48+02:00 py : add Gemma conversion from HF models (#5647)
7e4f339c404dbe029d4a117c03b37a9bf646cf0e 334f76fa385ed81095165e5ae068756214893901 Georgi Gerganov ggerganov@gmail.com 2024-02-22T23:21:39+02:00 GitHub noreply@github.com 2024-02-22T23:21:39+02:00 ggml : always define ggml_fp16_t as uint16_t (#5666)
334f76fa385ed81095165e5ae068756214893901 efd56b1c2139d50b9b4381a212feb75d69598fda Georgi Gerganov ggerganov@gmail.com 2024-02-22T23:21:05+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-22T23:21:05+02:00 sync : ggml
efd56b1c2139d50b9b4381a212feb75d69598fda 201294ae177b308fb3a99dc504dd6d27e8afa907 Georgi Gerganov ggerganov@gmail.com 2024-02-22T18:31:40+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-22T23:20:50+02:00 ggml : 32-bit arm compat (whisper/1891)
201294ae177b308fb3a99dc504dd6d27e8afa907 5a9e2f60ba3d8362ba17c77ac3092906d49b813f Someone sergei.kozlukov@aalto.fi 2024-02-22T19:44:10Z GitHub noreply@github.com 2024-02-22T11:44:10-08:00 nix: init singularity and docker images (#5056)
5a9e2f60ba3d8362ba17c77ac3092906d49b813f 373ee3fbbabc4c1508eed4f5c3795b23a20939a3 Georgi Gerganov ggerganov@gmail.com 2024-02-22T20:13:25+02:00 GitHub noreply@github.com 2024-02-22T20:13:25+02:00 py : minor fixes (#5668)
373ee3fbbabc4c1508eed4f5c3795b23a20939a3 4cb4d8b22d4fda971621a68c570ce84d66897c37 Xuan Son Nguyen thichthat@gmail.com 2024-02-22T19:10:21+01:00 GitHub noreply@github.com 2024-02-22T19:10:21+01:00 Add Gemma chat template (#5665)
4cb4d8b22d4fda971621a68c570ce84d66897c37 3a03541cedea474fa9d41214484cc3fbcf468a9e Someone sergei.kozlukov@aalto.fi 2024-02-22T16:32:09Z GitHub noreply@github.com 2024-02-22T08:32:09-08:00 workflows: nix: hardcode cachix ids, build unconditionally (#5663)
3a03541cedea474fa9d41214484cc3fbcf468a9e 56d03d92be57f5880b9ed94542d87bb6effae31f Georgi Gerganov ggerganov@gmail.com 2024-02-22T13:54:03+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-22T13:54:03+02:00 minor : fix trailing whitespace (#5638)
56d03d92be57f5880b9ed94542d87bb6effae31f a46f50747b2028f7f9c9883b26bfba12bf92556e Georgi Gerganov ggerganov@gmail.com 2024-02-22T10:35:54+02:00 GitHub noreply@github.com 2024-02-22T10:35:54+02:00 readme : update hot topics
a46f50747b2028f7f9c9883b26bfba12bf92556e c5688c6250430d2b8e0259efcf26c16dfa4c1f46 Xuan Son Nguyen thichthat@gmail.com 2024-02-22T09:33:24+01:00 GitHub noreply@github.com 2024-02-22T10:33:24+02:00 server : fallback to chatml, add AlphaMonarch chat template (#5628)
c5688c6250430d2b8e0259efcf26c16dfa4c1f46 4ef245a92a968ba0f18a5adfd41e51980ce4fdf5 Alexey Parfenov zxed@alkatrazstudio.net 2024-02-22T08:27:32Z GitHub noreply@github.com 2024-02-22T10:27:32+02:00 server : clarify some params in the docs (#5640)
4ef245a92a968ba0f18a5adfd41e51980ce4fdf5 973053d8b0d04809836b3339a50f68d9c842de90 Dat Quoc Nguyen 2412555+datquocnguyen@users.noreply.github.com 2024-02-22T18:15:13+10:00 GitHub noreply@github.com 2024-02-22T10:15:13+02:00 mpt : add optional bias tensors (#5638)
973053d8b0d04809836b3339a50f68d9c842de90 7c8bcc11dc61cf5930b70cd0168b84afcebe12a9 slaren slarengh@gmail.com 2024-02-22T00:42:09+01:00 GitHub noreply@github.com 2024-02-22T00:42:09+01:00 llama : fix loading models with shared tok_embd and output (#5651)
7c8bcc11dc61cf5930b70cd0168b84afcebe12a9 7fe4678b0244ba7b03eae66ebeaa947e2770bb1a Xuan Son Nguyen thichthat@gmail.com 2024-02-22T00:31:00+01:00 GitHub noreply@github.com 2024-02-22T00:31:00+01:00 Add docs for llama_chat_apply_template (#5645)
7fe4678b0244ba7b03eae66ebeaa947e2770bb1a ba2135ccae7462470b3865c6e41d2e1d734eac05 slaren slarengh@gmail.com 2024-02-21T22:52:39+01:00 GitHub noreply@github.com 2024-02-21T22:52:39+01:00 llama : fix session save/load with quantized KV (#5649)
ba2135ccae7462470b3865c6e41d2e1d734eac05 89febfed9322c8849520dc63c93ee4f5fd72556e slaren slarengh@gmail.com 2024-02-21T22:18:23+01:00 GitHub noreply@github.com 2024-02-21T22:18:23+01:00 gemma : allow offloading the output tensor (#5646)
89febfed9322c8849520dc63c93ee4f5fd72556e 5022cf242d689e15defd133f96c4345ad30c5d19 Jared Van Bortel jared@nomic.ai 2024-02-21T10:33:54-05:00 GitHub noreply@github.com 2024-02-21T10:33:54-05:00 examples : do not assume BOS when shifting context (#5622)
5022cf242d689e15defd133f96c4345ad30c5d19 1ecea255ebb70750b52688393f37a63606b90e3f Georgi Gerganov ggerganov@gmail.com 2024-02-21T16:52:39+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-21T16:52:52+02:00 sync : ggml
1ecea255ebb70750b52688393f37a63606b90e3f a00a35cef93e057eace8351a667d14d152a91ebc Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-21T15:47:48+01:00 GitHub noreply@github.com 2024-02-21T15:47:48+01:00 server: health: fix race condition on slots data using tasks queue (#5634)
a00a35cef93e057eace8351a667d14d152a91ebc eccd7a26ddbff19e4b8805648f5f14c501957859 Ettore Di Giacinto mudler@users.noreply.github.com 2024-02-21T15:39:10+01:00 GitHub noreply@github.com 2024-02-21T16:39:10+02:00 readme : add LocalAI to the availables UI (#5629)
eccd7a26ddbff19e4b8805648f5f14c501957859 c14f72db9c62d71d35eb1c141745c0bd0cb27b49 Georgi Gerganov ggerganov@gmail.com 2024-02-21T16:17:10+02:00 GitHub noreply@github.com 2024-02-21T16:17:10+02:00 sync : ggml (#5633)
c14f72db9c62d71d35eb1c141745c0bd0cb27b49 cc6cac08e38e32bf40bbe07e9e8f8f0130b5fd94 Georgi Gerganov ggerganov@gmail.com 2024-02-21T15:39:54+02:00 GitHub noreply@github.com 2024-02-21T15:39:54+02:00 readme : update hot topics
cc6cac08e38e32bf40bbe07e9e8f8f0130b5fd94 580111d42b3b6ad0a390bfb267d6e3077506eb31 Daniel Bevenius daniel.bevenius@gmail.com 2024-02-21T14:36:57+01:00 GitHub noreply@github.com 2024-02-21T15:36:57+02:00 llava : add --skip-unknown to 1.6 convert.py (#5632)
580111d42b3b6ad0a390bfb267d6e3077506eb31 88c46cbdac05cebd936511b1d3c74112e721615f postmasters namnguyen@google.com 2024-02-21T05:08:22-08:00 GitHub noreply@github.com 2024-02-21T15:08:22+02:00 llama : add `gemma` model (#5631)
88c46cbdac05cebd936511b1d3c74112e721615f a14679cc30c785e75d38028bae6ec39c6209ddef Meng, Hengyu hengyu.meng@intel.com 2024-02-21T17:52:06+08:00 GitHub noreply@github.com 2024-02-21T17:52:06+08:00 [SYCL] conext add name (#5624)
a14679cc30c785e75d38028bae6ec39c6209ddef 6560bed3f066c876682464762cad90f1e28e3f1b Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-21T11:39:52+02:00 GitHub noreply@github.com 2024-02-21T11:39:52+02:00 IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590)
6560bed3f066c876682464762cad90f1e28e3f1b 06bf2cf8c406e6b70dbf9b431a02fa0ad845b9df CJ Pais cj@cjpais.com 2024-02-20T11:07:22-08:00 GitHub noreply@github.com 2024-02-20T21:07:22+02:00 server : support llava 1.6 (#5553)
06bf2cf8c406e6b70dbf9b431a02fa0ad845b9df 4ed8e4fbef6a15afd993bfcd9ffa279841e18ef1 slaren slarengh@gmail.com 2024-02-20T20:06:17+01:00 GitHub noreply@github.com 2024-02-20T20:06:17+01:00 make : fix debug build with CUDA (#5616)
4ed8e4fbef6a15afd993bfcd9ffa279841e18ef1 9c405c9f9a7cfd23511fd6b2de05dc72481119b4 Daniel Bevenius daniel.bevenius@gmail.com 2024-02-20T18:30:27+01:00 GitHub noreply@github.com 2024-02-20T19:30:27+02:00 llava : add explicit instructions for llava-1.6 (#5611)
9c405c9f9a7cfd23511fd6b2de05dc72481119b4 5207b3fbc500f89dfe528693e96540956dbaed96 Xuan Son Nguyen thichthat@gmail.com 2024-02-20T15:58:27+01:00 GitHub noreply@github.com 2024-02-20T15:58:27+01:00 Server: use llama_chat_apply_template (#5593)
5207b3fbc500f89dfe528693e96540956dbaed96 8dbbd75754d43ec7b4bbe42fb287cc2553fdf0e9 Dane Madsen dane_madsen@hotmail.com 2024-02-20T21:00:23+11:00 GitHub noreply@github.com 2024-02-20T12:00:23+02:00 readme : update UI list (#5605)
8dbbd75754d43ec7b4bbe42fb287cc2553fdf0e9 c0a8c6db371cb3e4379900867b948879f5842201 Haoxiang Fei tonyfettes@tonyfettes.com 2024-02-19T22:58:36-11:00 GitHub noreply@github.com 2024-02-20T11:58:36+02:00 metal : add build system support for embedded metal library (#5604)
c0a8c6db371cb3e4379900867b948879f5842201 b9111bd209c7b11b0592450a6ed2e0ca545b2c84 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-20T08:48:19+01:00 GitHub noreply@github.com 2024-02-20T09:48:19+02:00 server : health endpoint configurable failure on no slot (#5594)
b9111bd209c7b11b0592450a6ed2e0ca545b2c84 633782b8d949f24b619e6c68ee37b5cc79167173 AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-02-20T07:01:25Z GitHub noreply@github.com 2024-02-20T12:31:25+05:30 Update ggml_sycl_op_mul_mat_vec_q (#5502)
633782b8d949f24b619e6c68ee37b5cc79167173 22f83f0c383e12106692b8afc224d61b8993a52c Mathijs de Bruin mathijs@mathijsfietst.nl 2024-02-13T20:28:02Z Philip Taron philip.taron@gmail.com 2024-02-19T14:49:49-08:00 nix: now that we can do so, allow MacOS to build Vulkan binaries
22f83f0c383e12106692b8afc224d61b8993a52c bb9dcd560a7e81265398b0d463c40f3e467daf19 0cc4m picard12@live.de 2024-02-10T22:18:33+01:00 Philip Taron philip.taron@gmail.com 2024-02-19T14:49:49-08:00 Enable Vulkan MacOS CI
bb9dcd560a7e81265398b0d463c40f3e467daf19 f50db6ae0bdcb5f8593ca6ca46dfa03b177faa2f 0cc4m picard12@live.de 2024-02-14T20:57:17+01:00 Philip Taron philip.taron@gmail.com 2024-02-19T14:49:49-08:00 Refactor validation and enumeration platform checks into functions to clean up ggml_vk_instance_init()
f50db6ae0bdcb5f8593ca6ca46dfa03b177faa2f d8c054517dc24f1316f3be12a98fff383e1e93e3 0cc4m picard12@live.de 2024-02-10T22:14:52+01:00 Philip Taron philip.taron@gmail.com 2024-02-19T14:49:49-08:00 Add check for VK_KHR_portability_enumeration for MoltenVK support
d8c054517dc24f1316f3be12a98fff383e1e93e3 42f664a3825dfde13a32c3577ab66d10c56f3aa6 Mathijs de Bruin mathijs@mathijsfietst.nl 2024-02-06T14:39:22Z Philip Taron philip.taron@gmail.com 2024-02-19T14:49:49-08:00 Add preprocessor checks for Apple devices.
42f664a3825dfde13a32c3577ab66d10c56f3aa6 5dde5408978eda22242b87e22e306d1c2d1a5834 Mathijs de Bruin mathijs@mathijsfietst.nl 2024-02-03T18:00:11Z Philip Taron philip.taron@gmail.com 2024-02-19T14:49:49-08:00 Resolve ErrorIncompatibleDriver with Vulkan on MacOS.
5dde5408978eda22242b87e22e306d1c2d1a5834 40c3a6c1e11040088b4a1ce0abc4651cb3011dd4 Mathijs de Bruin mathijs@mathijsfietst.nl 2024-02-03T17:56:46Z Philip Taron philip.taron@gmail.com 2024-02-19T14:49:49-08:00 Allow for Vulkan build with Accelerate.
40c3a6c1e11040088b4a1ce0abc4651cb3011dd4 f24ed14ee0ce28dfe98115c378b37da144912016 slaren slarengh@gmail.com 2024-02-19T23:40:26+01:00 GitHub noreply@github.com 2024-02-19T23:40:26+01:00 cuda : ignore peer access already enabled errors (#5597)
f24ed14ee0ce28dfe98115c378b37da144912016 9d679f0fccd4030779ed3c7684a40122fe41806c Jared Van Bortel jared@nomic.ai 2024-02-19T15:54:12-05:00 GitHub noreply@github.com 2024-02-19T15:54:12-05:00 make : pass CPPFLAGS directly to nvcc, not via -Xcompiler (#5598)
9d679f0fccd4030779ed3c7684a40122fe41806c 1387cf60f758efb218fa06b670182c38ff149b7b nopperl 54780682+nopperl@users.noreply.github.com 2024-02-19T14:14:07Z GitHub noreply@github.com 2024-02-19T16:14:07+02:00 examples : support minItems/maxItems in JSON grammar converter (#5039)
1387cf60f758efb218fa06b670182c38ff149b7b 6fd413791a754598a54a366145960f2e27eec015 Georgi Gerganov ggerganov@gmail.com 2024-02-19T15:23:17+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-19T15:23:17+02:00 llava : remove extra cont (#5587)
6fd413791a754598a54a366145960f2e27eec015 337c9cbd52918ae5fb9a9d9e25d7fae4e238c9f1 slaren slarengh@gmail.com 2024-02-19T14:02:36+01:00 Georgi Gerganov ggerganov@gmail.com 2024-02-19T15:09:43+02:00 llava : replace ggml_cpy with ggml_cont
337c9cbd52918ae5fb9a9d9e25d7fae4e238c9f1 a3145bdc305422973e25f0b066da6f469ed5dc45 Georgi Gerganov ggerganov@gmail.com 2024-02-19T14:54:21+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-19T15:09:43+02:00 sync : ggml
a3145bdc305422973e25f0b066da6f469ed5dc45 890559ab28e354052e16e770155ad007fd0856e8 Georgi Gerganov ggerganov@gmail.com 2024-02-19T14:53:48+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-19T15:09:43+02:00 ggml-alloc : apply ggml/731
890559ab28e354052e16e770155ad007fd0856e8 d0e3ce51f45bd6a646da1952d7e5d143a087db3e Didzis Gosko didzis@users.noreply.github.com 2024-02-11T16:41:41+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-19T15:09:43+02:00 metal : option to embed MSL source into compiled binary (whisper/1842)
d0e3ce51f45bd6a646da1952d7e5d143a087db3e 68a6b98b3c8af7e5baade3ee45fe1d2c7b9323a9 Georgi Gerganov ggerganov@gmail.com 2024-02-19T14:45:41+02:00 GitHub noreply@github.com 2024-02-19T14:45:41+02:00 ci : enable -Werror for CUDA builds (#5579)
68a6b98b3c8af7e5baade3ee45fe1d2c7b9323a9 70d45af0efce9ed360e1858b827989d971dd9caf Georgi Gerganov ggerganov@gmail.com 2024-02-19T13:41:51+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-19T13:41:51+02:00 make : fix CUDA build (#5580)
70d45af0efce9ed360e1858b827989d971dd9caf 13e2c771aa4212cd5405cf310203848d50f7f859 valiray 133289098+valiray@users.noreply.github.com 2024-02-19T02:37:10-08:00 GitHub noreply@github.com 2024-02-19T12:37:10+02:00 readme : fix typo in README-sycl.md (#5353)
13e2c771aa4212cd5405cf310203848d50f7f859 f53119cec4f073b6d214195ecbe1fad3abdf2b34 Abhilash Majumder 30946547+abhilash1910@users.noreply.github.com 2024-02-19T14:45:18+05:30 GitHub noreply@github.com 2024-02-19T11:15:18+02:00 cmake : remove obsolete sycl compile flags (#5581)
f53119cec4f073b6d214195ecbe1fad3abdf2b34 70847553963c85e86051d06df848236829f5f951 Georgi Gerganov ggerganov@gmail.com 2024-02-19T10:34:10+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-19T10:34:10+02:00 minor : fix trailing whitespace (#5538)
70847553963c85e86051d06df848236829f5f951 4480542b2271ba1438f0daff8e5f3a74b1dc8609 Daniel Bevenius daniel.bevenius@gmail.com 2024-02-19T09:31:59+01:00 GitHub noreply@github.com 2024-02-19T10:31:59+02:00 llava : avoid changing the original BakLLaVA model (#5577)
4480542b2271ba1438f0daff8e5f3a74b1dc8609 11b12de39bd787c0494da0cd405958fdfedc29c4 NawafAlansari 72708095+NawafAlansari@users.noreply.github.com 2024-02-19T03:25:38-05:00 GitHub noreply@github.com 2024-02-19T10:25:38+02:00 baby-llama : allocate graphs in ggml_context (#5573)
11b12de39bd787c0494da0cd405958fdfedc29c4 3a9cb4ca6408c29423373dd6cd7aa78a58286c00 Xuan Son Nguyen thichthat@gmail.com 2024-02-19T09:23:37+01:00 GitHub noreply@github.com 2024-02-19T10:23:37+02:00 llama : add llama_chat_apply_template() (#5538)
3a9cb4ca6408c29423373dd6cd7aa78a58286c00 769a716e30ba1da46f709df1c00727d6869d30e7 slaren slarengh@gmail.com 2024-02-19T09:04:45+01:00 GitHub noreply@github.com 2024-02-19T10:04:45+02:00 cuda, metal : fix nans in soft_max (#5574)
769a716e30ba1da46f709df1c00727d6869d30e7 f0d1fafc029a056cd765bdae58dcaa12312e9879 Mirko185 mirkosig@gmail.com 2024-02-19T08:39:31+01:00 GitHub noreply@github.com 2024-02-19T09:39:31+02:00 readme : update (#5572)
f0d1fafc029a056cd765bdae58dcaa12312e9879 a0c2dad9d43456c677e205c6240a5f8afb0121ac bmwl brian.marshall@tolko.com 2024-02-18T23:38:32-08:00 GitHub noreply@github.com 2024-02-19T09:38:32+02:00 ggml : android and old glibc NUMA incompatibility bugfixes (#5557)
a0c2dad9d43456c677e205c6240a5f8afb0121ac 14278f55d2e2c6a53022075c7f2719b71e1cd61d Jared Van Bortel jared@nomic.ai 2024-02-18T16:21:52-05:00 GitHub noreply@github.com 2024-02-18T16:21:52-05:00 build : pass all warning flags to nvcc via -Xcompiler (#5570)
14278f55d2e2c6a53022075c7f2719b71e1cd61d b1de96824bdbeb91ea458abcb3e5478690ad0727 Georgi Gerganov ggerganov@gmail.com 2024-02-18T22:58:57+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-18T22:58:57+02:00 ggml : restore vec dot stride arg names (#5453)
b1de96824bdbeb91ea458abcb3e5478690ad0727 7ad554f90e735cf2a0f612ce44f9aa4fad6ae46a Georgi Gerganov ggerganov@gmail.com 2024-02-18T22:39:30+02:00 GitHub noreply@github.com 2024-02-18T22:39:30+02:00 ci : fix wikitext url + compile warnings (#5569)
7ad554f90e735cf2a0f612ce44f9aa4fad6ae46a 5ee99c32f5e47c8d32634eff9a47fb32a24c276b Georgi Gerganov ggerganov@gmail.com 2024-02-18T21:39:58+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-18T21:39:58+02:00 metal : fix unused warnings (#0)
5ee99c32f5e47c8d32634eff9a47fb32a24c276b c145f8a132b2fe1d1e65987faddbd9a40bef7a12 Robey Holderith robey@flaminglunchbox.net 2024-02-18T11:11:16-08:00 GitHub noreply@github.com 2024-02-18T21:11:16+02:00 common, server : surface min_keep as its own parameter (#5567)
c145f8a132b2fe1d1e65987faddbd9a40bef7a12 689a091bbe0537ee9abff3e15a1d74f5f3561165 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-18T18:39:57+01:00 GitHub noreply@github.com 2024-02-18T19:39:57+02:00 server : slots monitoring endpoint (#5550)
689a091bbe0537ee9abff3e15a1d74f5f3561165 f3f28c5395cd25b371617981b341616dbdd31e85 Georgi Gerganov ggerganov@gmail.com 2024-02-18T19:38:06+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-18T19:38:06+02:00 sampling : do not set min_keep to n_probs (#5564)
f3f28c5395cd25b371617981b341616dbdd31e85 e75c6279d1c8e7abb82a331f5de7124eed402de2 Georgi Gerganov ggerganov@gmail.com 2024-02-18T19:17:00+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-18T19:17:00+02:00 cmake : fix GGML_USE_SYCL typo (#5555)
e75c6279d1c8e7abb82a331f5de7124eed402de2 36376abe05a12a8cb3af548a4af9b8d0e2e69597 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-18T17:31:28+01:00 GitHub noreply@github.com 2024-02-18T18:31:28+02:00 server : enhanced health endpoint (#5548)
36376abe05a12a8cb3af548a4af9b8d0e2e69597 66c1968f7a2e895675425e875b6589f1233a1b52 Pierrick Hymbert pierrick.hymbert@gmail.com 2024-02-18T17:30:09+01:00 GitHub noreply@github.com 2024-02-18T18:30:09+02:00 server : --n-predict option document and cap to max value (#5549)
66c1968f7a2e895675425e875b6589f1233a1b52 1dcc3fde004787e6fc4d84c9de0bb34cd2901a3e Daniel Hiltgen dhiltgen@users.noreply.github.com 2024-02-18T08:23:16-08:00 GitHub noreply@github.com 2024-02-18T18:23:16+02:00 server : graceful server shutdown (#5244)
1dcc3fde004787e6fc4d84c9de0bb34cd2901a3e 5d3de51f972055702a1859186fe7acb8f0b43dc4 Georgi Gerganov ggerganov@gmail.com 2024-02-18T18:21:52+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-18T18:21:52+02:00 common : fix ub (#5530)
5d3de51f972055702a1859186fe7acb8f0b43dc4 fc0c8d286a533363a9a663510b62af85ffad58b3 Herman Semenov GermanAizek@yandex.ru 2024-02-18T16:20:12Z GitHub noreply@github.com 2024-02-18T18:20:12+02:00 ggml, common, examples, tests : fixed type arguments in printf (#5528)
fc0c8d286a533363a9a663510b62af85ffad58b3 bd2d4e393b2b7d2a1b2e201058e26017c9728ead Daniel Bevenius daniel.bevenius@gmail.com 2024-02-18T17:19:23+01:00 GitHub noreply@github.com 2024-02-18T18:19:23+02:00 llava : update surgery script to not remove tensors (#5536)
bd2d4e393b2b7d2a1b2e201058e26017c9728ead c8e0d7efeb7634ecc2e9832e879ab9fca4510e71 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-18T18:16:55+02:00 GitHub noreply@github.com 2024-02-18T18:16:55+02:00 1.5 bit quantization (#5453)
c8e0d7efeb7634ecc2e9832e879ab9fca4510e71 8f1be0d42f23016cb6819dbae01126699c4bd9bc github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-02-18T00:17:07Z Philip Taron philip.taron@gmail.com 2024-02-18T06:39:58-08:00 flake.lock: Update
8f1be0d42f23016cb6819dbae01126699c4bd9bc 6e4e973b2615f8d390b1c4f4a7e05a119078bb0f Georgi Gerganov ggerganov@gmail.com 2024-02-17T23:04:16+02:00 GitHub noreply@github.com 2024-02-17T23:04:16+02:00 ggml : add ALiBi support for ggml_soft_max_ext (#5488)
6e4e973b2615f8d390b1c4f4a7e05a119078bb0f d250c9d61d4d9f7346930814cc4aef3f3673dc3e Ananta Bastola anantarajbastola@gmail.com 2024-02-17T16:03:14-05:00 GitHub noreply@github.com 2024-02-17T23:03:14+02:00 ci : add an option to fail on compile warning (#3952)
d250c9d61d4d9f7346930814cc4aef3f3673dc3e 5bf2b94dd4fb74378b78604023b31512fec55f8f clibdev 52199778+clibdev@users.noreply.github.com 2024-02-17T18:28:37+02:00 GitHub noreply@github.com 2024-02-17T18:28:37+02:00 gitignore : update for CLion IDE (#5544)
5bf2b94dd4fb74378b78604023b31512fec55f8f d2819d5577b35507be83d0c3f4d2d3c0ab1488ca Georgi Gerganov ggerganov@gmail.com 2024-02-16T19:05:56+02:00 GitHub noreply@github.com 2024-02-16T19:05:56+02:00 cmake : fix VULKAN and ROCm builds (#5525)
d2819d5577b35507be83d0c3f4d2d3c0ab1488ca 4cb072769804c77ab466bc8351c76ede9d5ba49d Georgi Gerganov ggerganov@gmail.com 2024-02-16T15:14:40+02:00 GitHub noreply@github.com 2024-02-16T15:14:40+02:00 scripts : add helpers script for bench comparing commits (#5521)
4cb072769804c77ab466bc8351c76ede9d5ba49d 65085c713e14f78cdda6abc275b1a5d8c2b8ca15 Herman Semenov GermanAizek@yandex.ru 2024-02-16T12:43:23Z GitHub noreply@github.com 2024-02-16T14:43:23+02:00 llava : removed excess free(NULL) operation (#5531)
65085c713e14f78cdda6abc275b1a5d8c2b8ca15 6dcc02d2444c779c18d49c364c5d5c5728b6b484 Herman Semenov GermanAizek@yandex.ru 2024-02-16T11:45:48Z GitHub noreply@github.com 2024-02-16T13:45:48+02:00 llama : minor fixed return int value (#5529)
6dcc02d2444c779c18d49c364c5d5c5728b6b484 5f5808ca7b7f23a1fa7a77241842bb84a0e55108 Alexey Parfenov zxed@alkatrazstudio.net 2024-02-16T11:33:25Z GitHub noreply@github.com 2024-02-16T13:33:25+02:00 server : add "samplers" param to control the samplers order (#5494)
5f5808ca7b7f23a1fa7a77241842bb84a0e55108 f486f6e1e5e9d01603d9325ab3e05f1edb362a95 Rőczey Barnabás 31726601+An0nie@users.noreply.github.com 2024-02-16T11:00:56+01:00 GitHub noreply@github.com 2024-02-16T12:00:56+02:00 server : fix system prompt cli (#5516)
f486f6e1e5e9d01603d9325ab3e05f1edb362a95 60ed04cf82dc91ade725dd7ad53f0ee81f76eccf bmwl brian.marshall@tolko.com 2024-02-16T01:31:07-08:00 GitHub noreply@github.com 2024-02-16T11:31:07+02:00 ggml : add numa options (#5377)
60ed04cf82dc91ade725dd7ad53f0ee81f76eccf 594845aab1c6775877f6d9545a51dc0f8d0b3d77 Daniel Bevenius daniel.bevenius@gmail.com 2024-02-16T10:24:39+01:00 GitHub noreply@github.com 2024-02-16T11:24:39+02:00 llava : fix clip-model-is-vision flag in README.md (#5509)
594845aab1c6775877f6d9545a51dc0f8d0b3d77 4524290e87b8e107cc2b56e1251751546f4b9051 Georgi Gerganov ggerganov@gmail.com 2024-02-16T09:57:55+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-16T09:57:55+02:00 ci : fix BERT model download and convert
4524290e87b8e107cc2b56e1251751546f4b9051 c06e45d72983d9ace7b1535f7e7ea258d212169e Douglas Hanley thesecretaryofwar@gmail.com 2024-02-15T11:21:49-06:00 GitHub noreply@github.com 2024-02-15T12:21:49-05:00 Use correct type of pooling for embedding models (#5500)
c06e45d72983d9ace7b1535f7e7ea258d212169e 9060a1e9dfca6038906e819be5fa42217f49028c Georgi Gerganov ggerganov@gmail.com 2024-02-15T18:49:08+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-15T18:49:08+02:00 clip : fix wrong loop condition
9060a1e9dfca6038906e819be5fa42217f49028c 9350a1cf21b1492c69b20175b73a419b897d6a3a slaren slarengh@gmail.com 2024-02-15T16:49:01+01:00 GitHub noreply@github.com 2024-02-15T16:49:01+01:00 cuda : print message when initialization fails (#5512)
9350a1cf21b1492c69b20175b73a419b897d6a3a 73122473ffd73030146276dbb85da7c8021a3ee4 Georgi Gerganov ggerganov@gmail.com 2024-02-15T15:41:15+02:00 GitHub noreply@github.com 2024-02-15T15:41:15+02:00 scripts : add hf.sh helper script (#5501)
73122473ffd73030146276dbb85da7c8021a3ee4 0d4177126b0556e202efb85bf3f768be81076400 Michaël de Vries vriesdemichael@gmail.com 2024-02-15T14:14:37+01:00 GitHub noreply@github.com 2024-02-15T14:14:37+01:00 fix(gguf-py): special tokens are no longer skipped when add_<token>_token is set to false (#5487)
0d4177126b0556e202efb85bf3f768be81076400 7930a8a6e89a04c77c51e3ae5dc1cd8e845b6b8f Elbios 141279586+Elbios@users.noreply.github.com 2024-02-15T09:01:57+01:00 GitHub noreply@github.com 2024-02-15T10:01:57+02:00 llava : fix memory management bug (#5491)
7930a8a6e89a04c77c51e3ae5dc1cd8e845b6b8f 704359e29985a06a389337a2617b7f3fa8eff908 John 78893154+cmp-nct@users.noreply.github.com 2024-02-15T08:59:18+01:00 GitHub noreply@github.com 2024-02-15T09:59:18+02:00 llaba : hotfix for llava-1.6 image number (#5495)
704359e29985a06a389337a2617b7f3fa8eff908 594fca3fefe27b8e95cfb1656eb0e160ad15a793 Neuman Vong neuman.vong@gmail.com 2024-02-15T17:11:15+11:00 GitHub noreply@github.com 2024-02-15T07:11:15+01:00 vulkan: Find optimal memory type but with fallback (#5381)
594fca3fefe27b8e95cfb1656eb0e160ad15a793 ccbb277f4642fc0d84c72dbc0d51ed2df418d6ce Rune 43761327+Rune-AI@users.noreply.github.com 2024-02-14T16:15:49+01:00 GitHub noreply@github.com 2024-02-14T17:15:49+02:00 readme : fix typo (#5490)
ccbb277f4642fc0d84c72dbc0d51ed2df418d6ce 8084d554406b767d36b3250b3b787462d5dd626f John 78893154+cmp-nct@users.noreply.github.com 2024-02-14T15:49:42+01:00 GitHub noreply@github.com 2024-02-14T16:49:42+02:00 llava : update README.md (#5489)
8084d554406b767d36b3250b3b787462d5dd626f aa2341298924ac89778252015efcb792f2df1e20 Michael Podvitskiy podvitskiymichael@gmail.com 2024-02-14T11:49:01+03:00 GitHub noreply@github.com 2024-02-14T10:49:01+02:00 cmake : ARM intrinsics detection for MSVC (#5401)
aa2341298924ac89778252015efcb792f2df1e20 f5ca054855dea83f424003162f26de376e5643f6 John 78893154+cmp-nct@users.noreply.github.com 2024-02-14T08:38:35+01:00 GitHub noreply@github.com 2024-02-14T09:38:35+02:00 llava : support v1.6 (#5267)
f5ca054855dea83f424003162f26de376e5643f6 6c00a066928b0475b865a2e3e709e2166e02d548 AT manyoso@users.noreply.github.com 2024-02-13T15:44:25-06:00 GitHub noreply@github.com 2024-02-13T22:44:25+01:00 Early return for zero size calls to get_tensor. (#5482)
6c00a066928b0475b865a2e3e709e2166e02d548 ea9c8e11436ad50719987fa23a289c74b7b40d40 John 78893154+cmp-nct@users.noreply.github.com 2024-02-13T18:56:38+01:00 GitHub noreply@github.com 2024-02-13T19:56:38+02:00 gguf : add python reader example (#5216)
ea9c8e11436ad50719987fa23a289c74b7b40d40 c4e6dd59e45ef7b14f7763fb073b517395dc176c Jared Van Bortel jared@nomic.ai 2024-02-13T12:03:53-05:00 GitHub noreply@github.com 2024-02-13T12:03:53-05:00 llama : add support for Nomic Embed (#5468)
c4e6dd59e45ef7b14f7763fb073b517395dc176c 037259be689353081e7bae3c1ab4ab18e7fbe8c9 Aarni Koskela akx@iki.fi 2024-02-13T18:18:16+02:00 GitHub noreply@github.com 2024-02-13T18:18:16+02:00 llama : allow raw byte in SPM vocabs; don't crash on nl 404 (#5478)
037259be689353081e7bae3c1ab4ab18e7fbe8c9 263978904c7472db1865409a7ff1129599f6a40b Aarni Koskela akx@iki.fi 2024-02-13T15:24:50+02:00 GitHub noreply@github.com 2024-02-13T15:24:50+02:00 llama : make load error reporting more granular (#5477)
263978904c7472db1865409a7ff1129599f6a40b cf45252a7cfcb998bade46a886e20477cecc538a Daniel Bevenius daniel.bevenius@gmail.com 2024-02-13T14:15:42+01:00 GitHub noreply@github.com 2024-02-13T15:15:42+02:00 finetune : rename feed-forward tensors (w1/w2/w3) (#4839)
cf45252a7cfcb998bade46a886e20477cecc538a 03bf161eb6dea6400ee49c6dc6b69bdcfa9fd3fc Georgi Gerganov ggerganov@gmail.com 2024-02-13T15:14:22+02:00 GitHub noreply@github.com 2024-02-13T15:14:22+02:00 tests : multi-thread the tokenizer tests (#5474)
03bf161eb6dea6400ee49c6dc6b69bdcfa9fd3fc ad014bba97ef6ef6c3e2f78b2fc463e91ae94579 Douglas Hanley thesecretaryofwar@gmail.com 2024-02-13T06:06:58-06:00 GitHub noreply@github.com 2024-02-13T14:06:58+02:00 llama : support batched embeddings (#5466)
ad014bba97ef6ef6c3e2f78b2fc463e91ae94579 49cc1f7d67de2da99f3ac185f9ff1319b7bf35f8 Johannes Gäßler johannesg@5d6.de 2024-02-13T12:38:37+01:00 GitHub noreply@github.com 2024-02-13T12:38:37+01:00 make: add error message for bad CUDA version (#5444)
49cc1f7d67de2da99f3ac185f9ff1319b7bf35f8 99b8b43d7b185a6483f28cf798a2d968b2e16ca7 Georgi Gerganov ggerganov@gmail.com 2024-02-13T13:01:29+02:00 GitHub noreply@github.com 2024-02-13T13:01:29+02:00 bert : add tests + fix quantization (#5475)
99b8b43d7b185a6483f28cf798a2d968b2e16ca7 895407f31b358e3d9335e847d13f033491ec8a5b Georgi Gerganov ggerganov@gmail.com 2024-02-13T11:20:24+02:00 GitHub noreply@github.com 2024-02-13T11:20:24+02:00 tests : disable moe test (#5473)
895407f31b358e3d9335e847d13f033491ec8a5b 099afc6274c859ca67146e725839f2d97a5ef313 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-13T09:07:57+02:00 GitHub noreply@github.com 2024-02-13T09:07:57+02:00 ggml-quants : fix compiler warnings (shadow variable) (#5472)
099afc6274c859ca67146e725839f2d97a5ef313 df334a11251b81fd0b6a0e51e7146e0ba9e973f2 Georgi Gerganov ggerganov@gmail.com 2024-02-12T20:14:39+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-12T20:14:39+02:00 llama : fix quantization when tensors are missing (#5423)
df334a11251b81fd0b6a0e51e7146e0ba9e973f2 dbd8828eb03b9aa8d0af7e4c533d3c2f5b38aba6 Georgi Gerganov ggerganov@gmail.com 2024-02-12T19:54:29+02:00 GitHub noreply@github.com 2024-02-12T19:54:29+02:00 swift : package no longer use ggml dependency (#5465)
dbd8828eb03b9aa8d0af7e4c533d3c2f5b38aba6 43fe07c1a4f3a58612e1d9543f7c6b556710f5d0 Lee 44310445+lx200916@users.noreply.github.com 2024-02-13T01:29:57+08:00 GitHub noreply@github.com 2024-02-12T19:29:57+02:00 py : fix persimmon `n_rot` conversion (#5460)
43fe07c1a4f3a58612e1d9543f7c6b556710f5d0 4a46d2b7923be83d6019251671ee63aa1fa0d6bc Abhilash Majumder 30946547+abhilash1910@users.noreply.github.com 2024-02-12T20:22:05+05:30 GitHub noreply@github.com 2024-02-12T20:22:05+05:30 ggml-sycl: Replace 3d ops with macro (#5458)
4a46d2b7923be83d6019251671ee63aa1fa0d6bc 3b169441dfe8e420f88d1592708cc2a871daadb9 Daniel Bevenius daniel.bevenius@gmail.com 2024-02-12T09:38:44+01:00 GitHub noreply@github.com 2024-02-12T10:38:44+02:00 llava : remove prog parameter from ArgumentParser (#5457)
3b169441dfe8e420f88d1592708cc2a871daadb9 3bdc4cd0f595a6096cca4a64aa75ffa8a3503465 Georgi Gerganov ggerganov@gmail.com 2024-02-12T09:16:06+02:00 GitHub noreply@github.com 2024-02-12T09:16:06+02:00 sync : ggml (#5452)
3bdc4cd0f595a6096cca4a64aa75ffa8a3503465 2891c8aa9af17f4ff636ff3868bc34ff72b56e25 Johannes Gäßler johannesg@5d6.de 2024-02-11T19:08:39+01:00 GitHub noreply@github.com 2024-02-11T19:08:39+01:00 CUDA: mul_mat_vec_q tiling, refactor mul mat logic (#5434)
2891c8aa9af17f4ff636ff3868bc34ff72b56e25 97a336507ed9b971d72262bec7e2b8b7016a054a Douglas Hanley thesecretaryofwar@gmail.com 2024-02-11T10:21:38-06:00 GitHub noreply@github.com 2024-02-11T11:21:38-05:00 Add support for BERT embedding models (#5423)
97a336507ed9b971d72262bec7e2b8b7016a054a c88c74f967028ae3d5ebade40ae586d20a961abc github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-02-11T00:17:31Z Philip Taron philip.taron@gmail.com 2024-02-11T07:50:41-08:00 flake.lock: Update
c88c74f967028ae3d5ebade40ae586d20a961abc a803333a4e6fc534c93afe90d741bc2388bdec87 Sergio López slp@sinrega.org 2024-02-11T15:12:00+01:00 GitHub noreply@github.com 2024-02-11T15:12:00+01:00 vulkan: only use M-sized matmul on Apple GPUs (#5412)
a803333a4e6fc534c93afe90d741bc2388bdec87 684780141a08200ec98eba3e982dbafd1d0b5000 Alexey Parfenov zxed@alkatrazstudio.net 2024-02-11T13:43:31Z GitHub noreply@github.com 2024-02-11T15:43:31+02:00 common : use enums for sampler types (#5418)
684780141a08200ec98eba3e982dbafd1d0b5000 85910c5b30f6e268321be8df044f5528a6efac52 Alexey Parfenov zxed@alkatrazstudio.net 2024-02-11T13:38:14Z GitHub noreply@github.com 2024-02-11T15:38:14+02:00 server : allow to specify tokens as strings in logit_bias (#5003)
85910c5b30f6e268321be8df044f5528a6efac52 139b62a839825ef20084ed75ed624db7a5ad554a Georgi Gerganov ggerganov@gmail.com 2024-02-11T15:35:50+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-11T15:35:50+02:00 main : ctrl+C print timing in non-interactive mode (#3873)
139b62a839825ef20084ed75ed624db7a5ad554a 0f2411f154db46780d3aaa3a0664691b2170c83f Georgi Gerganov ggerganov@gmail.com 2024-02-11T15:33:43+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-11T15:33:43+02:00 common : fix compile warning
0f2411f154db46780d3aaa3a0664691b2170c83f a07d0fee1f05c5c1dc49948ae1a3293db017275f Georgi Gerganov ggerganov@gmail.com 2024-02-11T15:33:01+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-11T15:33:01+02:00 ggml : fix compile warnings (unused vars) (#4966)
a07d0fee1f05c5c1dc49948ae1a3293db017275f e4640d8fdf56f14a6db3d092bcd3d2d315cb5d04 snadampal 87143774+snadampal@users.noreply.github.com 2024-02-11T07:22:33-06:00 GitHub noreply@github.com 2024-02-11T15:22:33+02:00 ggml : add mmla kernels for quantized GEMM (#4966)
e4640d8fdf56f14a6db3d092bcd3d2d315cb5d04 907e08c1109f498b01036367804cff3082c44524 Johannes Gäßler johannesg@5d6.de 2024-02-11T12:44:51+01:00 GitHub noreply@github.com 2024-02-11T12:44:51+01:00 lookup: add print for drafting performance (#5450)
907e08c1109f498b01036367804cff3082c44524 f026f8120f97090d34a52b3dc023c82e0ede3f7d Xuan Son Nguyen thichthat@gmail.com 2024-02-11T11:16:22+01:00 GitHub noreply@github.com 2024-02-11T12:16:22+02:00 server : add llama2 chat template (#5425)
f026f8120f97090d34a52b3dc023c82e0ede3f7d cd9aea63b577a83def84dbd6dcd90a6fa02af745 Ian Bull irbull@eclipsesource.com 2024-02-10T02:53:28-08:00 GitHub noreply@github.com 2024-02-10T12:53:28+02:00 metal : use autoreleasepool to avoid memory leaks (#5437)
cd9aea63b577a83def84dbd6dcd90a6fa02af745 43b65f5eb85e8741aba573a8f65bb8efad245d31 Georgi Gerganov ggerganov@gmail.com 2024-02-10T09:53:05+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-10T09:53:05+02:00 scripts : update sync scripts with new backends
43b65f5eb85e8741aba573a8f65bb8efad245d31 4633d93af08d890ecd00fa6e4f61d76f21cded4c Georgi Gerganov ggerganov@gmail.com 2024-02-10T09:30:36+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-10T09:30:36+02:00 sync : ggml
4633d93af08d890ecd00fa6e4f61d76f21cded4c 4b7b38bef5addbd31f453871d79647fbae6bec8a Michael Podvitskiy podvitskiymichael@gmail.com 2024-02-09T10:42:27+01:00 Georgi Gerganov ggerganov@gmail.com 2024-02-10T09:29:21+02:00 ggml : add abort_callback for cpu backend (ggml/725)
4b7b38bef5addbd31f453871d79647fbae6bec8a e00d2a62dd1441e3b089570ec06d05c18800d368 Neuman Vong neuman.vong@gmail.com 2024-02-10T05:30:19+11:00 GitHub noreply@github.com 2024-02-09T19:30:19+01:00 vulkan: Set limit for task concurrency (#5427)
e00d2a62dd1441e3b089570ec06d05c18800d368 7c777fcd5dd4af7079e33390cf6a19c328a2666f Daniel Bevenius daniel.bevenius@gmail.com 2024-02-09T14:00:59+01:00 GitHub noreply@github.com 2024-02-09T15:00:59+02:00 llava : add requirements.txt and update README.md (#5428)
7c777fcd5dd4af7079e33390cf6a19c328a2666f e5ca3937c685d6e012ac4db40555d6ec100ff03c Riley Stewart ristew@users.noreply.github.com 2024-02-09T02:49:49-08:00 GitHub noreply@github.com 2024-02-09T12:49:49+02:00 server : fix prompt caching for repeated prompts (#5420)
e5ca3937c685d6e012ac4db40555d6ec100ff03c e4124c24775f2cb5b3d7acc93bf9dc5471c172ef Paul Tsochantaris ptsochantaris@icloud.com 2024-02-09T10:48:06Z GitHub noreply@github.com 2024-02-09T12:48:06+02:00 llama : do not cap thread count when MoE on CPU (#5419)
e4124c24775f2cb5b3d7acc93bf9dc5471c172ef b2f87cb64db47d799b6f3656855c9caf9792ab2a Marko Tasic mtasic85@gmail.com 2024-02-09T11:17:00+01:00 GitHub noreply@github.com 2024-02-09T12:17:00+02:00 readme : add JavaScript/Wasm repo (#5415)
b2f87cb64db47d799b6f3656855c9caf9792ab2a 44fbe34360dd760f9e68b4271f21533436397f84 Michael Podvitskiy podvitskiymichael@gmail.com 2024-02-09T10:56:43+01:00 GitHub noreply@github.com 2024-02-09T11:56:43+02:00 ggml : fix `error C2078: too many initializers` for MSVC ARM64 (#5404)
44fbe34360dd760f9e68b4271f21533436397f84 8e6a9d2de0096af7120606c74ee2f26684e87b41 0cc4m picard12@live.de 2024-02-09T06:52:33+01:00 GitHub noreply@github.com 2024-02-09T06:52:33+01:00 Fix Vulkan crash on APUs with very little device memory (#5424)
8e6a9d2de0096af7120606c74ee2f26684e87b41 41f308f58edc2a04bcf9e245100b0a9b10e9a0fb Johannes Gäßler johannesg@5d6.de 2024-02-08T21:56:40+01:00 GitHub noreply@github.com 2024-02-08T21:56:40+01:00 CUDA: more warps for mmvq on NVIDIA (#5394)
41f308f58edc2a04bcf9e245100b0a9b10e9a0fb 6e99f2a04f1871d637dd77eb4d81de31a5510253 slaren slarengh@gmail.com 2024-02-08T21:33:03+01:00 GitHub noreply@github.com 2024-02-08T21:33:03+01:00 llama : do not print "offloading layers" message in CPU-only builds (#5416)
6e99f2a04f1871d637dd77eb4d81de31a5510253 ff4ff05c5ff4311c05a8ce1f984c7d8def4f07a5 Abhilash Majumder 30946547+abhilash1910@users.noreply.github.com 2024-02-08T22:39:10+05:30 GitHub noreply@github.com 2024-02-08T22:39:10+05:30 Fix f16_sycl cpy call from Arc (#5411)
ff4ff05c5ff4311c05a8ce1f984c7d8def4f07a5 b7b74cef36a93ae01e0b9af8986d131761742d0e Daniel Bevenius daniel.bevenius@gmail.com 2024-02-08T15:20:03+01:00 GitHub noreply@github.com 2024-02-08T16:20:03+02:00 llava : add missing .py, and fix paths in README.md (#5414)
b7b74cef36a93ae01e0b9af8986d131761742d0e 4aa43fab569215a13495a7f1a0f8afc541b16d03 Johannes Gäßler johannesg@5d6.de 2024-02-08T11:36:54+01:00 GitHub noreply@github.com 2024-02-08T11:36:54+01:00 fix trailing whitespace (#5407)
4aa43fab569215a13495a7f1a0f8afc541b16d03 a6e514a85f0fda38ff78ec91782877ea3d19ed98 runfuture runfuture@users.noreply.github.com 2024-02-08T18:36:19+08:00 GitHub noreply@github.com 2024-02-08T12:36:19+02:00 llama : fix MiniCPM (#5392)
a6e514a85f0fda38ff78ec91782877ea3d19ed98 26d4efd11e48908e14e2ee9471a7fc4c57079a1d Daniel Bevenius daniel.bevenius@gmail.com 2024-02-08T09:58:19+01:00 GitHub noreply@github.com 2024-02-08T09:58:19+01:00 llava: fix typo/formatting in README.md (#5405)
26d4efd11e48908e14e2ee9471a7fc4c57079a1d 8504d2d0da8cc7a1f2eee0e9e56949f960510b75 Johannes Gäßler johannesg@5d6.de 2024-02-08T09:46:30+01:00 GitHub noreply@github.com 2024-02-08T09:46:30+01:00 sampling: fix top_k <= 0 (#5388)
8504d2d0da8cc7a1f2eee0e9e56949f960510b75 c4fbb6717c684196bd13b72d21747557130914e8 Georgi Gerganov ggerganov@gmail.com 2024-02-08T09:46:47+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-08T09:46:47+02:00 tests : .gitignore obj files
c4fbb6717c684196bd13b72d21747557130914e8 8c933b70c21e05b685d476d0a1f36b34cbda7365 Michael Podvitskiy podvitskiymichael@gmail.com 2024-02-07T22:39:23+01:00 GitHub noreply@github.com 2024-02-07T16:39:23-05:00 CMAKE_OSX_ARCHITECTURES for MacOS cross compilation (#5393)
8c933b70c21e05b685d476d0a1f36b34cbda7365 b906596bb775b17656c2e51d5ab1b347faab6860 Ebey Abraham ebey97@gmail.com 2024-02-07T21:11:30Z GitHub noreply@github.com 2024-02-07T22:11:30+01:00 fix typo in readme (#5399)
b906596bb775b17656c2e51d5ab1b347faab6860 aa7ab99be29b633263803f2e185265734c2d9427 Kamil Tomšík info@tomsik.cz 2024-02-07T19:44:52+01:00 GitHub noreply@github.com 2024-02-07T13:44:52-05:00 Add Ava in the list of llama.cpp UIs (#4362)
aa7ab99be29b633263803f2e185265734c2d9427 10afa6f1d11ebc9fcc1085f468170002cbf6e2b5 Johannes Gäßler johannesg@5d6.de 2024-02-07T12:40:26+01:00 GitHub noreply@github.com 2024-02-07T12:40:26+01:00 CUDA: fixed mmvq kernel for bs 2,3,4 and -sm row (#5386)
10afa6f1d11ebc9fcc1085f468170002cbf6e2b5 0ef46da632c32faa1a538e5dc180994e8bbb46e1 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-02-07T18:16:55+08:00 GitHub noreply@github.com 2024-02-07T18:16:55+08:00 [SYCL] update install make by w64devkit (#5297)
0ef46da632c32faa1a538e5dc180994e8bbb46e1 ee1628bdfea8b0079fed0140ac2f00ef1b465b57 Xiao-Yong Jin jinxiaoyong@gmail.com 2024-02-07T02:17:25-06:00 GitHub noreply@github.com 2024-02-07T10:17:25+02:00 llava-cli : always tokenize special tokens (#5382)
ee1628bdfea8b0079fed0140ac2f00ef1b465b57 ed0bf32290ee5b30ffad5becd99cbecef74aedd7 0cc4m picard12@live.de 2024-02-07T07:54:50+01:00 GitHub noreply@github.com 2024-02-07T07:54:50+01:00 Basic Vulkan Multi-GPU implementation (#5321)
ed0bf32290ee5b30ffad5becd99cbecef74aedd7 9a697d842bc0cfce8268ebd2ba703ffc1c904f98 Eve 139727413+netrunnereve@users.noreply.github.com 2024-02-07T06:21:30Z GitHub noreply@github.com 2024-02-07T08:21:30+02:00 readme : modernize (#5379)
9a697d842bc0cfce8268ebd2ba703ffc1c904f98 316c7faf7740fa98ea68f1445f4505810f706b9e Ben Williams ben@719ben.com 2024-02-06T22:16:48-08:00 GitHub noreply@github.com 2024-02-07T08:16:48+02:00 readme : update ui list (#5354)
316c7faf7740fa98ea68f1445f4505810f706b9e f3e2b4fa3f81a410ecb7dec929c259ef8d8dbb7d runfuture runfuture@users.noreply.github.com 2024-02-07T14:15:56+08:00 GitHub noreply@github.com 2024-02-07T08:15:56+02:00 llama : add MiniCPM support (#5346)
f3e2b4fa3f81a410ecb7dec929c259ef8d8dbb7d f68664ac241a6b5c233d8f1051eef20929b06008 Justin Parker jparkerweb@gmail.com 2024-02-07T01:15:19-05:00 GitHub noreply@github.com 2024-02-07T08:15:19+02:00 server : update `/props` with "total_slots" value (#5373)
f68664ac241a6b5c233d8f1051eef20929b06008 213d1439fadefe182f69c5f7e8dd3b4b6572ebcb Sang-Kil Park sang.park@42dot.ai 2024-02-07T13:28:00+09:00 GitHub noreply@github.com 2024-02-06T23:28:00-05:00 convert : fix TypeError on GPT-2 vocab.json (#5288)
213d1439fadefe182f69c5f7e8dd3b4b6572ebcb 17c97fb0620448b37516a3f53fea6c482b0a30a4 Alexey Parfenov zxed@alkatrazstudio.net 2024-02-06T18:08:38Z GitHub noreply@github.com 2024-02-06T20:08:38+02:00 server : remove model.json endpoint (#5371)
17c97fb0620448b37516a3f53fea6c482b0a30a4 b08f22c882a1443e6b97081f3ce718a4d1a741f8 Johannes Gäßler johannesg@5d6.de 2024-02-06T18:43:06+01:00 GitHub noreply@github.com 2024-02-06T19:43:06+02:00 CUDA: mul_mat_vec_q max. batch size 8 -> 4 (#5370)
b08f22c882a1443e6b97081f3ce718a4d1a741f8 f57fadc009cbff741a1961cb7896c47d73978d2c Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-06T19:00:16+02:00 GitHub noreply@github.com 2024-02-06T19:00:16+02:00 Update README.md (#5366)
f57fadc009cbff741a1961cb7896c47d73978d2c 2e9c0bd6b301155ce749e162527fc55e9fb5b832 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-06T17:28:02+02:00 GitHub noreply@github.com 2024-02-06T17:28:02+02:00 Slight quantization improvement for Q4_K and Q5_K (#5361)
2e9c0bd6b301155ce749e162527fc55e9fb5b832 2c516611f1d0f1e5e9754f8ea1cf97cb1b17bf2c BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2024-02-06T09:06:48-05:00 GitHub noreply@github.com 2024-02-06T16:06:48+02:00 readme : add phi, orion 14b, internlm2, and yi-VL to readme (#5362)
2c516611f1d0f1e5e9754f8ea1cf97cb1b17bf2c 8a79c591de9b7ff3242a94f68b7fb5a17ed8c2be Johannes Gäßler johannesg@5d6.de 2024-02-06T14:44:06+01:00 GitHub noreply@github.com 2024-02-06T14:44:06+01:00 CUDA: mul_mat_vec_q for batch sizes > 1 (#5351)
8a79c591de9b7ff3242a94f68b7fb5a17ed8c2be 31e790322133a4b1d0684527ea446e765e8a96cf Justin Parker jparkerweb@gmail.com 2024-02-06T04:20:59-05:00 GitHub noreply@github.com 2024-02-06T11:20:59+02:00 server : include total "num_slots" in props endpoint (#5349)
31e790322133a4b1d0684527ea446e765e8a96cf 4ffc7a17d4e80c5f3f905139cb570ed9b6934fcb Michael Coppola m18coppola@gmail.com 2024-02-06T04:20:00-05:00 GitHub noreply@github.com 2024-02-06T11:20:00+02:00 server : add `dynatemp_range` and `dynatemp_exponent` (#5352)
4ffc7a17d4e80c5f3f905139cb570ed9b6934fcb 906cff55c2848fda091d888a1585915ec0c9ea9e Niall Coates 1349685+Niall-@users.noreply.github.com 2024-02-06T08:16:23Z GitHub noreply@github.com 2024-02-06T10:16:23+02:00 server : various fixes for the prompt field in /completion (#5300)
906cff55c2848fda091d888a1585915ec0c9ea9e 098f6d737b65134cf220d12b9b706e8cfc5e4610 Georgi Gerganov ggerganov@gmail.com 2024-02-06T07:47:22+02:00 GitHub noreply@github.com 2024-02-06T07:47:22+02:00 py : handle byte tokens in `get_token_type` (#5341)
098f6d737b65134cf220d12b9b706e8cfc5e4610 78b00dda6c0d62c34f5371d47718defff6ed2b22 Johannes Gäßler johannesg@5d6.de 2024-02-05T19:33:00+01:00 GitHub noreply@github.com 2024-02-05T19:33:00+01:00 make: Use ccache for faster compilation (#5318)
78b00dda6c0d62c34f5371d47718defff6ed2b22 c6b395535a6874d749ef47c33eacd466cb252cd5 Johannes Gäßler johannesg@5d6.de 2024-02-05T15:55:10+01:00 GitHub noreply@github.com 2024-02-05T15:55:10+01:00 README: updated introduction (#5343)
c6b395535a6874d749ef47c33eacd466cb252cd5 abb61944a5f64dec62c893ed0db10790169b672a Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-05T14:09:47+02:00 GitHub noreply@github.com 2024-02-05T14:09:47+02:00 ggml : make use of ggml-quants.h possible in C++ code (#5338)
abb61944a5f64dec62c893ed0db10790169b672a 89503dcb5f764a5cc7093db1f395f5121876a2cc Dr. Tom Murphy VII Ph.D 499244+tom7@users.noreply.github.com 2024-02-05T06:13:57-05:00 GitHub noreply@github.com 2024-02-05T13:13:57+02:00 ggml : avoid duplicating function calls using MIN/MAX macros (#5325)
89503dcb5f764a5cc7093db1f395f5121876a2cc 7e1ae372f36d98fa66b1d778c5862904b4d80c88 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-05T12:32:27+02:00 GitHub noreply@github.com 2024-02-05T12:32:27+02:00 iq3_xxs: quards for the no-imatrix situation (#5334)
7e1ae372f36d98fa66b1d778c5862904b4d80c88 6fdfa2ecc684000a25a4ad91823bc82a6652b645 Guoteng 32697156+SolenoidWGT@users.noreply.github.com 2024-02-05T17:04:06+08:00 GitHub noreply@github.com 2024-02-05T11:04:06+02:00 py : fix internlm2-hf convert to gguf (#5305)
6fdfa2ecc684000a25a4ad91823bc82a6652b645 a2d60c9158435ae9a6f14632f07f1acf7a3becef Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-05T10:46:06+02:00 GitHub noreply@github.com 2024-02-05T10:46:06+02:00 iq2_xxs: tune quantization (#5320)
a2d60c9158435ae9a6f14632f07f1acf7a3becef e6f81775323f6f4e4a30abf022a6028fa86b79ac Alexey Parfenov zxed@alkatrazstudio.net 2024-02-05T08:10:22Z GitHub noreply@github.com 2024-02-05T10:10:22+02:00 server : allow to get default generation settings for completion (#5307)
e6f81775323f6f4e4a30abf022a6028fa86b79ac 30679d438d5225b3aecf5cec6482cbc9f8f87ba5 l3utterfly gc.pthzfoldr@gmail.com 2024-02-05T17:00:47+09:00 GitHub noreply@github.com 2024-02-05T10:00:47+02:00 common : add dynamic temperature parameters to main example cli (#5295)
30679d438d5225b3aecf5cec6482cbc9f8f87ba5 4be04c8965578edc09194fab769b4b922b8444f5 Georgi Gerganov ggerganov@gmail.com 2024-02-05T09:48:03+02:00 Georgi Gerganov ggerganov@gmail.com 2024-02-05T09:48:03+02:00 scripts : fix typos, cleanup (#5303)
4be04c8965578edc09194fab769b4b922b8444f5 5d55b0cd827bb0fcfedfa329a82bd5d6ef2c93ca Нияз Гарифзянов 112617865+garrnizon@users.noreply.github.com 2024-02-05T10:43:57+03:00 GitHub noreply@github.com 2024-02-05T09:43:57+02:00 scripts : add non-interactive server-llm.sh (#5303)
5d55b0cd827bb0fcfedfa329a82bd5d6ef2c93ca 4833ac209da6a427de64f97e8f403dcdc5de6bc3 chiranko 96988916+chiranko@users.noreply.github.com 2024-02-05T15:41:38+08:00 GitHub noreply@github.com 2024-02-05T09:41:38+02:00 readme : add CodeShell models to the supported models list (#5330)
4833ac209da6a427de64f97e8f403dcdc5de6bc3 9392ebd49ea5ae236a55b47cbf6a13247e8a3b8c AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-02-05T07:08:24Z GitHub noreply@github.com 2024-02-05T12:38:24+05:30 [SYCL] Fix cpy with dims of 3 (#5289)
9392ebd49ea5ae236a55b47cbf6a13247e8a3b8c 5ed26e1fc9fab4ce96ecf2d84183fe45bdcab0d4 github-actions[bot] github-actions[bot]@users.noreply.github.com 2024-02-04T00:17:24Z Philip Taron philip.taron@gmail.com 2024-02-04T08:45:35-08:00 flake.lock: Update
5ed26e1fc9fab4ce96ecf2d84183fe45bdcab0d4 277fad30c60ef3559dc2d01b19d05e659d40a824 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-02-04T10:39:58+02:00 GitHub noreply@github.com 2024-02-04T10:39:58+02:00 Adding some imatrix tools (#5302)
277fad30c60ef3559dc2d01b19d05e659d40a824 3c0d25c4756742ebf15ad44700fabc0700c638bd Welby Seely welbyseely@gmail.com 2024-02-03T23:18:51-05:00 GitHub noreply@github.com 2024-02-03T23:18:51-05:00 cmake : use set() for LLAMA_WIN_VER (#5298)
3c0d25c4756742ebf15ad44700fabc0700c638bd 3cc5ed353c07201d8d5b98b0a4713ab633da6d04 Johannes Gäßler johannesg@5d6.de 2024-02-03T20:15:13+01:00 GitHub noreply@github.com 2024-02-03T20:15:13+01:00 make: add nvcc info print (#5310)
3cc5ed353c07201d8d5b98b0a4713ab633da6d04 60ecf099eddfe70fec797ef6790572e452054add Johannes Gäßler johannesg@5d6.de 2024-02-03T20:14:59+01:00 GitHub noreply@github.com 2024-02-03T20:14:59+01:00 make: fix nvcc optimization flags for host code (#5309)
60ecf099eddfe70fec797ef6790572e452054add e920ed393d989ed35625ddaf182ebb52cda07fcd Martin Schwaighofer mschwaig@users.noreply.github.com 2024-01-28T12:59:43+01:00 Philip Taron philip.taron@gmail.com 2024-02-03T13:13:07-06:00 add Vulkan support to Nix flake
e920ed393d989ed35625ddaf182ebb52cda07fcd 52bb63c7082c859c3f1dfc527227e6a95b299c7c 0cc4m picard12@live.de 2024-02-03T18:15:00+01:00 GitHub noreply@github.com 2024-02-03T18:15:00+01:00 Vulkan Intel Fixes, Optimizations and Debugging Flags (#5301)
52bb63c7082c859c3f1dfc527227e6a95b299c7c 1ec3332ade60aeb1494ace2211cf1a966db6d770 Michael Klimenko mklimenko29@gmail.com 2024-02-03T12:23:37+01:00 GitHub noreply@github.com 2024-02-03T13:23:37+02:00 refactor : switch to emplace_back to avoid extra object (#5291)
1ec3332ade60aeb1494ace2211cf1a966db6d770 6a66c5071a74a96c4f52cf1015a092acd18c3714 Jared Van Bortel jared@nomic.ai 2024-02-03T06:22:06-05:00 GitHub noreply@github.com 2024-02-03T13:22:06+02:00 YaRN : store rope scaling type as int32_t in memory (#5285)
6a66c5071a74a96c4f52cf1015a092acd18c3714 a305dba8ff642e57f538f42010868fe0bc5262a1 BADR contact@pythops.com 2024-02-03T12:20:26+01:00 GitHub noreply@github.com 2024-02-03T13:20:26+02:00 readme : add tenere in the ui tools list (#5284)
a305dba8ff642e57f538f42010868fe0bc5262a1 191221178f51b6e81122c5bda0fd79620e547d07 AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-02-03T08:11:37Z GitHub noreply@github.com 2024-02-03T16:11:37+08:00 Fix im2col with 32fp (#5286)
191221178f51b6e81122c5bda0fd79620e547d07 e437b37fd0b2b97e6c6ff1045ec7f901faa6498a kalomaze 66376113+kalomaze@users.noreply.github.com 2024-02-02T08:15:30-06:00 GitHub noreply@github.com 2024-02-02T16:15:30+02:00 perplexity : fix KL divergence calculations on Windows (#5273)
e437b37fd0b2b97e6c6ff1045ec7f901faa6498a 2d40085c26794e29c434480b9e06738e89e5686f Georgi Gerganov ggerganov@gmail.com 2024-02-02T14:23:40+02:00 GitHub noreply@github.com 2024-02-02T14:23:40+02:00 scripts : parse wtype in server-llm.sh (#5167)
2d40085c26794e29c434480b9e06738e89e5686f b05102fe8cfa9893851c6bf6efd15cdc20b6afa2 Mirror Azure 54669636+MirrorAzure@users.noreply.github.com 2024-02-02T14:39:09+03:00 GitHub noreply@github.com 2024-02-02T13:39:09+02:00 py : add check for '.attn.masked_bias' layers to GPT2model (#5281)
b05102fe8cfa9893851c6bf6efd15cdc20b6afa2 6b91b1e0a92ac2e4e269eec6361ca53a61ced6c6 AidanBeltonS 87009434+AidanBeltonS@users.noreply.github.com 2024-02-02T08:39:48Z GitHub noreply@github.com 2024-02-02T16:39:48+08:00 Tidy ggml-sycl (#5261)
6b91b1e0a92ac2e4e269eec6361ca53a61ced6c6 e805f0fa9951081ce0a86378a7aa52b6f636b82d Xuan Son Nguyen thichthat@gmail.com 2024-02-02T08:56:31+01:00 GitHub noreply@github.com 2024-02-02T09:56:31+02:00 docker : add build for SYCL, Vulkan + update readme (#5228)
e805f0fa9951081ce0a86378a7aa52b6f636b82d af3ba5d94627d337e32a95129e31a3064c459f6b Meng, Hengyu hengyu.meng@intel.com 2024-02-02T15:54:14+08:00 GitHub noreply@github.com 2024-02-02T15:54:14+08:00 [SYCL] get MAX_MEM_ALLOC from device property (#5270)
af3ba5d94627d337e32a95129e31a3064c459f6b e1e721094d8169636d55f68efe37f222cd3f0677 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-02-02T15:53:27+08:00 GitHub noreply@github.com 2024-02-02T15:53:27+08:00 [SYCL] update guide of SYCL backend (#5254)
e1e721094d8169636d55f68efe37f222cd3f0677 128dcbd3c9c4b12f42b560a4430427d7b2828628 Ian Bull irbull@gmail.com 2024-02-01T23:20:13-08:00 GitHub noreply@github.com 2024-02-02T09:20:13+02:00 llama : fix memory leak in llama_batch_free (#5252)
128dcbd3c9c4b12f42b560a4430427d7b2828628 4d0924a8902010d31bd737b6f1f594943d120d0f Neo Zhang Jianyu jianyu.zhang@intel.com 2024-02-02T03:48:53+08:00 GitHub noreply@github.com 2024-02-01T20:48:53+01:00 add --no-mmap in llama-bench (#5257)
4d0924a8902010d31bd737b6f1f594943d120d0f 8ca511cadee2c67f0bd8c7034a2513778ee9a1b7 0cc4m picard12@live.de 2024-02-01T19:25:24+01:00 GitHub noreply@github.com 2024-02-01T19:25:24+01:00 Vulkan Phi Fix for AMD Proprietary Drivers (#5260)
8ca511cadee2c67f0bd8c7034a2513778ee9a1b7 d71ac90985854b0905e1abba778e407e17f9f887 slaren slarengh@gmail.com 2024-02-01T18:30:17+01:00 GitHub noreply@github.com 2024-02-01T18:30:17+01:00 cuda : fix LLAMA_CUDA_F16 (#5262)
d71ac90985854b0905e1abba778e407e17f9f887 ce32060198b7e2d6a13a9b8e1e1369e3c295ae2a Ali Nehzat ali.nehzat@thanks.dev 2024-02-02T02:18:53+11:00 GitHub noreply@github.com 2024-02-01T17:18:53+02:00 make : generate .a library for static linking (#5205)
ce32060198b7e2d6a13a9b8e1e1369e3c295ae2a 1cfb5372cf5707c8ec6dde7c874f4a44a6c4c915 Guoteng 32697156+SolenoidWGT@users.noreply.github.com 2024-02-01T17:19:51+08:00 GitHub noreply@github.com 2024-02-01T11:19:51+02:00 llama : support InternLM2 (#5184)
1cfb5372cf5707c8ec6dde7c874f4a44a6c4c915 d3bac7d58408c602ec1f1e423695f1df8410bb03 Eve 139727413+netrunnereve@users.noreply.github.com 2024-01-31T19:21:55Z GitHub noreply@github.com 2024-01-31T20:21:55+01:00 Fix broken Vulkan Cmake (properly) (#5230)
d3bac7d58408c602ec1f1e423695f1df8410bb03 5cb04dbc16d1da38c8fdcc0111b40e67d00dd1c3 Georgi Gerganov ggerganov@gmail.com 2024-01-31T18:47:10+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-31T18:47:10+02:00 llama : reorder build_orion() at correct place (#5118)
5cb04dbc16d1da38c8fdcc0111b40e67d00dd1c3 efb7bdbbd061d087c788598b97992c653f992ddd Georgi Gerganov ggerganov@gmail.com 2024-01-31T17:30:17+02:00 GitHub noreply@github.com 2024-01-31T17:30:17+02:00 llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240)
efb7bdbbd061d087c788598b97992c653f992ddd 15606309a05ccf7fadbaad5538cb7c32acb1e06b Georgi Gerganov ggerganov@gmail.com 2024-01-31T15:35:41+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-31T15:35:41+02:00 metal : add im2col F32 dst support (#5132)
15606309a05ccf7fadbaad5538cb7c32acb1e06b b2b9f025e7821e78bd501d75d01838c26de07a57 JidongZhang-THU 1119708529@qq.com 2024-01-31T21:10:15+08:00 GitHub noreply@github.com 2024-01-31T15:10:15+02:00 llava : add MobileVLM support (#5132)
b2b9f025e7821e78bd501d75d01838c26de07a57 dabcc5b471348e4ae03ddacc41e19ad75fb2f041 Neo Zhang Jianyu jianyu.zhang@intel.com 2024-01-31T21:04:46+08:00 GitHub noreply@github.com 2024-01-31T18:34:46+05:30 format license text, restore apache license by legal suggestion (#5233)
dabcc5b471348e4ae03ddacc41e19ad75fb2f041 f8e9140cb46eebaa867e1184a9946e4840eec772 slaren slarengh@gmail.com 2024-01-31T13:43:03+01:00 GitHub noreply@github.com 2024-01-31T13:43:03+01:00 ggml : limit n_threads to the max n_tasks (#5238)
f8e9140cb46eebaa867e1184a9946e4840eec772 d62520eb2cc1d7168a30edec6110e1daefbd959f 0cc4m picard12@live.de 2024-01-31T11:44:19+01:00 GitHub noreply@github.com 2024-01-31T11:44:19+01:00 Vulkan Fixes (#5223)
d62520eb2cc1d7168a30edec6110e1daefbd959f 01684139c352561840ae55ec627ab58abc3e06ab Yiming Cui conandiy@vip.qq.com 2024-01-31T11:04:21+08:00 GitHub noreply@github.com 2024-01-30T22:04:21-05:00 Fix typos of IQ2_XXS and IQ3_XXS in llama.cpp (#5231)
01684139c352561840ae55ec627ab58abc3e06ab e8dc55d0065d076d4c20f3c4bfca562701b4edfe Neo Zhang Jianyu jianyu.zhang@intel.com 2024-01-31T10:38:07+08:00 GitHub noreply@github.com 2024-01-31T08:08:07+05:30 support SYCL backend windows build (#5208)
e8dc55d0065d076d4c20f3c4bfca562701b4edfe e0085fdf7c758f0bc2746fc106fb29dd9df959de Jared Van Bortel jared@nomic.ai 2024-01-30T19:04:37-05:00 GitHub noreply@github.com 2024-01-30T19:04:37-05:00 kompute : llama-bench support and ggml_cpu_has_kompute() (#5226)
e0085fdf7c758f0bc2746fc106fb29dd9df959de e6f291d15844398f8326940fe5ad7f2e02b5aa56 Georgi Gerganov ggerganov@gmail.com 2024-01-30T21:19:26+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-30T21:19:26+02:00 Revert "server : change deps.sh xxd files to string literals (#5221)"
e6f291d15844398f8326940fe5ad7f2e02b5aa56 4003be0e5feef320f3707786f22722b73cff9356 Georgi Gerganov ggerganov@gmail.com 2024-01-30T20:17:30+02:00 GitHub noreply@github.com 2024-01-30T20:17:30+02:00 server : fix context shift (#5195)
4003be0e5feef320f3707786f22722b73cff9356 fea4fd4ba7f6b754ac795387b275e1a014a77bde JohnnyB jboero@users.noreply.github.com 2024-01-30T12:15:05-06:00 GitHub noreply@github.com 2024-01-30T20:15:05+02:00 server : change deps.sh xxd files to string literals (#5221)
fea4fd4ba7f6b754ac795387b275e1a014a77bde 8f8ddfcfadc830b936318c3ea9fe2e8e3365aa85 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-30T19:15:28+02:00 GitHub noreply@github.com 2024-01-30T19:15:28+02:00 ggml : fix IQ3_XXS on Metal (#5219)
8f8ddfcfadc830b936318c3ea9fe2e8e3365aa85 6fb50ebbf036ac57a20fe8d8da31731a543582d5 Georgi Gerganov ggerganov@gmail.com 2024-01-30T16:21:57+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-30T16:21:57+02:00 sync : ggml (#0)
6fb50ebbf036ac57a20fe8d8da31731a543582d5 625a699b5456994bc32a8093d53818f60ceda6d1 Georgi Gerganov ggerganov@gmail.com 2024-01-29T21:08:18+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-30T16:20:25+02:00 gguf : fix comparison (ggml/715)
625a699b5456994bc32a8093d53818f60ceda6d1 a4b07c057a553b1ac253051efc3f040351e2eae1 John Balis phobossystems@gmail.com 2024-01-29T06:37:33-06:00 Georgi Gerganov ggerganov@gmail.com 2024-01-30T16:20:25+02:00 `ggml_cuda_cpy` support for 4d tensors and float16->float32 upcasting (ggml/686)
a4b07c057a553b1ac253051efc3f040351e2eae1 549a1e6cd5b39fe0dc3d4ea5515c65f17797a31e Georgi Gerganov ggerganov@gmail.com 2024-01-29T14:00:10+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-30T16:20:25+02:00 gguf : add input validation, prevent integer overflows (ggml/709)
549a1e6cd5b39fe0dc3d4ea5515c65f17797a31e 5f14ee0b0cd06f1c4790e6123df4b38ace637e88 Georgi Gerganov ggerganov@gmail.com 2024-01-29T13:29:46+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-30T16:20:25+02:00 ci : fix yolo URLs + fix metal capture (ggml/712)
5f14ee0b0cd06f1c4790e6123df4b38ace637e88 8e14e3ddb3744566aef7bc0fa734180e47ae6bdf Jack Mousseau jmousseau@users.noreply.github.com 2024-01-29T01:22:23-08:00 Georgi Gerganov ggerganov@gmail.com 2024-01-30T16:20:25+02:00 metal : add debug capture backend function (ggml/694)
8e14e3ddb3744566aef7bc0fa734180e47ae6bdf f4d7e5497485ce6ce0e322533930b7da4657dd2d Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-30T15:15:07+02:00 GitHub noreply@github.com 2024-01-30T15:15:07+02:00 Faster AVX2 dot product for IQ2_XS (#5187)
f4d7e5497485ce6ce0e322533930b7da4657dd2d 2256f36b79a932a478d4dcdf02c1e5a60056e5f3 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-30T15:14:12+02:00 GitHub noreply@github.com 2024-01-30T15:14:12+02:00 SOTA 3-bit quants (#5196)
2256f36b79a932a478d4dcdf02c1e5a60056e5f3 7359016c7c0f65dc30cf79791212b06f15866450 0cc4m picard12@live.de 2024-01-30T13:59:30+01:00 GitHub noreply@github.com 2024-01-30T13:59:30+01:00 Vulkan Windows APU Memory Handling (#5199)
7359016c7c0f65dc30cf79791212b06f15866450 813416991ab0d1caa0d12f93ac4e8a24a2add0a3 Vladimir Malyutin first-leon@yandex.ru 2024-01-30T17:57:07+07:00 GitHub noreply@github.com 2024-01-30T12:57:07+02:00 quantize : fix typo (#5211)
813416991ab0d1caa0d12f93ac4e8a24a2add0a3 5589921ef84a4fb1c6d1c9c34d626a5a83033db6 divinity76 divinity76@gmail.com 2024-01-30T10:18:02+01:00 GitHub noreply@github.com 2024-01-30T11:18:02+02:00 main : allow empty --prompt-cache file (#5176)
5589921ef84a4fb1c6d1c9c34d626a5a83033db6 49f44b5c55d801e3d51ddcf409d866047d718905 Romain Neutron romain@neutron.io 2024-01-30T10:16:38+01:00 GitHub noreply@github.com 2024-01-30T11:16:38+02:00 readme : minor (#5204)
49f44b5c55d801e3d51ddcf409d866047d718905 6685cc41c237544623b4b5651eceb9c4280728cc Georgi Gerganov ggerganov@gmail.com 2024-01-30T11:14:44+02:00 GitHub noreply@github.com 2024-01-30T11:14:44+02:00 readme : update hot topics
6685cc41c237544623b4b5651eceb9c4280728cc ceebbb5b21b971941b2533210b74bf359981006c Wu Jian Ping wujp@greatld.com 2024-01-30T17:11:46+08:00 GitHub noreply@github.com 2024-01-30T11:11:46+02:00 server : improve README (#5209)
ceebbb5b21b971941b2533210b74bf359981006c 6daa69ee81851ab26ca8aefca1a4202941fc0262 Paul Tsochantaris ptsochantaris@icloud.com 2024-01-29T22:19:29Z GitHub noreply@github.com 2024-01-29T23:19:29+01:00 ggml alloc: Fix for null dereference on alloc failure (#5200)
6daa69ee81851ab26ca8aefca1a4202941fc0262 fbf1ddec69f7001cc707de17fa74d7200813bbac Jared Van Bortel jared@nomic.ai 2024-01-29T17:11:27-05:00 GitHub noreply@github.com 2024-01-29T17:11:27-05:00 kompute : fix fallback to CPU (#5201)
fbf1ddec69f7001cc707de17fa74d7200813bbac 2aed77eb06a329f0d82bb1c467f4244904d4073f Jared Van Bortel jared@nomic.ai 2024-01-29T15:50:50-05:00 GitHub noreply@github.com 2024-01-29T15:50:50-05:00 Nomic Vulkan backend (#4456)
2aed77eb06a329f0d82bb1c467f4244904d4073f c82d18e863fcde91b4b1109b1d0c73ea4470c405 divinity76 divinity76@gmail.com 2024-01-29T15:45:41+01:00 GitHub noreply@github.com 2024-01-29T09:45:41-05:00 fix typo "RLIMIT_MLOCK" (#5175)
c82d18e863fcde91b4b1109b1d0c73ea4470c405 14fef85e2d5361e6b4dd7a9ecf22bb817362997d Wu Jian Ping wujjpp@hotmail.com 2024-01-29T21:48:10+08:00 GitHub noreply@github.com 2024-01-29T15:48:10+02:00 server : embeddings compatibility for OpenAI (#5190)
14fef85e2d5361e6b4dd7a9ecf22bb817362997d e76627bcce9f77adb6034cb127b7ec93d4287b69 Georgi Gerganov ggerganov@gmail.com 2024-01-29T15:35:54+02:00 GitHub noreply@github.com 2024-01-29T15:35:54+02:00 py : fix except (#5194)
e76627bcce9f77adb6034cb127b7ec93d4287b69 fbe7dfa53caff0a7e830b676e6e949917a5c71b4 Sang-Kil Park sang.park@42dot.ai 2024-01-29T18:24:19+09:00 GitHub noreply@github.com 2024-01-29T11:24:19+02:00 py : improve BPE tokenizer support (#5189)
fbe7dfa53caff0a7e830b676e6e949917a5c71b4 172ac82629815d038702b049070f4c8c02662da5 slaren slarengh@gmail.com 2024-01-29T09:05:13+01:00 GitHub noreply@github.com 2024-01-29T10:05:13+02:00 ggml : add max buffer sizes to opencl and metal backends (#5181)
172ac82629815d038702b049070f4c8c02662da5 d2f650cb5b04ee2726663e79b47da5efe196ce00 Eve 139727413+netrunnereve@users.noreply.github.com 2024-01-29T08:04:47Z GitHub noreply@github.com 2024-01-29T10:04:47+02:00 cmake : fix Vulkan build (#5182)
d2f650cb5b04ee2726663e79b47da5efe196ce00 35dec26cc25a9ff7d8c3ed52326b94f772b911ce Paul Tsochantaris ptsochantaris@icloud.com 2024-01-28T19:50:16Z GitHub noreply@github.com 2024-01-28T21:50:16+02:00 metal : free metal objects (#5161)
35dec26cc25a9ff7d8c3ed52326b94f772b911ce d460510c7222d43a458a17e01d4bbe72437cdd3c Georgi Gerganov ggerganov@gmail.com 2024-01-28T19:48:05+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-28T19:48:05+02:00 sync : ggml
d460510c7222d43a458a17e01d4bbe72437cdd3c 2307523d322af762ae06648b29ec5a9eb1c73032 Georgi Gerganov ggerganov@gmail.com 2024-01-28T18:44:58+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-28T19:47:31+02:00 ggml : minor type fix (int64_t -> size_t)
2307523d322af762ae06648b29ec5a9eb1c73032 0f648573dde61c510560f68244f70ece7e60d8c1 0cc4m picard12@live.de 2024-01-28T18:03:59+01:00 GitHub noreply@github.com 2024-01-28T19:03:59+02:00 ggml : add Vulkan backend (#2059)
0f648573dde61c510560f68244f70ece7e60d8c1 b764b8f1d079ba44d912801ce6d29bd0d94d51cf Abhilash Majumder 30946547+abhilash1910@users.noreply.github.com 2024-01-28T21:26:23+05:30 GitHub noreply@github.com 2024-01-28T17:56:23+02:00 ggml : add unified SYCL backend for Intel GPUs (#2690)
b764b8f1d079ba44d912801ce6d29bd0d94d51cf 9241c3a2ace544aef708334e54bbdddb90208ee8 Georgi Gerganov ggerganov@gmail.com 2024-01-28T16:54:54+02:00 GitHub noreply@github.com 2024-01-28T14:54:54Z flake.lock: Update (#5162)
9241c3a2ace544aef708334e54bbdddb90208ee8 b2b2bf988c098851b4f3831f0cf38394bff75121 Johannes Gäßler johannesg@5d6.de 2024-01-28T09:59:49+01:00 GitHub noreply@github.com 2024-01-28T09:59:49+01:00 Apply min_p to unsorted tokens (#5115)
b2b2bf988c098851b4f3831f0cf38394bff75121 af4980bfedfd8df43b9e4cd1442895e85fee37bc Johannes Gäßler johannesg@5d6.de 2024-01-28T09:35:14+01:00 GitHub noreply@github.com 2024-01-28T09:35:14+01:00 Tests for min_p, sampling queue (#5147)
af4980bfedfd8df43b9e4cd1442895e85fee37bc f2e69d28c01303ca9dc79907f89ef120a6ac4a92 Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2024-01-28T00:30:44-08:00 GitHub noreply@github.com 2024-01-28T10:30:44+02:00 readme : add link to rust bindings (#5148)
f2e69d28c01303ca9dc79907f89ef120a6ac4a92 39baaf55a160909bb9428bd981014218761a20cb sharpHL 132747147+sharpHL@users.noreply.github.com 2024-01-28T16:00:30+08:00 GitHub noreply@github.com 2024-01-28T10:00:30+02:00 llama : add support for Orion-14B (#5118)
39baaf55a160909bb9428bd981014218761a20cb 6db2b41a76ee78d5efdd5c3cddd5d7ad3f646855 Kyle Mistele kyle@mistele.com 2024-01-28T01:55:31-06:00 GitHub noreply@github.com 2024-01-28T09:55:31+02:00 docker : add server-first container images (#5157)
6db2b41a76ee78d5efdd5c3cddd5d7ad3f646855 753eafed0ebd07af6903771327a1786a7c02cf98 John 78893154+cmp-nct@users.noreply.github.com 2024-01-27T16:09:18+01:00 GitHub noreply@github.com 2024-01-27T17:09:18+02:00 llava : support for Yi-VL and fix for mobileVLM (#5093)
753eafed0ebd07af6903771327a1786a7c02cf98 e9764230054e01553bdead6f2bfd8e001869599d Georgi Gerganov ggerganov@gmail.com 2024-01-27T16:59:20+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-27T17:00:24+02:00 sync : ggml
e9764230054e01553bdead6f2bfd8e001869599d 35a2ee914308c85ab5cb576467381443ad23f0ac Judd foldl@users.noreply.github.com 2024-01-26T21:04:01+08:00 Georgi Gerganov ggerganov@gmail.com 2024-01-27T16:59:00+02:00 ggml : check ggml_add src1 type (ggml/708)
35a2ee914308c85ab5cb576467381443ad23f0ac ec903c034131848da9222536ff18da07ec0882a0 Michael Klimenko mklimenko29@gmail.com 2024-01-27T15:25:55+01:00 GitHub noreply@github.com 2024-01-27T15:25:55+01:00 Remove unused data and add fixes (#5154)
ec903c034131848da9222536ff18da07ec0882a0 a1d6df129bcd3d42cda38c09217d8d4ec4ea3bdd Maximilian Winter maximilian.winter.91@gmail.com 2024-01-27T14:38:05+01:00 GitHub noreply@github.com 2024-01-27T15:38:05+02:00 server : add self-extend support (#5104)
a1d6df129bcd3d42cda38c09217d8d4ec4ea3bdd bbe7c56c9993af86aa2d84cbe1fd69e1b4300cea 0cc4m picard12@live.de 2024-01-26T23:07:32+01:00 GitHub noreply@github.com 2024-01-26T23:07:32+01:00 Add OpenCL add kernel (#5151)
bbe7c56c9993af86aa2d84cbe1fd69e1b4300cea 62fead3ea0a30c8d424f4a8373fa14165c7c707f Jared Van Bortel jared@nomic.ai 2024-01-26T15:34:06-05:00 GitHub noreply@github.com 2024-01-26T15:34:06-05:00 cmake : pass CPU architecture flags to nvcc (#5146)
62fead3ea0a30c8d424f4a8373fa14165c7c707f 15b4538ff29b280a395a1406d711497d8eaa2564 slaren slarengh@gmail.com 2024-01-26T18:59:43+01:00 GitHub noreply@github.com 2024-01-26T18:59:43+01:00 cuda : fix tensor size calculation for non-split buffer (#5145)
15b4538ff29b280a395a1406d711497d8eaa2564 7032f4f6349c17a8352f9f93f7d2122f45469e59 slaren slarengh@gmail.com 2024-01-26T18:18:26+01:00 GitHub noreply@github.com 2024-01-26T19:18:26+02:00 ggml-alloc : add 10% margin to the buffer sizes (#5149)
7032f4f6349c17a8352f9f93f7d2122f45469e59 5f1925a8cef81eb9b372faaae34b0dd76d5361d4 snadampal 87143774+snadampal@users.noreply.github.com 2024-01-26T11:17:59-06:00 GitHub noreply@github.com 2024-01-26T19:17:59+02:00 ggml : update softmax n_task calculation (#5126)
5f1925a8cef81eb9b372faaae34b0dd76d5361d4 3b7c914de25c6851396d7f9178249f1ed278120e Georgi Gerganov ggerganov@gmail.com 2024-01-26T17:09:44+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-26T17:09:44+02:00 scripts : move run-with-preset.py from root to scripts folder
3b7c914de25c6851396d7f9178249f1ed278120e 48c857aa10aea73210a4a72da3f1a6f99269e75d Georgi Gerganov ggerganov@gmail.com 2024-01-26T14:48:15+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-26T14:48:15+02:00 tests : gitignore test-c.o
48c857aa10aea73210a4a72da3f1a6f99269e75d 413e7b0559f922bd4de5e9eec548829d111651b1 Xuan Son Nguyen thichthat@gmail.com 2024-01-26T13:42:20+01:00 GitHub noreply@github.com 2024-01-26T14:42:20+02:00 server : refactored the task processing logic (#5065)
413e7b0559f922bd4de5e9eec548829d111651b1 6dd3c28c9cd1ef74b49d79f47d668759346a3c6c crasm crasm@git.vczf.net 2024-01-26T07:18:00-05:00 GitHub noreply@github.com 2024-01-26T14:18:00+02:00 ci : add model tests + script wrapper (#4586)
6dd3c28c9cd1ef74b49d79f47d668759346a3c6c 38b431de232d1b736b5af19b8c7d72f7075a70bc Paul Tsochantaris ptsochantaris@icloud.com 2024-01-26T12:16:07Z GitHub noreply@github.com 2024-01-26T14:16:07+02:00 metal : remove unused `n_buffers` and `buffers` (#5129)
38b431de232d1b736b5af19b8c7d72f7075a70bc aad0b01d7380a7cdfe0dd42307b18c7b6bac9575 Riceball LEE snowyu.lee@gmail.com 2024-01-26T17:10:28+08:00 GitHub noreply@github.com 2024-01-26T11:10:28+02:00 gguf : fix "general.alignment" type in gguf_reader.py (#5136)
aad0b01d7380a7cdfe0dd42307b18c7b6bac9575 1182cf4d4f6ee383b92695c2e3fe438086dcdba7 Georgi Gerganov ggerganov@gmail.com 2024-01-26T10:52:33+02:00 GitHub noreply@github.com 2024-01-26T10:52:33+02:00 readme : update hot topics
1182cf4d4f6ee383b92695c2e3fe438086dcdba7 fe54033b69b83164cabb5f3ed92dc0ff7ea47605 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-26T09:14:39+02:00 GitHub noreply@github.com 2024-01-26T09:14:39+02:00 Another bucket sort (#5109)
fe54033b69b83164cabb5f3ed92dc0ff7ea47605 5eaf9964fc797d4585c214db32a463d557f3ed33 XiaotaoChen chenxiaotao1234@gmail.com 2024-01-26T04:14:32+08:00 GitHub noreply@github.com 2024-01-25T22:14:32+02:00 readme : add MobileVLM 1.7B/3B to the supported models list (#5107)
5eaf9964fc797d4585c214db32a463d557f3ed33 d292f4f2047963f558dd516f1baaa71793e9acf2 l3utterfly gc.pthzfoldr@gmail.com 2024-01-26T05:06:22+09:00 GitHub noreply@github.com 2024-01-25T22:06:22+02:00 llama : dynamic temperature sampling (#4972)
d292f4f2047963f558dd516f1baaa71793e9acf2 256d1bb0ddce6a0a21f5a7503019bdd5c1933cba Jared Van Bortel jared@nomic.ai 2024-01-25T14:51:24-05:00 GitHub noreply@github.com 2024-01-25T14:51:24-05:00 examples : make pydantic scripts pass mypy and support py3.8 (#5099)
256d1bb0ddce6a0a21f5a7503019bdd5c1933cba faa3526a1eba458120987ed8269e5616385a76f4 Valentin Konovalov valle.ketsujin@gmail.com 2024-01-25T12:05:51-05:00 GitHub noreply@github.com 2024-01-25T19:05:51+02:00 android : use release cmake build type by default (#5123)
faa3526a1eba458120987ed8269e5616385a76f4 ddc5a5033f948dc7ab0a3a6ec2d914d13c274077 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-25T17:58:53+02:00 GitHub noreply@github.com 2024-01-25T17:58:53+02:00 Fix Q3_K_XS for MoE models (#5113)
ddc5a5033f948dc7ab0a3a6ec2d914d13c274077 cd4fddb29f81d6a1f6d51a0c016bc6b486d68def Georgi Gerganov ggerganov@gmail.com 2024-01-25T11:26:17+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-25T11:26:17+02:00 metal : show compile log messages
cd4fddb29f81d6a1f6d51a0c016bc6b486d68def c9b316c78fba31e65879a2ec91cbafd341b88cce Engininja2 139037756+Engininja2@users.noreply.github.com 2024-01-24T16:18:15-06:00 GitHub noreply@github.com 2024-01-24T23:18:15+01:00 cuda : fix 2-bit quants on amd hip (#5105)
c9b316c78fba31e65879a2ec91cbafd341b88cce bf63d695b804b1c995c7ae4427a8a86936ea6d25 Michael Hueschen m@mhueschen.dev 2024-01-22T16:44:10-07:00 Someone newkozlukov@gmail.com 2024-01-24T12:39:29Z nix-shell: use addToSearchPath
bf63d695b804b1c995c7ae4427a8a86936ea6d25 1387ea21178f9f154944013d4dd9764b54c69deb Michael Hueschen m@mhueschen.dev 2024-01-22T03:17:05-07:00 Someone newkozlukov@gmail.com 2024-01-24T12:39:29Z nix: add cc to devShell LD_LIBRARY_PATH
1387ea21178f9f154944013d4dd9764b54c69deb 26d607608d794efa56df3bdb6043a2f94c1d632c slaren slarengh@gmail.com 2024-01-24T12:48:14+01:00 GitHub noreply@github.com 2024-01-24T12:48:14+01:00 llama : pre-allocate input tensors in a separate buffer (#5100)
26d607608d794efa56df3bdb6043a2f94c1d632c 44879ee885f48ecf4675dd216b373dce0a6f3690 Georgi Gerganov ggerganov@gmail.com 2024-01-23T15:50:56+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-23T15:50:56+02:00 metal : disable support for MUL_MAT F32 x F16
44879ee885f48ecf4675dd216b373dce0a6f3690 9ecdd12e95aee20d6dfaf5f5a0f0ce5ac1fb2747 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-23T15:17:20+02:00 GitHub noreply@github.com 2024-01-23T15:17:20+02:00 Additional KL-divergence statistics (#5081)
9ecdd12e95aee20d6dfaf5f5a0f0ce5ac1fb2747 89758723c75ba594e401f6513751beeba7ca1d28 Johannes Gäßler johannesg@5d6.de 2024-01-23T13:31:56+01:00 GitHub noreply@github.com 2024-01-23T13:31:56+01:00 CUDA: more info when no device code (#5088)
89758723c75ba594e401f6513751beeba7ca1d28 2bed4aa3f37cb4e39e16e9ec7b595a7738fd5faf Georgi Gerganov ggerganov@gmail.com 2024-01-23T14:12:57+02:00 GitHub noreply@github.com 2024-01-23T14:12:57+02:00 minor : clean-up some warnings and style (#5094)
2bed4aa3f37cb4e39e16e9ec7b595a7738fd5faf 125d03a5036a02a983c8e98c2cdc126e061afb8e Xuan Son Nguyen thichthat@gmail.com 2024-01-23T08:11:39+01:00 GitHub noreply@github.com 2024-01-23T09:11:39+02:00 devops : add intel oneapi dockerfile (#5068)
125d03a5036a02a983c8e98c2cdc126e061afb8e 011e8ec577fd135cbc02993d3ea9840c516d6a1c Michael Coppola m18coppola@gmail.com 2024-01-23T01:51:27-05:00 GitHub noreply@github.com 2024-01-23T08:51:27+02:00 llama.vim : added api key support (#5090)
011e8ec577fd135cbc02993d3ea9840c516d6a1c 6f9939d119b2d004c264952eb510bd106455531e slaren slarengh@gmail.com 2024-01-22T23:42:41+01:00 GitHub noreply@github.com 2024-01-22T23:42:41+01:00 llama : fix not enough space in buffer with Qwen (#5086)
6f9939d119b2d004c264952eb510bd106455531e 780e24a22eb595b705cbe8284771e9ceff1c4dd2 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-22T16:10:14+02:00 GitHub noreply@github.com 2024-01-22T16:10:14+02:00 KL-divergence (#5076)
780e24a22eb595b705cbe8284771e9ceff1c4dd2 3ce7e8f8e7ccfce07e5947ac5f1f3f4628cf68ea Reinforce-II fate@eastal.com 2024-01-22T21:15:08+08:00 GitHub noreply@github.com 2024-01-22T15:15:08+02:00 ggml : parallelize FP32 conversion when using BLAS (#5045)
3ce7e8f8e7ccfce07e5947ac5f1f3f4628cf68ea b2d80e105a59b54822edf7ce7f3ed5f317e96e21 XiaotaoChen chenxiaotao1234@gmail.com 2024-01-22T21:09:35+08:00 GitHub noreply@github.com 2024-01-22T15:09:35+02:00 llava : MobileVLM support (#4954)
b2d80e105a59b54822edf7ce7f3ed5f317e96e21 28603cd2833cedd4434f398d847f87fc83546dbb Someone Serge sergei.kozlukov@aalto.fi 2024-01-21T03:41:37Z Someone newkozlukov@gmail.com 2024-01-22T12:19:30Z flake.nix: add a comment about flakes vs nix
28603cd2833cedd4434f398d847f87fc83546dbb 5e97ec91ae3038720a5b15cde4c52d2a53ec2137 Someone Serge sergei.kozlukov@aalto.fi 2024-01-21T03:29:38Z Someone newkozlukov@gmail.com 2024-01-22T12:19:30Z nix: add a comment on the many nixpkgs-with-cuda instances
5e97ec91ae3038720a5b15cde4c52d2a53ec2137 7251870780e2d572dd6f239d7a0bfe438c82fa74 Someone Serge sergei.kozlukov@aalto.fi 2024-01-21T03:15:13Z Someone newkozlukov@gmail.com 2024-01-22T12:19:30Z nix: add a comment about makeScope
7251870780e2d572dd6f239d7a0bfe438c82fa74 fe8b3c0d4b0d806e8b46660e24eaf4b90b8b385f Someone Serge sergei.kozlukov@aalto.fi 2024-01-13T17:45:01Z Someone newkozlukov@gmail.com 2024-01-22T12:19:30Z nix: refactor the cleanSource rules
fe8b3c0d4b0d806e8b46660e24eaf4b90b8b385f f4dd059259d0234913b9e9780e1662811744c09d Someone Serge sergei.kozlukov@aalto.fi 2024-01-13T17:38:32Z Someone newkozlukov@gmail.com 2024-01-22T12:19:30Z workflows: nix-ci: drop the redundant "paths" filter
f4dd059259d0234913b9e9780e1662811744c09d f7276f7500f7ea588836dd1fc6f126334c517878 Someone Serge sergei.kozlukov@aalto.fi 2024-01-13T17:16:54Z Someone newkozlukov@gmail.com 2024-01-22T12:19:30Z workflows: nix-build-aarch64: rate limit
f7276f7500f7ea588836dd1fc6f126334c517878 15bceec2d73d4166340b46b27677c45ac1b4cdad Someone Serge sergei.kozlukov@aalto.fi 2024-01-13T17:10:19Z Someone newkozlukov@gmail.com 2024-01-22T12:19:30Z workflows: nix-ci: rebuild on flake.lock updates
15bceec2d73d4166340b46b27677c45ac1b4cdad d6bd4d46ddb6926087c11e0f6633ab1c81da58c3 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-22T14:18:43+02:00 GitHub noreply@github.com 2024-01-22T14:18:43+02:00 imatrix : keep intermediate imatrix results (#5077)
d6bd4d46ddb6926087c11e0f6633ab1c81da58c3 152d9d05e097e35f1cac21262946d57faec7542a compilade 113953597+compilade@users.noreply.github.com 2024-01-22T06:21:52-05:00 GitHub noreply@github.com 2024-01-22T13:21:52+02:00 llama : support StableLM 2 1.6B (#5052)
152d9d05e097e35f1cac21262946d57faec7542a 66d575c45c5a370d668f9c3283cdf348e2329fa2 Daniel Bevenius daniel.bevenius@gmail.com 2024-01-22T12:11:01+01:00 GitHub noreply@github.com 2024-01-22T13:11:01+02:00 finetune : print sample-start/include-sample-start (#5072)
66d575c45c5a370d668f9c3283cdf348e2329fa2 57744932c64266359ee905518de7e096c0295d8c Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-22T12:43:33+02:00 GitHub noreply@github.com 2024-01-22T12:43:33+02:00 llama : add Q3_K_XS (#5060)
57744932c64266359ee905518de7e096c0295d8c 3466c6ebcf668cac453f891b493ead19283347a8 bobqianic 129547291+bobqianic@users.noreply.github.com 2024-01-22T08:55:05Z GitHub noreply@github.com 2024-01-22T10:55:05+02:00 ci : fix Windows CI by updating Intel SDE version (#5053)
3466c6ebcf668cac453f891b493ead19283347a8 504dc37be8446fb09b1ede70300250ad41be32a2 Shijie 821898965@qq.com 2024-01-22T15:33:19+08:00 GitHub noreply@github.com 2024-01-22T09:33:19+02:00 llama : add more qwen2 models (#5071)
504dc37be8446fb09b1ede70300250ad41be32a2 05490fad7f7f60ff2bed9ad05cd81b44e82ccde3 iSma ismail.senhaji@gmail.com 2024-01-21T22:37:13+01:00 GitHub noreply@github.com 2024-01-21T21:37:13Z Revert LLAMA_NATIVE to OFF in flake.nix (#5066)
05490fad7f7f60ff2bed9ad05cd81b44e82ccde3 6c5629d4d2d15557c38a0e609b30c1a42abad80d kuronekosaiko EvanChanJ@163.com 2024-01-22T00:28:14+08:00 GitHub noreply@github.com 2024-01-21T17:28:14+01:00 add safetensors support to convert-lora-to-ggml.py (#5062)
6c5629d4d2d15557c38a0e609b30c1a42abad80d 7dcbe39d36b76389f6c5cd3b151928472b7e22ff bobqianic 129547291+bobqianic@users.noreply.github.com 2024-01-21T15:17:35Z GitHub noreply@github.com 2024-01-21T10:17:35-05:00 add `#include <string>` to unicode.h (#5051)
7dcbe39d36b76389f6c5cd3b151928472b7e22ff 726c0fa9a2da976e9c5d5c51e185d9dd453fc9e5 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-21T14:42:44+02:00 GitHub noreply@github.com 2024-01-21T14:42:44+02:00 Add ability to evauate multiple choice tasks (#5047)
726c0fa9a2da976e9c5d5c51e185d9dd453fc9e5 942c0107a7301434c0a5e7da46bc4cf2393aa556 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-21T08:01:20+02:00 GitHub noreply@github.com 2024-01-21T08:01:20+02:00 Slightly faster imatrix (#5050)
942c0107a7301434c0a5e7da46bc4cf2393aa556 b43ebde3b0ccbc42d9dd782b32e2fd8eb35b43b5 Georgi Gerganov ggerganov@gmail.com 2024-01-21T05:17:27+02:00 GitHub noreply@github.com 2024-01-21T03:17:27Z flake.lock: Update (#5054)
b43ebde3b0ccbc42d9dd782b32e2fd8eb35b43b5 97c1549808d2742d37584a3c9df28154bdf34417 Jared Van Bortel jared@nomic.ai 2024-01-20T18:14:18-05:00 GitHub noreply@github.com 2024-01-20T18:14:18-05:00 convert : partially revert PR #4818 (#5041)
97c1549808d2742d37584a3c9df28154bdf34417 6df465a91d402370bcba6676b19fad85b06ce7e0 Jared Van Bortel jared@nomic.ai 2024-01-20T10:08:08-05:00 GitHub noreply@github.com 2024-01-20T17:08:08+02:00 perplexity : fix MSVC build after #5020 (#5043)
6df465a91d402370bcba6676b19fad85b06ce7e0 77bc1bbd05f0c31cb45773eb5eb59b9ff2b07e1b slaren slarengh@gmail.com 2024-01-20T16:05:49+01:00 GitHub noreply@github.com 2024-01-20T17:05:49+02:00 llama : run all KQV ops on the CPU with no KV offload (#5049)
77bc1bbd05f0c31cb45773eb5eb59b9ff2b07e1b 48e2b1337257bb77573bf76cfffcff3a5efa8704 Herman Semenov GermanAizek@yandex.ru 2024-01-20T08:11:31Z GitHub noreply@github.com 2024-01-20T10:11:31+02:00 cmake : add support for ccache (#5002)
48e2b1337257bb77573bf76cfffcff3a5efa8704 cca894f16a5eade15afd07b015e4cb3d8658943f adel boussaken netdur@gmail.com 2024-01-20T09:05:43+01:00 GitHub noreply@github.com 2024-01-20T03:05:43-05:00 Add a dart/flutter binding to README.md (#4882)
cca894f16a5eade15afd07b015e4cb3d8658943f 381ee195721d8e747ee31a60c0751822b3072f02 Kylin 56434533+KyL0N@users.noreply.github.com 2024-01-20T15:01:46+08:00 GitHub noreply@github.com 2024-01-20T09:01:46+02:00 cuda : fix compile error in jetson platform (#4975)
381ee195721d8e747ee31a60c0751822b3072f02 a5cacb22b2114fd9adf61c00cbb237384d86bced Uzo Nweke uzoechi@gmail.com 2024-01-19T13:20:50-05:00 GitHub noreply@github.com 2024-01-19T20:20:50+02:00 finetune : fix ggml_allocr lifetimes (tmp workaround) (#5033)
a5cacb22b2114fd9adf61c00cbb237384d86bced 9b75cb2b3ccbed3df2e14c1202168db3e5145095 Georgi Gerganov ggerganov@gmail.com 2024-01-19T15:24:47+02:00 GitHub noreply@github.com 2024-01-19T15:24:47+02:00 imatrix : add README.md
9b75cb2b3ccbed3df2e14c1202168db3e5145095 de9a147df14e62f54f879d2d15e6c4793107f4fc Shijie 821898965@qq.com 2024-01-19T19:53:13+08:00 GitHub noreply@github.com 2024-01-19T13:53:13+02:00 llama : support upcoming Qwen2 (#5037)
de9a147df14e62f54f879d2d15e6c4793107f4fc 7051aacfac0057fa5fac9ea46c55bffc3892d810 Georgi Gerganov ggerganov@gmail.com 2024-01-19T13:52:22+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-19T13:52:22+02:00 py : fix flake8 lint
7051aacfac0057fa5fac9ea46c55bffc3892d810 2b3b999cacc7ad1207c32fbdf3479a19c06e1a34 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-19T11:39:11+02:00 GitHub noreply@github.com 2024-01-19T11:39:11+02:00 winogrande: evaluate log-probs in parallel (#5036)
2b3b999cacc7ad1207c32fbdf3479a19c06e1a34 993fba81807e55d27b570945af8e416d535eced1 chiranko 96988916+chiranko@users.noreply.github.com 2024-01-19T17:07:27+08:00 GitHub noreply@github.com 2024-01-19T11:07:27+02:00 llama : add CodeShell support (#5016)
993fba81807e55d27b570945af8e416d535eced1 8b20858e5e9c44b99b4b31ae9c40b8f20d01d94f Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-19T11:02:39+02:00 GitHub noreply@github.com 2024-01-19T11:02:39+02:00 perplexity: avoid unnecessary alloocations and logit copies (#5035)
8b20858e5e9c44b99b4b31ae9c40b8f20d01d94f 57e2a7a52a819883f40dada8a2edc24ecf48186b Georgi Gerganov ggerganov@gmail.com 2024-01-19T10:45:06+02:00 GitHub noreply@github.com 2024-01-19T10:45:06+02:00 perplexity : faster Winogrande via batching (#5024)
57e2a7a52a819883f40dada8a2edc24ecf48186b 9b6ea4263ab45e02ff905bf7a29dc143ca1facc3 John 78893154+cmp-nct@users.noreply.github.com 2024-01-18T23:12:15+01:00 GitHub noreply@github.com 2024-01-19T00:12:15+02:00 llama : fix falcon arch for tied output embeddings (#4978)
9b6ea4263ab45e02ff905bf7a29dc143ca1facc3 821f0a271e7c9ee737945245dd7abfa22cc9b5b0 Georgi Gerganov ggerganov@gmail.com 2024-01-18T23:36:07+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-18T23:36:07+02:00 cmake : add ggml public headers (#5011)
821f0a271e7c9ee737945245dd7abfa22cc9b5b0 96d7f56d2918ffde1995dbb32392571deb76d7fc Xuan Son Nguyen thichthat@gmail.com 2024-01-18T21:33:05+01:00 GitHub noreply@github.com 2024-01-18T22:33:05+02:00 server : defer tasks when "slot unavailable" (#5018)
96d7f56d2918ffde1995dbb32392571deb76d7fc 2d5419d08ab1131623e6a1d554607b7663435e87 slaren slarengh@gmail.com 2024-01-18T21:12:15+01:00 GitHub noreply@github.com 2024-01-18T21:12:15+01:00 llama : fix mlock with no-mmap with Metal (#5025)
2d5419d08ab1131623e6a1d554607b7663435e87 d391ae9b4919e24624cc963d82162450848beaf4 Georgi Gerganov ggerganov@gmail.com 2024-01-18T21:45:51+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-18T21:45:51+02:00 imatrix : fix assert for src0 non-cont check
d391ae9b4919e24624cc963d82162450848beaf4 e9240cdfa06a50c1b5dbafa367cb8cd698e65103 Georgi Gerganov ggerganov@gmail.com 2024-01-18T20:49:00+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-18T20:49:00+02:00 perplexity : fix winogrande N tasks option
e9240cdfa06a50c1b5dbafa367cb8cd698e65103 b46757735d30f5c6ed4f20ebeccc684e02d4f3bf Georgi Gerganov ggerganov@gmail.com 2024-01-18T20:45:39+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-18T20:45:39+02:00 scripts : add get-winogrande.sh
b46757735d30f5c6ed4f20ebeccc684e02d4f3bf 3e945cc1e9c06d2001031360e4e303e9548fb02c David Sommers 12738+databyte@users.noreply.github.com 2024-01-18T12:20:59-05:00 GitHub noreply@github.com 2024-01-18T19:20:59+02:00 convert.py : fix llama/llama2 conversion due to vocab_size=-1 (#5019)
3e945cc1e9c06d2001031360e4e303e9548fb02c ad19812cda4062c9f154ef16315df41fbe6a770a Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-18T19:18:21+02:00 GitHub noreply@github.com 2024-01-18T19:18:21+02:00 HellaSwag: speed up by parallelizing log-prob evaluation (#5020)
ad19812cda4062c9f154ef16315df41fbe6a770a 682986a08eb5cb04865d2e713449f17304d266d8 Georgi Gerganov ggerganov@gmail.com 2024-01-18T15:33:01+02:00 GitHub noreply@github.com 2024-01-18T15:33:01+02:00 perplexity : faster HellaSwag via batching (#5017)
682986a08eb5cb04865d2e713449f17304d266d8 dcad445d0c83ad49bca1b58cf9c139cfcebee5d4 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-18T13:46:27+02:00 GitHub noreply@github.com 2024-01-18T13:46:27+02:00 Add Winogrande evaluation (#5015)
dcad445d0c83ad49bca1b58cf9c139cfcebee5d4 1e605f4102c7ea8dc0dff82f5eaa6a71973d549f Georgi Gerganov ggerganov@gmail.com 2024-01-18T11:44:49+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-18T11:44:49+02:00 scritps : add helper script to get hellaswag data in txt format
1e605f4102c7ea8dc0dff82f5eaa6a71973d549f 6b6916b215251e09bd57cdbf870dc8a73345edc2 Paul Tsochantaris ptsochantaris@icloud.com 2024-01-18T08:47:24Z GitHub noreply@github.com 2024-01-18T10:47:24+02:00 metal : fix memory leak, dangling pointer and unused autorel (#5007)
6b6916b215251e09bd57cdbf870dc8a73345edc2 38566680cdfe982a495562332c25b9227de9cf8d Georgi Gerganov ggerganov@gmail.com 2024-01-17T20:54:50+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-17T20:54:50+02:00 sync : ggml
38566680cdfe982a495562332c25b9227de9cf8d ba69bbc84ced580fe4fdb0713ca2d95634325b7a Georgi Gerganov ggerganov@gmail.com 2024-01-17T18:54:56+02:00 GitHub noreply@github.com 2024-01-17T18:54:56+02:00 ggml : add IQ2 to test-backend-ops + refactoring (#4990)
ba69bbc84ced580fe4fdb0713ca2d95634325b7a 44a1a4a41a4c0b03afaa7d9e06bcbc7cf95aa1e6 Georgi Gerganov ggerganov@gmail.com 2024-01-17T18:46:30+02:00 GitHub noreply@github.com 2024-01-17T18:46:30+02:00 imatrix : offload to GPU support (#4957)
44a1a4a41a4c0b03afaa7d9e06bcbc7cf95aa1e6 c918fe8dca8fa1c4602427e0a4b88e20046f6c34 Georgi Gerganov ggerganov@gmail.com 2024-01-17T18:39:41+02:00 GitHub noreply@github.com 2024-01-17T18:39:41+02:00 backend : add eval callback (#4935)
c918fe8dca8fa1c4602427e0a4b88e20046f6c34 0f83e727af0a7cadf90b7ecc1f8e35de1d0880bc Georgi Gerganov ggerganov@gmail.com 2024-01-17T18:38:39+02:00 GitHub noreply@github.com 2024-01-17T18:38:39+02:00 metal : create autorelease pool during library build (#4970)
0f83e727af0a7cadf90b7ecc1f8e35de1d0880bc 4f4bf35f46600441dec2f941e667291eeb9a18d8 Georgi Gerganov ggerganov@gmail.com 2024-01-17T18:37:36+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-17T18:37:36+02:00 py : fix whitespace
4f4bf35f46600441dec2f941e667291eeb9a18d8 2b3a665d3917edf393761a24c4835447894df74a Georgi Gerganov ggerganov@gmail.com 2024-01-17T15:45:03+02:00 GitHub noreply@github.com 2024-01-17T15:45:03+02:00 py : fix missing added_tokens_dict for SPM and BPE vocabs (#4971)
2b3a665d3917edf393761a24c4835447894df74a 75632936659772d5b2ce54b0b65319fecbaac2e6 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-17T12:36:37+02:00 GitHub noreply@github.com 2024-01-17T12:36:37+02:00 llama : use Q4_K for attn_v for Q2_K_S when n_gqa >= 4 (#4996)
75632936659772d5b2ce54b0b65319fecbaac2e6 f46c0c1b0ea0bc67e24e4bf026a7e898c1af22a9 Paul Tsochantaris ptsochantaris@icloud.com 2024-01-17T08:07:24Z GitHub noreply@github.com 2024-01-17T10:07:24+02:00 metal : remove unnecessary nil check (#4986)
f46c0c1b0ea0bc67e24e4bf026a7e898c1af22a9 5c999609013a30c06e6fd28be8db5c2074bcc196 David Renshaw dwrenshaw@gmail.com 2024-01-17T02:17:50-05:00 GitHub noreply@github.com 2024-01-17T09:17:50+02:00 llama : fix copy/paste error in llama_sampling_params comment (#4994)
5c999609013a30c06e6fd28be8db5c2074bcc196 bee938da74c33f42242c3a1058ac0a0a6eeef531 Georgi Gerganov ggerganov@gmail.com 2024-01-16T20:59:31+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-16T20:59:31+02:00 py : remove unnecessary hasattr (#4903)
bee938da74c33f42242c3a1058ac0a0a6eeef531 cec8a4847062fbd76253e3b085683f39d91e80d3 Philip Taron philip.taron@gmail.com 2024-01-16T09:56:21-08:00 GitHub noreply@github.com 2024-01-16T09:56:21-08:00 nix: remove nixConfig from flake.nix (#4984)
cec8a4847062fbd76253e3b085683f39d91e80d3 334a835a1ccc8106a5fa355683a965efb1bfa24b Daniel Bevenius daniel.bevenius@gmail.com 2024-01-16T18:54:24+01:00 GitHub noreply@github.com 2024-01-16T19:54:24+02:00 finetune : add training data file to log message (#4979)
334a835a1ccc8106a5fa355683a965efb1bfa24b 4feb4b33eeb1756e46084a4db9230b279af1a480 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-16T19:51:26+02:00 GitHub noreply@github.com 2024-01-16T19:51:26+02:00 ggml : importance matrix support for legacy quants (#4969)
4feb4b33eeb1756e46084a4db9230b279af1a480 959ef0c0df725c013c7f712eaa7790b8e38a8e20 Maximilian Winter maximilian.winter.91@gmail.com 2024-01-16T18:41:42+01:00 GitHub noreply@github.com 2024-01-16T19:41:42+02:00 examples : add complete parallel function calling example (#4974)
959ef0c0df725c013c7f712eaa7790b8e38a8e20 c37b3474e61d609d43cccc3bde5d559e80e4f5d1 Georgi Gerganov ggerganov@gmail.com 2024-01-16T19:34:54+02:00 GitHub noreply@github.com 2024-01-16T19:34:54+02:00 perplexity : fix kv cache handling for hellaswag (#4981)
c37b3474e61d609d43cccc3bde5d559e80e4f5d1 158f8c9e21302114bac3c646f80ea85b52ffa0bd Georgi Gerganov ggerganov@gmail.com 2024-01-16T19:13:54+02:00 GitHub noreply@github.com 2024-01-16T09:13:54-08:00 flake.lock: update flake-parts, flake-parts/nixpkgs-lib, and nixpkgs (#4920)
158f8c9e21302114bac3c646f80ea85b52ffa0bd 862f5e41ab1fdf12d6f59455aad3f5dd8258f805 Paul Tsochantaris ptsochantaris@icloud.com 2024-01-16T17:05:19Z GitHub noreply@github.com 2024-01-16T19:05:19+02:00 metal : localized logic in `ggml_metal_graph_compute` (#4924)
862f5e41ab1fdf12d6f59455aad3f5dd8258f805 3a48d558a69c88ac17efcaa5900cd9eb19596ac4 Neuman Vong neuman.vong@gmail.com 2024-01-17T00:47:34+11:00 GitHub noreply@github.com 2024-01-16T15:47:34+02:00 android : introduce starter project example (#4926)
3a48d558a69c88ac17efcaa5900cd9eb19596ac4 7c8d3abd1a17c28fc56b1a4814bc4b29f91d7454 Alex Azarov alex@azarov.by 2024-01-16T14:41:27+01:00 GitHub noreply@github.com 2024-01-16T15:41:27+02:00 metal : replace loop of dispatch_async with dispatch_apply (#4934)
7c8d3abd1a17c28fc56b1a4814bc4b29f91d7454 122ed4840cc6d209df6043e027f9f8a03aee01da Alex Azarov alex@azarov.by 2024-01-16T14:33:02+01:00 GitHub noreply@github.com 2024-01-16T15:33:02+02:00 metal : log `recommendedMaxWorkingSetSize` on iOS 16+ (#4936)
122ed4840cc6d209df6043e027f9f8a03aee01da a0b3ac8c48b66206b9c5921ce57bd5c0ea6557c3 Maximilian Winter maximilian.winter.91@gmail.com 2024-01-16T13:10:48+01:00 GitHub noreply@github.com 2024-01-16T14:10:48+02:00 examples : fix and improv docs for the grammar generator (#4909)
a0b3ac8c48b66206b9c5921ce57bd5c0ea6557c3 d75c232e1da56f19ac4d2530dadbe0ab3a11fde5 Justine Tunney jtunney@gmail.com 2024-01-16T03:16:33-08:00 GitHub noreply@github.com 2024-01-16T13:16:33+02:00 ggml : introduce GGML_CALL function annotation (#4850)
d75c232e1da56f19ac4d2530dadbe0ab3a11fde5 e0324285a569d0583cf2f4a07a2402221ee25f58 Daniel Bevenius daniel.bevenius@gmail.com 2024-01-16T12:14:19+01:00 GitHub noreply@github.com 2024-01-16T13:14:19+02:00 finetune : use LLAMA_FILE_MAGIC_GGLA (#4961)
e0324285a569d0583cf2f4a07a2402221ee25f58 3e5ca7931c68152e4ec18d126e9c832dd84914c8 stduhpf stephduh@live.fr 2024-01-16T12:04:32+01:00 GitHub noreply@github.com 2024-01-16T13:04:32+02:00 speculative : threading options (#4959)
3e5ca7931c68152e4ec18d126e9c832dd84914c8 4483396751c79dea540808b9cb9238245d06da2b ngc92 7938269+ngc92@users.noreply.github.com 2024-01-15T20:40:48+02:00 GitHub noreply@github.com 2024-01-15T19:40:48+01:00 pass cpu-architecture arguments only to host code (C;C++) (#4943)
4483396751c79dea540808b9cb9238245d06da2b d9aa4ffa6e0296d42f1f676dd85de97c8491eb73 David Friehs david@friehs.info 2024-01-15T14:06:52+01:00 GitHub noreply@github.com 2024-01-15T15:06:52+02:00 llama : apply classifier-free guidance to logits directly (#4951)
d9aa4ffa6e0296d42f1f676dd85de97c8491eb73 ddb008d845cd50bb090bf051f570130524042936 Victor Z. Peng ziliangdotme@gmail.com 2024-01-15T04:41:46-08:00 GitHub noreply@github.com 2024-01-15T14:41:46+02:00 awq-py : fix typo in awq-py/README.md (#4947)
ddb008d845cd50bb090bf051f570130524042936 2faaef39799c97a53bec3898141478700da25757 Georgi Gerganov ggerganov@gmail.com 2024-01-15T13:27:00+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-15T13:27:00+02:00 cuda : fix dequantize kernel names (#4938)
2faaef39799c97a53bec3898141478700da25757 4a3156de2fac9a8ee4279de7804d4e352dcfe121 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-15T10:09:38+02:00 GitHub noreply@github.com 2024-01-15T10:09:38+02:00 llama : check for 256 divisibility for IQ2_XS, IQ2_XXS (#4950)
4a3156de2fac9a8ee4279de7804d4e352dcfe121 a836c8f534ab789b02da149fbdaf7735500bff74 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-15T07:48:06+02:00 GitHub noreply@github.com 2024-01-15T07:48:06+02:00 CUDA: faster dequantize kernels for Q4_0 and Q4_1 (#4938)
a836c8f534ab789b02da149fbdaf7735500bff74 467a882fd2e5b6172897b49aa45aa29bd3f27685 David Pflug david@pflug.email 2024-01-14T10:46:00-05:00 GitHub noreply@github.com 2024-01-14T17:46:00+02:00 llama : fix missing quotes (#4937)
467a882fd2e5b6172897b49aa45aa29bd3f27685 bb0c1392479398f9aba86d9ec98db0b95ede6e6d Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-14T16:21:12+02:00 GitHub noreply@github.com 2024-01-14T16:21:12+02:00 Add ability to use importance matrix for all k-quants (#4930)
bb0c1392479398f9aba86d9ec98db0b95ede6e6d 9408cfdad6b1c090a7e1419d4434edc260b7e47e Georgi Gerganov ggerganov@gmail.com 2024-01-14T13:26:53+02:00 GitHub noreply@github.com 2024-01-14T13:26:53+02:00 llama : check LLAMA_TRACE env for extra logging (#4929)
9408cfdad6b1c090a7e1419d4434edc260b7e47e 03c526749041c863b0cd842b26b8907e1ea0e0b1 Georgi Gerganov ggerganov@gmail.com 2024-01-14T11:08:09+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-14T11:08:41+02:00 scripts : sync-ggml-am.sh option to skip commits
03c526749041c863b0cd842b26b8907e1ea0e0b1 a128c38de862431f1aae9ccc40b792fbc1b8b682 Georgi Gerganov ggerganov@gmail.com 2024-01-14T11:03:19+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-14T11:03:19+02:00 llama : use LLAMA_LOG_ macros for logging
a128c38de862431f1aae9ccc40b792fbc1b8b682 5f5fe1bd608fa2ed42af97b5f2ea31be6625fc48 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-14T10:53:39+02:00 GitHub noreply@github.com 2024-01-14T10:53:39+02:00 Fix ffn_down quantization mix for MoE models (#4927)
5f5fe1bd608fa2ed42af97b5f2ea31be6625fc48 ac32902a87147f78d63c931aa8a23dee762660e7 Alex Azarov alex@azarov.by 2024-01-14T09:44:39+01:00 GitHub noreply@github.com 2024-01-14T10:44:39+02:00 metal : correctly set SIMD support flags on iOS (#4923)
ac32902a87147f78d63c931aa8a23dee762660e7 147b17ac94a24d524e367cda26a9ff6245689f34 Karthik Kumar Viswanathan 195178+guilt@users.noreply.github.com 2024-01-14T00:41:44-08:00 GitHub noreply@github.com 2024-01-14T10:41:44+02:00 llama : support WinXP build with MinGW 8.1.0 (#3419)
147b17ac94a24d524e367cda26a9ff6245689f34 807179ec583dcb882f97d9704577c06beb2c5ec9 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-14T09:45:56+02:00 GitHub noreply@github.com 2024-01-14T09:45:56+02:00 2-bit quantizations (#4897)
807179ec583dcb882f97d9704577c06beb2c5ec9 76484fbfd355df388f71d6edaa98e1692a74de7e Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-14T09:44:30+02:00 GitHub noreply@github.com 2024-01-14T09:44:30+02:00 Make Q3_K_S be the same as olf Q3_K_L for Mixtral-8x7B (#4906)
76484fbfd355df388f71d6edaa98e1692a74de7e c71d608ce7a1584bf5072f197919dd24f3a6163f Georgi Gerganov ggerganov@gmail.com 2024-01-14T00:14:46+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-14T00:14:46+02:00 sync : ggml
c71d608ce7a1584bf5072f197919dd24f3a6163f 4be5ef556de830c5c4f6e45c05ef4427823fe607 Johannes Gäßler johannesg@5d6.de 2024-01-13T21:41:37+01:00 GitHub noreply@github.com 2024-01-13T21:41:37+01:00 ggml: cache sin/cos for RoPE (#4908)
4be5ef556de830c5c4f6e45c05ef4427823fe607 0ea069b87bd296c556824e57455433b6c0357340 Georgi Gerganov ggerganov@gmail.com 2024-01-13T20:45:45+02:00 GitHub noreply@github.com 2024-01-13T20:45:45+02:00 metal : remove old API (#4919)
0ea069b87bd296c556824e57455433b6c0357340 f172de03f11465dc6c5a0fc3a22f8ec254c6832c Georgi Gerganov ggerganov@gmail.com 2024-01-13T19:31:26+02:00 GitHub noreply@github.com 2024-01-13T19:31:26+02:00 server : fix prompt caching with system prompt (#4914)
f172de03f11465dc6c5a0fc3a22f8ec254c6832c 2d57de525541247132e354f561ff48775fba5d85 Georgi Gerganov ggerganov@gmail.com 2024-01-13T18:47:38+02:00 GitHub noreply@github.com 2024-01-13T18:47:38+02:00 llama : fix detokenization of non-special added-tokens (#4916)
2d57de525541247132e354f561ff48775fba5d85 df845cc982e7e2ea7b9900e29d55b15338faa78d Georgi Gerganov ggerganov@gmail.com 2024-01-13T18:46:37+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-13T18:46:37+02:00 metal : disable log for loaded kernels (#4794)
df845cc982e7e2ea7b9900e29d55b15338faa78d 6b48ed089377330cdb362970a51c1c89b6d857a8 David Friehs david@friehs.info 2024-01-13T17:29:43+01:00 GitHub noreply@github.com 2024-01-13T18:29:43+02:00 llama : minimize size used for state save/load (#4820)
6b48ed089377330cdb362970a51c1c89b6d857a8 722d33f34ec74c6f7046109f936d0928ffe171bc Someone sergei.kozlukov@aalto.fi 2024-01-13T16:29:16Z GitHub noreply@github.com 2024-01-13T16:29:16Z workflows: unbreak nix-build-aarch64, and split it out (#4915)
722d33f34ec74c6f7046109f936d0928ffe171bc c30b1ef39aeba497a943416d2897d69fee055b96 Yann Follet 131855179+YannFollet@users.noreply.github.com 2024-01-14T00:09:08+08:00 GitHub noreply@github.com 2024-01-13T18:09:08+02:00 main : add parameter --no-display-prompt (#4541)
c30b1ef39aeba497a943416d2897d69fee055b96 b38b5e93ae31019e87f692b69d27124eae6aac02 texmex76 40733439+texmex76@users.noreply.github.com 2024-01-13T17:06:20+01:00 GitHub noreply@github.com 2024-01-13T18:06:20+02:00 gguf : fix potential infinite for-loop (#4600)
b38b5e93ae31019e87f692b69d27124eae6aac02 7dc78764e2ff86512e6e31cb0fcb8087df4b4708 Georgi Gerganov ggerganov@gmail.com 2024-01-13T18:03:45+02:00 GitHub noreply@github.com 2024-01-13T18:03:45+02:00 metal : refactor kernel loading code (#4794)
7dc78764e2ff86512e6e31cb0fcb8087df4b4708 356327feb3f66980ab687040495d722696d98970 Johannes Gäßler johannesg@5d6.de 2024-01-13T15:52:53+01:00 GitHub noreply@github.com 2024-01-13T15:52:53+01:00 compare-llama-bench: tweak output format (#4910)
356327feb3f66980ab687040495d722696d98970 ee8243adaa9a9f51ff449213383874e49efe368f Ziad Ben Hadj-Alouane zied.benhadjalouane@gmail.com 2024-01-13T09:20:46-05:00 GitHub noreply@github.com 2024-01-13T16:20:46+02:00 server : fix deadlock that occurs in multi-prompt scenarios (#4905)
ee8243adaa9a9f51ff449213383874e49efe368f 15ebe59210e7fd9817ff67f51fa1a5ee2d004294 makomk makosoft@googlemail.com 2024-01-13T14:16:11Z GitHub noreply@github.com 2024-01-13T16:16:11+02:00 server : fix crash with multimodal models without BOS token (#4904)
15ebe59210e7fd9817ff67f51fa1a5ee2d004294 de473f5f8e19ba5e659cdf5af65fb9251dce16c5 Georgi Gerganov ggerganov@gmail.com 2024-01-13T13:44:37+02:00 GitHub noreply@github.com 2024-01-13T13:44:37+02:00 convert : update phi-2 to latest HF repo (#4903)
de473f5f8e19ba5e659cdf5af65fb9251dce16c5 f238461236f4e0e18cac1a554af23c7deadc9b01 Georgi Gerganov ggerganov@gmail.com 2024-01-12T22:02:43+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-12T22:02:43+02:00 sync : ggml
f238461236f4e0e18cac1a554af23c7deadc9b01 fa5c1fb44a2724292da545d6b7cf2a1ac0e0b989 Georgi Gerganov ggerganov@gmail.com 2024-01-12T14:02:30+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-12T22:02:11+02:00 ggml : fix 32-bit ARM compat for IQ2_XS (whisper/1758)
fa5c1fb44a2724292da545d6b7cf2a1ac0e0b989 52ee4540c0f2e11d52c839db6eb51d014ce060e1 slaren slarengh@gmail.com 2024-01-12T20:38:34+01:00 Georgi Gerganov ggerganov@gmail.com 2024-01-12T22:02:11+02:00 backend_sched : fix assignments
52ee4540c0f2e11d52c839db6eb51d014ce060e1 3fe81781e3bf98b8e44946240a19f3a6ad08a11a Maximilian Winter maximilian.winter.91@gmail.com 2024-01-12T20:46:45+01:00 GitHub noreply@github.com 2024-01-12T21:46:45+02:00 examples : add pydantic models to GBNF grammar generator (#4883)
3fe81781e3bf98b8e44946240a19f3a6ad08a11a e7e4df031b9e29d4b55a4e0b0295187f6b213db1 Johannes Gäßler johannesg@5d6.de 2024-01-12T20:38:54+01:00 GitHub noreply@github.com 2024-01-12T20:38:54+01:00 CUDA: faster q8_0 -> f16 dequantization (#4895)
e7e4df031b9e29d4b55a4e0b0295187f6b213db1 584d674be622fbf1578694ada6e62eebedbfd377 slaren slarengh@gmail.com 2024-01-12T20:07:38+01:00 GitHub noreply@github.com 2024-01-12T20:07:38+01:00 llama : ggml-backend integration (#4766)
584d674be622fbf1578694ada6e62eebedbfd377 930f907d3ece1eb5b0a1ec5e209983a66dcbfa68 Georgi Gerganov ggerganov@gmail.com 2024-01-12T20:54:12+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-12T20:54:12+02:00 llama : remove redundant assert for StableLM (#4901)
930f907d3ece1eb5b0a1ec5e209983a66dcbfa68 e790eef21ce659f5c16d59f8a5c8dcf6cde0692a Daniel Bevenius daniel.bevenius@gmail.com 2024-01-12T18:54:53+01:00 GitHub noreply@github.com 2024-01-12T19:54:53+02:00 export-lora : use LLAMA_FILE_MAGIC_GGLA (#4894)
e790eef21ce659f5c16d59f8a5c8dcf6cde0692a 5537d9d36bfdb4379555431f574d3d78ce6e7955 Zay 95888118+isaiahbjork@users.noreply.github.com 2024-01-12T05:48:00-07:00 GitHub noreply@github.com 2024-01-12T14:48:00+02:00 llama.swiftui : update models layout (#4826)
5537d9d36bfdb4379555431f574d3d78ce6e7955 1b280c9fffd682b6924010a4437f0275f2921fa9 Georgi Gerganov ggerganov@gmail.com 2024-01-12T14:33:21+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-12T14:33:21+02:00 gitignore : imatrix
1b280c9fffd682b6924010a4437f0275f2921fa9 3cabe80630c7eeb57713cd02249053a8cf6894fa Johannes Gäßler johannesg@5d6.de 2024-01-12T12:30:41+01:00 GitHub noreply@github.com 2024-01-12T12:30:41+01:00 CUDA: fix softmax compile for old CUDA versions (#4862)
3cabe80630c7eeb57713cd02249053a8cf6894fa 4315a94366708828f949f9db89d2a8d99b634459 Georgi Gerganov ggerganov@gmail.com 2024-01-12T13:10:19+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-12T13:11:15+02:00 llama : fix typo "imp_embd" -> "inp_embd"
4315a94366708828f949f9db89d2a8d99b634459 2d00741e12c5db4a33dfccd1125f5de4adec9a5b howlger eclipse@voormann.de 2024-01-12T12:05:32+01:00 GitHub noreply@github.com 2024-01-12T13:05:32+02:00 common : streamline the formatting of help (#4890)
2d00741e12c5db4a33dfccd1125f5de4adec9a5b f445c0e68cf8e1faca0b2aa8dfb9d48231cec301 Georgi Gerganov ggerganov@gmail.com 2024-01-12T13:03:38+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-12T13:03:38+02:00 py : fix lint (#4889)
f445c0e68cf8e1faca0b2aa8dfb9d48231cec301 326b418b59b6d48d854c4461a2303e8ac0a311e6 Georgi Gerganov ggerganov@gmail.com 2024-01-12T13:01:56+02:00 GitHub noreply@github.com 2024-01-12T13:01:56+02:00 llama : fix llm_build_k_shift to use correct n_rot (#4889)
326b418b59b6d48d854c4461a2303e8ac0a311e6 1d118386fea031f01550f8cd47a5c86296e5333f Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-12T06:59:57+01:00 GitHub noreply@github.com 2024-01-12T06:59:57+01:00 Importance Matrix calculation (#4861)
1d118386fea031f01550f8cd47a5c86296e5333f 7edefbd79cc6dea96640edc54c6b94b2b2496d8b Georgi Gerganov ggerganov@gmail.com 2024-01-11T23:23:49+02:00 GitHub noreply@github.com 2024-01-11T23:23:49+02:00 server : fix infill when prompt is empty (#4833)
7edefbd79cc6dea96640edc54c6b94b2b2496d8b 3ca63b4538dfc78aaec88cd2c3e3f8417c1924e3 Georgi Gerganov ggerganov@gmail.com 2024-01-11T22:46:26+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T22:46:26+02:00 main : better name for variable n_print (#4874)
3ca63b4538dfc78aaec88cd2c3e3f8417c1924e3 b0377875488b33f7114138687d828da1de61775d Georgi Gerganov ggerganov@gmail.com 2024-01-11T22:43:05+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T22:43:05+02:00 main : disable token count by default (#4874)
b0377875488b33f7114138687d828da1de61775d 469e75d0a35b08de549a4fd87f082ca7a8a539ba Georgi Gerganov ggerganov@gmail.com 2024-01-11T21:58:28+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T21:58:28+02:00 swift : track ggml release branch (#4867)
469e75d0a35b08de549a4fd87f082ca7a8a539ba 49662cbed3e95f5976c070b85b9fd53fd577038d Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-11T20:43:15+01:00 GitHub noreply@github.com 2024-01-11T21:43:15+02:00 llama : restore intended k-quants mixes for MoE models (#4872)
49662cbed3e95f5976c070b85b9fd53fd577038d 3ba5b8ca8e6181a5c712c5b77595a29f1d3e2b97 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-11T20:39:39+01:00 GitHub noreply@github.com 2024-01-11T21:39:39+02:00 ggml : SOTA 2-bit quants (add IQ2_XS) (#4856)
3ba5b8ca8e6181a5c712c5b77595a29f1d3e2b97 4330bd83feb39683de4bd7a34cfcf672ff8ac3e4 Georgi Gerganov ggerganov@gmail.com 2024-01-11T21:31:31+02:00 GitHub noreply@github.com 2024-01-11T21:31:31+02:00 swift : pin ggml commit + remove ggml.h from spm-headers (#4878)
4330bd83feb39683de4bd7a34cfcf672ff8ac3e4 27379455c38cb13f24de92dbd6fcdd04eeb1b9d9 Laura Tijntje_7@msn.com 2024-01-11T19:02:48+01:00 GitHub noreply@github.com 2024-01-11T20:02:48+02:00 server : implement credentialed CORS (#4514)
27379455c38cb13f24de92dbd6fcdd04eeb1b9d9 eab67950068e4b125007d027232c47d2a5831cd0 Michael Coppola m18coppola@gmail.com 2024-01-11T12:51:17-05:00 GitHub noreply@github.com 2024-01-11T19:51:17+02:00 server : support for multiple api keys (#4864)
eab67950068e4b125007d027232c47d2a5831cd0 d8d90aa343c22fe01429d3540e47ded87e9dcb9d Behnam M 58621210+ibehnam@users.noreply.github.com 2024-01-11T12:41:39-05:00 GitHub noreply@github.com 2024-01-11T19:41:39+02:00 server : add `LOG_INFO` when model is successfully loaded (#4881)
d8d90aa343c22fe01429d3540e47ded87e9dcb9d 43f76bf1c362c067fce46bb8dcda0b64af8a9533 Someone sergei.kozlukov@aalto.fi 2024-01-11T17:22:34Z GitHub noreply@github.com 2024-01-11T17:22:34Z ci: nix-flake-update: new token with pr permissions (#4879)
43f76bf1c362c067fce46bb8dcda0b64af8a9533 2f043328e3116724d15b915b5c6078e2df860a69 pudepiedj pudepiedj@gmail.com 2024-01-11T16:14:52Z GitHub noreply@github.com 2024-01-11T18:14:52+02:00 main : print total token count and tokens consumed so far (#4874)
2f043328e3116724d15b915b5c6078e2df860a69 2a7c94db5fb67b2f8882d2d16a11bf5d8d12d397 Isaac McFadyen isaac@imcf.me 2024-01-11T09:33:26-05:00 GitHub noreply@github.com 2024-01-11T16:33:26+02:00 server : fix typo in model name (#4876)
2a7c94db5fb67b2f8882d2d16a11bf5d8d12d397 64802ec00d6383784a9dacf616095eaced16c3c3 Paul Tsochantaris ptsochantaris@icloud.com 2024-01-11T14:31:52Z GitHub noreply@github.com 2024-01-11T16:31:52+02:00 metal : put encoder debug group behind a define (#4873)
64802ec00d6383784a9dacf616095eaced16c3c3 3267c2abc72e34608224408ace3c048831050f97 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:08+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:08+02:00 sync : ggml
3267c2abc72e34608224408ace3c048831050f97 f85a973aa139ae6f37e8b8e1966f1d278b5e0372 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:34:59+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:05+02:00 metal : fix deprecation warning (ggml/690)
f85a973aa139ae6f37e8b8e1966f1d278b5e0372 5362e43962e84d61e20b91f34991d7ccaef4a7d5 Timothy Cronin 40186632+4imothy@users.noreply.github.com 2024-01-11T02:27:48-05:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:05+02:00 ggml : remove ggml_cpy_inplace and ggml_cont_inplace (ggml/693)
5362e43962e84d61e20b91f34991d7ccaef4a7d5 e739de790921e6abbc8c70398303cacd74913f61 Jack Mousseau jmousseau@users.noreply.github.com 2024-01-10T06:19:19-08:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:05+02:00 metal : wrap each operation in debug group (ggml/690)
e739de790921e6abbc8c70398303cacd74913f61 c910e3c28a1caee8cb1398143d582dd9ab697e68 leejet leejet714@gmail.com 2024-01-10T21:13:42+08:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:05+02:00 ggml : change GGML_MAX_NAME at compile time (ggml/682)
c910e3c28a1caee8cb1398143d582dd9ab697e68 f34432ca1e0b288129390c1db8296a82aaf1e632 Halalaluyafail3 55773281+Halalaluyafail3@users.noreply.github.com 2024-01-09T11:16:37-05:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:05+02:00 Fix execlp call (ggml/689)
f34432ca1e0b288129390c1db8296a82aaf1e632 7a9f75c38b5e62fe27b8a5a3ed823b4a3714024b Erik Scholz Green-Sky@users.noreply.github.com 2024-01-05T16:00:00+01:00 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:39:05+02:00 fix : cuda order of synchronization when setting a buffer (ggml/679)
7a9f75c38b5e62fe27b8a5a3ed823b4a3714024b 5c1980d8d4c4e0c0af77359f81cc44d90b3f250b Behnam M 58621210+ibehnam@users.noreply.github.com 2024-01-11T02:12:05-05:00 GitHub noreply@github.com 2024-01-11T09:12:05+02:00 server : update readme to document the new `/health` endpoint (#4866)
5c1980d8d4c4e0c0af77359f81cc44d90b3f250b cd108e641dbdedd8c5641c4cec1762f751f38136 Georgi Gerganov ggerganov@gmail.com 2024-01-11T09:10:34+02:00 GitHub noreply@github.com 2024-01-11T09:10:34+02:00 server : fix build + rename enums (#4870)
cd108e641dbdedd8c5641c4cec1762f751f38136 57d016ba2d46a6e22517a31a75cebb48f9e234b6 Behnam M 58621210+ibehnam@users.noreply.github.com 2024-01-10T14:56:05-05:00 GitHub noreply@github.com 2024-01-10T21:56:05+02:00 server : add a `/health` endpoint (#4860)
57d016ba2d46a6e22517a31a75cebb48f9e234b6 329ff615699d32f596d4ebf8baba654c30064e0d Brian mofosyne@gmail.com 2024-01-11T01:09:53+11:00 GitHub noreply@github.com 2024-01-10T16:09:53+02:00 llama : add additional suffixes for model params (#4834)
329ff615699d32f596d4ebf8baba654c30064e0d d34633d8db6c2e400355de4862cd699154ecc73f Austin 77757836+teleprint-me@users.noreply.github.com 2024-01-10T08:39:09-05:00 GitHub noreply@github.com 2024-01-10T15:39:09+02:00 llama : recognize 1B phi models (#4847)
d34633d8db6c2e400355de4862cd699154ecc73f 4f56458d34cb13dcbf69aca650e9bf77d5497e6f John 78893154+cmp-nct@users.noreply.github.com 2024-01-10T14:37:09+01:00 GitHub noreply@github.com 2024-01-10T15:37:09+02:00 clip : support more quantization types (#4846)
4f56458d34cb13dcbf69aca650e9bf77d5497e6f 6efb8eb30e7025b168f3fda3ff83b9b386428ad6 Johannes Gäßler johannesg@5d6.de 2024-01-10T01:04:33+01:00 GitHub noreply@github.com 2024-01-10T01:04:33+01:00 Python script to compare commits with llama-bench (#4844)
6efb8eb30e7025b168f3fda3ff83b9b386428ad6 36e5a08b203542dca53cca4eaf172c5dc4bbc991 Austin 77757836+teleprint-me@users.noreply.github.com 2024-01-09T13:46:46-05:00 GitHub noreply@github.com 2024-01-09T20:46:46+02:00 convert.py : fix vanilla LLaMA model conversion (#4818)
36e5a08b203542dca53cca4eaf172c5dc4bbc991 4dccb38d9abab7f9f2d1f9a6977df4185d490132 Justine Tunney jtunney@gmail.com 2024-01-09T09:59:14-08:00 GitHub noreply@github.com 2024-01-09T19:59:14+02:00 llava-cli : don't crash if --image flag is invalid (#4835)
4dccb38d9abab7f9f2d1f9a6977df4185d490132 9a818f7c42761984ac99e08e613cc20634f8410e Georgi Gerganov ggerganov@gmail.com 2024-01-09T19:37:08+02:00 GitHub noreply@github.com 2024-01-09T19:37:08+02:00 metal : improve dequantize precision to match CPU (#4836)
9a818f7c42761984ac99e08e613cc20634f8410e 18adb4e9bb340b7b4565d8b6715b4449283e7641 Georgi Gerganov ggerganov@gmail.com 2024-01-09T19:20:45+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-09T19:21:13+02:00 scripts : improve get-pg.sh (#4838)
18adb4e9bb340b7b4565d8b6715b4449283e7641 d9653894dffbfd3a58616f31b0967b34faf6f611 iohub rickyang.pro@gmail.com 2024-01-10T00:45:54+08:00 GitHub noreply@github.com 2024-01-09T18:45:54+02:00 readme : add 3rd party collama reference to UI list (#4840)
d9653894dffbfd3a58616f31b0967b34faf6f611 128de3585b0f58b1e562733448fc00109f23a95d Georgi Gerganov ggerganov@gmail.com 2024-01-09T16:23:05+02:00 GitHub noreply@github.com 2024-01-09T16:23:05+02:00 scripts : script to get Paul Graham essays in txt format (#4838)
128de3585b0f58b1e562733448fc00109f23a95d 8c5833031857c9e9ada61948bae894ab9c785f86 Behnam M 58621210+ibehnam@users.noreply.github.com 2024-01-09T05:02:05-05:00 GitHub noreply@github.com 2024-01-09T12:02:05+02:00 server : update readme about token probs (#4777)
8c5833031857c9e9ada61948bae894ab9c785f86 18c2e1752c3b387689e9e73d7d8a1a3b1511ce23 Zsapi martin1.zsapka@gmail.com 2024-01-09T10:12:43+01:00 GitHub noreply@github.com 2024-01-09T11:12:43+02:00 server : add api-key flag to documentation (#4832)
18c2e1752c3b387689e9e73d7d8a1a3b1511ce23 8f900abfc09851e281bc9027e0ab2f16bf079b29 Georgi Gerganov ggerganov@gmail.com 2024-01-09T10:42:06+02:00 GitHub noreply@github.com 2024-01-09T10:42:06+02:00 ggml : fix vld1q_s8_x4 32-bit compat (#4828)
8f900abfc09851e281bc9027e0ab2f16bf079b29 1fc2f265ff9377a37fd2c61eae9cd813a3491bea Johannes Gäßler johannesg@5d6.de 2024-01-09T08:58:55+01:00 GitHub noreply@github.com 2024-01-09T08:58:55+01:00 CUDA: faster softmax via shared memory + fp16 math (#4742)
1fc2f265ff9377a37fd2c61eae9cd813a3491bea a9a8c5de3d2028701c239d821b220214fcaefbf1 howlger eclipse@voormann.de 2024-01-08T20:05:53+01:00 GitHub noreply@github.com 2024-01-08T21:05:53+02:00 common : fix the short form of `--grp-attn-w`, not `-gat` (#4825)
a9a8c5de3d2028701c239d821b220214fcaefbf1 dd5ae06405c5565b99889bdb3f168f4351252cfb Georgi Gerganov ggerganov@gmail.com 2024-01-08T20:25:17+02:00 GitHub noreply@github.com 2024-01-08T20:25:17+02:00 readme : add link to SOTA models
dd5ae06405c5565b99889bdb3f168f4351252cfb 668b31fc7d86245435ad6574e0e1126e734049e2 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2024-01-08T16:02:32+01:00 GitHub noreply@github.com 2024-01-08T16:02:32+01:00 SOTA 2-bit quants (#4773)
668b31fc7d86245435ad6574e0e1126e734049e2 42ea63c5a3da01d4a94e906d8565868012c79f4f Georgi Gerganov ggerganov@gmail.com 2024-01-08T16:40:51+02:00 GitHub noreply@github.com 2024-01-08T16:40:51+02:00 swift : exclude ggml-metal.metal from the package (#4822)
42ea63c5a3da01d4a94e906d8565868012c79f4f 52531fdff88764282c1b233174721aab8347252d Georgi Gerganov ggerganov@gmail.com 2024-01-08T15:57:36+02:00 GitHub noreply@github.com 2024-01-08T15:57:36+02:00 llama.swiftui : update readme
52531fdff88764282c1b233174721aab8347252d b0034d93ce2949ce7d9c098ca02e56f66cd484e2 Georgi Gerganov ggerganov@gmail.com 2024-01-08T11:18:32+02:00 GitHub noreply@github.com 2024-01-08T11:18:32+02:00 main : add self-extend support (#4815)
b0034d93ce2949ce7d9c098ca02e56f66cd484e2 b7e7982953f80a656e03feb5cfb17a17a173eb26 Georgi Gerganov ggerganov@gmail.com 2024-01-08T11:14:04+02:00 GitHub noreply@github.com 2024-01-08T11:14:04+02:00 examples : add passkey test (#3856)
b7e7982953f80a656e03feb5cfb17a17a173eb26 226460cc0d5b185bc6685fb76f418fd9418d7add Lars Grammel lars.grammel@gmail.com 2024-01-07T21:24:11+01:00 GitHub noreply@github.com 2024-01-07T22:24:11+02:00 readme : add lgrammel/modelfusion JS/TS client for llama.cpp (#4814)
226460cc0d5b185bc6685fb76f418fd9418d7add d5a410e8556191672465f7ff58682ea2474038b0 slaren slarengh@gmail.com 2024-01-07T17:59:01+01:00 GitHub noreply@github.com 2024-01-07T17:59:01+01:00 llama-bench : add no-kv-offload parameter (#4812)
d5a410e8556191672465f7ff58682ea2474038b0 9dede37d812604897496dd9d276ae9fbe13d1042 Johannes Gäßler johannesg@5d6.de 2024-01-07T17:24:08+01:00 GitHub noreply@github.com 2024-01-07T17:24:08+01:00 CUDA: fixed redundant value dequantization (#4809)
9dede37d812604897496dd9d276ae9fbe13d1042 3c36213df850a2353e95572b3636797c79b7c815 Georgi Gerganov ggerganov@gmail.com 2024-01-07T14:29:36+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-07T14:29:36+02:00 llama : remove unused vars (#4796)
3c36213df850a2353e95572b3636797c79b7c815 72d8407b3696dd1293bd233b6db392be108bc377 Georgi Gerganov ggerganov@gmail.com 2024-01-07T11:21:53+02:00 GitHub noreply@github.com 2024-01-07T11:21:53+02:00 llama : remove redundant GQA check (#4796)
72d8407b3696dd1293bd233b6db392be108bc377 d117d4dc5dadb46831036bfa4d6e5e8c86babaf1 Alex Azarov alexander.azarov@mapbox.com 2024-01-07T09:20:50+01:00 GitHub noreply@github.com 2024-01-07T10:20:50+02:00 llama.swiftui : use llama.cpp as SPM package (#4804)
d117d4dc5dadb46831036bfa4d6e5e8c86babaf1 3418c03ecc149fd657527ebb06776239b60a3f3b Georgi Gerganov ggerganov@gmail.com 2024-01-07T09:50:31+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-07T09:51:12+02:00 llama : print tensor meta for debugging
3418c03ecc149fd657527ebb06776239b60a3f3b 63ee677efd92060b14894b984597c62e3742b8da Alex Azarov alexander.azarov@mapbox.com 2024-01-07T08:46:55+01:00 GitHub noreply@github.com 2024-01-07T09:46:55+02:00 llama.swiftui : add visionOS target (#4805)
63ee677efd92060b14894b984597c62e3742b8da 67984921a70a7e680a24494aeb7575a66e90685d Konstantin Zhuravlyov konstantin.zhuravlyov@amd.com 2024-01-07T01:52:42-05:00 GitHub noreply@github.com 2024-01-07T08:52:42+02:00 ggml : use __builtin_amdgcn_sudot4 in __dp4a for gfx11 (#4787)
67984921a70a7e680a24494aeb7575a66e90685d c75ca5d96f902564cbbbdd7f5cade80d53c288bb Georgi Gerganov ggerganov@gmail.com 2024-01-07T08:45:26+02:00 GitHub noreply@github.com 2024-01-07T08:45:26+02:00 server : fix n_predict check (#4798)
c75ca5d96f902564cbbbdd7f5cade80d53c288bb 96e80dabc6e73ff68b09b68947b1fc25883c5094 Daniel Illescas Romero illescas.daniel@protonmail.com 2024-01-06T16:12:59+01:00 GitHub noreply@github.com 2024-01-06T17:12:59+02:00 llama.swiftui : use correct pointer for llama_token_eos (#4797)
96e80dabc6e73ff68b09b68947b1fc25883c5094 eec22a1c6378d9a013943cbddb4330c0da621442 Georgi Gerganov ggerganov@gmail.com 2024-01-06T11:40:24+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-06T11:40:24+02:00 examples : improve base-translate.sh script (#4783)
eec22a1c6378d9a013943cbddb4330c0da621442 be36bb946a6336238e92706464de6a30495fe825 a-n-n-a-l-e-e 150648636+a-n-n-a-l-e-e@users.noreply.github.com 2024-01-05T08:04:40-08:00 GitHub noreply@github.com 2024-01-05T18:04:40+02:00 cmake : check for openblas64 (#4134)
be36bb946a6336238e92706464de6a30495fe825 91d38876dfa10332359ac671b62353aeceb448d3 Ikko Eltociear Ashimine eltociear@gmail.com 2024-01-06T01:02:44+09:00 GitHub noreply@github.com 2024-01-05T18:02:44+02:00 flake.nix : fix typo (#4700)
91d38876dfa10332359ac671b62353aeceb448d3 d061bf9405cc5fd50792fb2dbdff9c9ea53d6bf9 Georgi Gerganov ggerganov@gmail.com 2024-01-05T16:30:52+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-05T18:02:06+02:00 metal : switch back to default.metallib (ggml/681)
d061bf9405cc5fd50792fb2dbdff9c9ea53d6bf9 1bf681f90ef4cf37b36e6d604d3e30fc57eda650 Georgi Gerganov ggerganov@gmail.com 2024-01-05T15:36:04+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-05T18:02:06+02:00 ggml : fix q2_k bpw in comments (ggml/680)
1bf681f90ef4cf37b36e6d604d3e30fc57eda650 c1d7cb28d3fed97fbc95fc3c43f0c5e2113e546c Finn Voorhees finnvoorhees@gmail.com 2024-01-03T08:39:43-05:00 Georgi Gerganov ggerganov@gmail.com 2024-01-05T18:02:06+02:00 ggml : add error handling to graph_compute (whisper/1714)
c1d7cb28d3fed97fbc95fc3c43f0c5e2113e546c 3681f22443d917e7328456b69c276d6927dafeec Georgi Gerganov ggerganov@gmail.com 2024-01-05T15:18:21+02:00 GitHub noreply@github.com 2024-01-05T15:18:21+02:00 ggml : do not sched_yield when calling BLAS (#4761)
3681f22443d917e7328456b69c276d6927dafeec b3a7c20b5c035250257d2b62851c379b159c899a Georgi Gerganov ggerganov@gmail.com 2024-01-05T15:11:10+02:00 GitHub noreply@github.com 2024-01-05T15:11:10+02:00 examples : add few-shot translation example (#4783)
b3a7c20b5c035250257d2b62851c379b159c899a 012cf349aec8ffb47c9def5dc018240fa3721e8b Daniel Bevenius daniel.bevenius@gmail.com 2024-01-04T20:45:37+01:00 GitHub noreply@github.com 2024-01-04T21:45:37+02:00 finetune : remove unused includes (#4756)
012cf349aec8ffb47c9def5dc018240fa3721e8b a91928014fcf51fe297823fcff0788de4f14206e Georgi Gerganov ggerganov@gmail.com 2024-01-04T19:56:33+02:00 GitHub noreply@github.com 2024-01-04T19:56:33+02:00 server : send token probs for "stream == false" (#4714)
a91928014fcf51fe297823fcff0788de4f14206e 3c0b585561d74a56977cf3a3844535ecc9e37972 Johannes Gäßler johannesg@5d6.de 2024-01-04T09:43:23+01:00 GitHub noreply@github.com 2024-01-04T09:43:23+01:00 Print backend name on test-backend-ops failure (#4751)
3c0b585561d74a56977cf3a3844535ecc9e37972 e5804313a1edaf00726ed0b96ecced07accbf50c singularity 12184989+singularity-s0@users.noreply.github.com 2024-01-04T16:22:38+08:00 GitHub noreply@github.com 2024-01-04T10:22:38+02:00 llama.swiftui : support loading custom model from file picker (#4767)
e5804313a1edaf00726ed0b96ecced07accbf50c dc891b7f7a23158d54f9383790b92c79cc5906c1 Michael Coppola m18coppola@gmail.com 2024-01-04T03:17:09-05:00 GitHub noreply@github.com 2024-01-04T10:17:09+02:00 server : fix options in README.md (#4765)
dc891b7f7a23158d54f9383790b92c79cc5906c1 46cea79e1f32499bb24b9fab12123cd386e96728 Georgi Gerganov ggerganov@gmail.com 2024-01-04T10:12:26+02:00 GitHub noreply@github.com 2024-01-04T10:12:26+02:00 ggml : include stdlib.h before intrin.h (#4736)
46cea79e1f32499bb24b9fab12123cd386e96728 cb1e2818e0e12ec99f7236ec5d4f3ffd8bcc2f4a singularity 12184989+singularity-s0@users.noreply.github.com 2024-01-04T15:58:16+08:00 GitHub noreply@github.com 2024-01-04T09:58:16+02:00 llama.swiftui : fix build of ggml.metallib (#4754)
cb1e2818e0e12ec99f7236ec5d4f3ffd8bcc2f4a ece9a45e8ffb73ad461c792720c2fec28b0137bc Daniel Bevenius daniel.bevenius@gmail.com 2024-01-03T18:53:40+01:00 GitHub noreply@github.com 2024-01-03T19:53:40+02:00 train : fix typo in overlapping-samples help msg (#4758)
ece9a45e8ffb73ad461c792720c2fec28b0137bc 7bed7eba359b0fa8e575345dc5467a46b4ba509f Ashraful Islam ashraful.meche@gmail.com 2024-01-03T11:30:02-06:00 GitHub noreply@github.com 2024-01-03T19:30:02+02:00 swift : update Package.swift to use ggml as dependency (#4691)
7bed7eba359b0fa8e575345dc5467a46b4ba509f d55356d3baa58a6c3a9171cb67a67094b9aa9dff Georgi Gerganov ggerganov@gmail.com 2024-01-03T14:18:46+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-03T14:38:38+02:00 cuda : simplify expression
d55356d3baa58a6c3a9171cb67a67094b9aa9dff 75e3fd85814c367b55aea11e7bb38cb7b82c6aa0 Georgi Gerganov ggerganov@gmail.com 2024-01-03T13:01:44+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-03T14:38:38+02:00 cuda : mark I16 and I32 ops as unsupported
75e3fd85814c367b55aea11e7bb38cb7b82c6aa0 289313716ff7ccf6aee284f686a0fe8cbc7714af Georgi Gerganov ggerganov@gmail.com 2024-01-03T11:37:44+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-03T14:38:38+02:00 sync : ggml
289313716ff7ccf6aee284f686a0fe8cbc7714af ab62fc3e5520f5a143c36cb23c269f11aa4dafd6 Georgi Gerganov ggerganov@gmail.com 2024-01-03T11:35:46+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-03T14:38:38+02:00 metal : add kernel_get_rows_i32
ab62fc3e5520f5a143c36cb23c269f11aa4dafd6 5f66ebca9c41a17385341da4b553a8eb5f07edee Georgi Gerganov ggerganov@gmail.com 2024-01-03T11:25:54+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-03T14:38:38+02:00 scripts : fix sync order + metal sed
5f66ebca9c41a17385341da4b553a8eb5f07edee f2eb19bd8bc9f5730d6e05d7a52a9e19bf5ac099 Guillaume Wenzek gwenzek@users.noreply.github.com 2023-12-29T18:07:03+01:00 Georgi Gerganov ggerganov@gmail.com 2024-01-03T14:38:38+02:00 ggml : extend ggml_get_rows, ggml_repeat, ggml_concat (ggml/639)
f2eb19bd8bc9f5730d6e05d7a52a9e19bf5ac099 f3f62f0d835d559e80714bbeb05d03125574e3dd Justin Parker jparkerweb@gmail.com 2024-01-03T03:43:19-05:00 GitHub noreply@github.com 2024-01-03T10:43:19+02:00 server : throw an error when `slot unavailable` (#4741)
f3f62f0d835d559e80714bbeb05d03125574e3dd 0ef3ca2ac62016c0c545de1c89dc2e3e130f4a99 Georgi Gerganov ggerganov@gmail.com 2024-01-02T21:07:47+02:00 GitHub noreply@github.com 2024-01-02T21:07:47+02:00 metal : optimize ggml_mul_mat_id (faster Mixtral PP) (#4725)
0ef3ca2ac62016c0c545de1c89dc2e3e130f4a99 540938f8904707dd74cb3be18495f853b312e72f Phil H 5756783+phiharri@users.noreply.github.com 2024-01-02T15:48:49Z GitHub noreply@github.com 2024-01-02T17:48:49+02:00 server : add token counts to html footer (#4738)
540938f8904707dd74cb3be18495f853b312e72f 0040d42eeb237197054cc7790df5776eacfa608e Georgi Gerganov ggerganov@gmail.com 2024-01-02T16:26:45+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-02T16:26:45+02:00 llama : llama_model_desc print number of experts
0040d42eeb237197054cc7790df5776eacfa608e 83e633c27efdf0eb0ba54249e784b0ea760b1007 Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2024-01-02T06:15:16-08:00 GitHub noreply@github.com 2024-01-02T16:15:16+02:00 llama : replace all API facing `int`'s with `int32_t` (#4577)
83e633c27efdf0eb0ba54249e784b0ea760b1007 32866c5edde402f42ff4233bb89dcfcede34fd22 postmasters namnguyen@google.com 2024-01-02T03:51:28-08:00 GitHub noreply@github.com 2024-01-02T13:51:28+02:00 llama : differentiate the KV dims in the attention (#4657)
32866c5edde402f42ff4233bb89dcfcede34fd22 5d7002d4372ebf107cfaf46fcd90df27b204f330 Georgi Gerganov ggerganov@gmail.com 2024-01-02T13:28:15+02:00 Georgi Gerganov ggerganov@gmail.com 2024-01-02T13:28:15+02:00 editorconfig : fix whitespace and indentation #4710
5d7002d4372ebf107cfaf46fcd90df27b204f330 26f3071d714f0b27ad7f021a46a66a1085480258 minarchist minarchist@users.noreply.github.com 2024-01-02T04:38:15-06:00 GitHub noreply@github.com 2024-01-02T12:38:15+02:00 server : add --override-kv parameter (#4710)
26f3071d714f0b27ad7f021a46a66a1085480258 775ac8712a7b42cfead2585f42cec0dfd56644ab Nam D. Tran 42194884+namtranase@users.noreply.github.com 2024-01-02T16:23:38+07:00 GitHub noreply@github.com 2024-01-02T11:23:38+02:00 py : re-enable mmap in convert hf (#4732)
775ac8712a7b42cfead2585f42cec0dfd56644ab 58ba655af054715c0516ee270ad028ad9e74f357 Daniel Bevenius daniel.bevenius@gmail.com 2024-01-02T10:16:55+01:00 GitHub noreply@github.com 2024-01-02T10:16:55+01:00 finetune: fix typo in README.md (#4733)
58ba655af054715c0516ee270ad028ad9e74f357 edd1ab7bc34c10a780ee7f9a4499f7689cdad36d Georgi Gerganov ggerganov@gmail.com 2024-01-02T10:57:44+02:00 GitHub noreply@github.com 2024-01-02T10:57:44+02:00 metal : enable shader debugging (cmake option) (#4705)
edd1ab7bc34c10a780ee7f9a4499f7689cdad36d 198ed7ebfc89b8f2b35a8b1655d57bfb57530c1a Someone Serge sergei.kozlukov@aalto.fi 2023-12-31T17:42:22Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 flake.lock: update
198ed7ebfc89b8f2b35a8b1655d57bfb57530c1a d8361747317c5cb2e00e7fb3b59ff4dce5a176a5 Someone Serge sergei.kozlukov@aalto.fi 2023-12-30T18:25:25Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 flake.nix: suggest the binary caches
d8361747317c5cb2e00e7fb3b59ff4dce5a176a5 06f2a5d1909a1385b1a16dab4ade68377e121bdd Someone Serge sergei.kozlukov@aalto.fi 2023-12-30T18:01:07Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 workflows: nix-ci: add a qemu job for jetsons
06f2a5d1909a1385b1a16dab4ade68377e121bdd c5239944bab0ff71915df8f2dc7e42fc2c138ff6 Someone Serge sergei.kozlukov@aalto.fi 2023-12-30T17:36:08Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 workflows: nix-flakestry: drop tag filters
c5239944bab0ff71915df8f2dc7e42fc2c138ff6 1e9ae54cf24d27afe3900d1250634a2a33423db1 Someone Serge sergei.kozlukov@aalto.fi 2023-12-30T16:38:36Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 workflows: weekly `nix flake update`
1e9ae54cf24d27afe3900d1250634a2a33423db1 7adedecbe39bd552bc14142f496246d55a43ac4e Someone Serge sergei.kozlukov@aalto.fi 2023-12-30T17:19:11Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 workflows: nix-ci: add a job for eval
7adedecbe39bd552bc14142f496246d55a43ac4e 356ea17e0f92bfbbf28a4f69261bed48eff68d9c Someone Serge sergei.kozlukov@aalto.fi 2023-12-26T19:17:26Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 workflows: nix-ci: init; build flake outputs
356ea17e0f92bfbbf28a4f69261bed48eff68d9c a5c088d8c698299b973d2709153e5d95295606d9 Someone Serge sergei.kozlukov@aalto.fi 2023-12-29T16:21:50Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 flake.nix: expose checks
a5c088d8c698299b973d2709153e5d95295606d9 1e3900ebacb3a0b385271389686403c97ad76d88 Someone Serge sergei.kozlukov@aalto.fi 2023-12-26T23:34:40Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 flake.nix: rocm not yet supported on aarch64, so hide the output
1e3900ebacb3a0b385271389686403c97ad76d88 e39106c0554cbd0e9310e08fb3b2a577ea4b6273 Someone Serge sergei.kozlukov@aalto.fi 2023-12-29T16:15:37Z Philip Taron philip.taron@gmail.com 2023-12-31T13:14:58-08:00 flake.nix: expose full scope in legacyPackages
e39106c0554cbd0e9310e08fb3b2a577ea4b6273 9fbda719de18a9400a064c28759c39d55d687d3e Georgi Gerganov ggerganov@gmail.com 2023-12-31T11:43:31+02:00 GitHub noreply@github.com 2023-12-31T11:43:31+02:00 ggml : add ggml_vdotq_s32 alias (#4715)
9fbda719de18a9400a064c28759c39d55d687d3e 39d8bc71edcb8b6f99d46fa4216af7a15232e218 Georgi Gerganov ggerganov@gmail.com 2023-12-30T23:24:42+02:00 GitHub noreply@github.com 2023-12-30T23:24:42+02:00 clip : refactor + bug fixes (#4696)
39d8bc71edcb8b6f99d46fa4216af7a15232e218 24a447e20af425fa44cf10feaa632b6bb596c80f Johannes Gäßler johannesg@5d6.de 2023-12-30T13:52:01+01:00 GitHub noreply@github.com 2023-12-30T13:52:01+01:00 CUDA: fixed tensor cores not being used on RDNA3 (#4697)
24a447e20af425fa44cf10feaa632b6bb596c80f a20f3c7465d6d1b33767757c2760643b799a81bf automaticcat daogiatuank54@gmail.com 2023-12-30T15:07:48+07:00 GitHub noreply@github.com 2023-12-30T10:07:48+02:00 ggml : add ggml_cpu_has_avx_vnni() (#4589)
a20f3c7465d6d1b33767757c2760643b799a81bf 0235b9b571f3cc7d2b8836409a5404b41ce1379c Johannes Gäßler johannesg@5d6.de 2023-12-29T23:12:53+01:00 GitHub noreply@github.com 2023-12-29T23:12:53+01:00 CUDA: fix tensor core logic for Pascal and HIP (#4682)
0235b9b571f3cc7d2b8836409a5404b41ce1379c ce18d727a47f2473ca863a6f78bf3ad480008f72 Georgi Gerganov ggerganov@gmail.com 2023-12-29T18:53:34+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-29T18:53:34+02:00 clip : use ggml_backend_buffer_is_host (#4205)
ce18d727a47f2473ca863a6f78bf3ad480008f72 91bb39cec7e4dfb9e2293509ef60298a67f0b1b7 Steward Garcia 57494570+FSSRepo@users.noreply.github.com 2023-12-29T11:52:15-05:00 GitHub noreply@github.com 2023-12-29T18:52:15+02:00 clip : enable gpu backend (#4205)
91bb39cec7e4dfb9e2293509ef60298a67f0b1b7 04ac0607e913ab91234dfb240e12a76509e30982 hydai z54981220@gmail.com 2023-12-30T00:31:19+08:00 GitHub noreply@github.com 2023-12-29T17:31:19+01:00 cuda: fix vmm oom issue on NVIDIA AGX Orin (#4687)
04ac0607e913ab91234dfb240e12a76509e30982 68eccbdc5b56f2a2450f9a8463f9934388cafabf crasm crasm@git.vczf.us 2023-12-29T09:50:29-05:00 GitHub noreply@github.com 2023-12-29T16:50:29+02:00 python : add check-requirements.sh and GitHub workflow (#4585)
68eccbdc5b56f2a2450f9a8463f9934388cafabf 97bbca6e8522d18041fcde6c3d0907a52ce36446 Philip Taron philip.taron@gmail.com 2023-12-29T06:42:26-08:00 GitHub noreply@github.com 2023-12-29T16:42:26+02:00 flake.nix : rewrite (#4605)
97bbca6e8522d18041fcde6c3d0907a52ce36446 4af4801566bc262a38fb77f51edf278ac323c2bd Cuong Trinh Manh nguoithichkhampha@gmail.com 2023-12-29T21:39:15+07:00 GitHub noreply@github.com 2023-12-29T16:39:15+02:00 cmake : fix ld warning duplicate libraries libllama.a (#4671)
4af4801566bc262a38fb77f51edf278ac323c2bd db49ff8ed7f0bb201176703441cc02911b08ef2a Justine Tunney jtunney@gmail.com 2023-12-29T06:38:38-08:00 GitHub noreply@github.com 2023-12-29T16:38:38+02:00 llava-cli : refactor to use sampling library (#4669)
db49ff8ed7f0bb201176703441cc02911b08ef2a 60f55e888c29cbd87c4238dd19e85d0eef87245d Justine Tunney jtunney@gmail.com 2023-12-29T06:24:12-08:00 GitHub noreply@github.com 2023-12-29T16:24:12+02:00 server : replace sleep with condition variables (#4673)
60f55e888c29cbd87c4238dd19e85d0eef87245d b93edd22f55d3e5268263c3edcdae1818505c078 SakuraUmi yukinon244@gmail.com 2023-12-29T22:22:44+08:00 GitHub noreply@github.com 2023-12-29T16:22:44+02:00 server : fix OpenAI server sampling w.r.t. penalty. (#4675)
b93edd22f55d3e5268263c3edcdae1818505c078 82d6eab224862a7044069fb9211dc4b29124264b Karthik Sethuraman k.seth1993@gmail.com 2023-12-29T06:22:10-08:00 GitHub noreply@github.com 2023-12-29T16:22:10+02:00 server : allow to generate multimodal embeddings (#4681)
82d6eab224862a7044069fb9211dc4b29124264b afd997ab6011dfefe9e917425b04ef4d83614841 andrijdavid david@geek.mg 2023-12-29T15:18:20+01:00 GitHub noreply@github.com 2023-12-29T16:18:20+02:00 main-cmake-pkg : fix build issue (#4665)
afd997ab6011dfefe9e917425b04ef4d83614841 c8255f8a6b2a3b3ebc6cb340cc2487f39fc95ffc Peter Sugihara peter@campsh.com 2023-12-29T05:58:56-08:00 GitHub noreply@github.com 2023-12-29T15:58:56+02:00 llama.swiftui : fix infinite loop, ouput timings, buff UI (#4674)
c8255f8a6b2a3b3ebc6cb340cc2487f39fc95ffc 441f51dca004debf8b275f1bdc08e0f1af7fd8f8 Georgi Gerganov ggerganov@gmail.com 2023-12-29T15:12:35+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-29T15:12:35+02:00 scripts : print list of sync commits
441f51dca004debf8b275f1bdc08e0f1af7fd8f8 38b3de4658292582a8941a2be5c77b40ce6ac0f2 Tamotsu Takahashi ttakah+github@gmail.com 2023-12-29T19:23:27+09:00 Georgi Gerganov ggerganov@gmail.com 2023-12-29T15:11:53+02:00 ci : build with CLBlast + ggml-opencl use GGML_API (whisper/1576)
38b3de4658292582a8941a2be5c77b40ce6ac0f2 afc8c192919f04613a92d40391bff4c8cd99856b Georgi Gerganov ggerganov@gmail.com 2023-12-29T14:56:41+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-29T14:56:41+02:00 sync : ggml
afc8c192919f04613a92d40391bff4c8cd99856b ca38b8d334baa724bd6c9402470931d26427466f bssrdf merlintiger@hotmail.com 2023-12-29T03:32:31-05:00 Georgi Gerganov ggerganov@gmail.com 2023-12-29T14:54:19+02:00 ggml : fix some mul mat cases + add tests for src1 F16 (ggml/669)
ca38b8d334baa724bd6c9402470931d26427466f 65e5f6dadbba4b496bba27f573e473c66b446496 Georgi Gerganov ggerganov@gmail.com 2023-12-29T14:41:36+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-29T14:54:05+02:00 scripts : do not sync commits from this repo
65e5f6dadbba4b496bba27f573e473c66b446496 ea5497df5d138c83b2b0ca70aefdc4b1175c1001 Justine Tunney jtunney@gmail.com 2023-12-28T11:20:00-08:00 GitHub noreply@github.com 2023-12-28T15:20:00-04:00 Fix OpenAI server sampling w.r.t. temp and seed (#4668)
ea5497df5d138c83b2b0ca70aefdc4b1175c1001 f6793491b5af6da75edad34d6f503ef86d31b09f manikbhandari mbbhandarimanik2@gmail.com 2023-12-28T09:03:57-05:00 GitHub noreply@github.com 2023-12-28T15:03:57+01:00 gpt2 : Add gpt2 architecture integration (#4555)
f6793491b5af6da75edad34d6f503ef86d31b09f 879b690a9e1eb1ab0a29b58236fc76978fb4d902 Nam D. Tran 42194884+namtranase@users.noreply.github.com 2023-12-27T22:39:45+07:00 GitHub noreply@github.com 2023-12-27T17:39:45+02:00 llama : add AWQ for llama, llama2, mpt, and mistral models (#4593)
879b690a9e1eb1ab0a29b58236fc76978fb4d902 b47879b0dda43f2d26415e88b6840295817e552a Daniel Bevenius daniel.bevenius@gmail.com 2023-12-27T15:16:55+01:00 GitHub noreply@github.com 2023-12-27T16:16:55+02:00 finetune : fix output formatting in print_params (#4653)
b47879b0dda43f2d26415e88b6840295817e552a 951010fa53a0ffe81b7d2e87c4349e0d3cb3d19d Georgi Gerganov ggerganov@gmail.com 2023-12-27T11:15:31+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-27T11:44:22+02:00 scripts : add sync-ggml-am.sh
951010fa53a0ffe81b7d2e87c4349e0d3cb3d19d f56d6077d0c37e6606ac0a4fa3169de70593acfe Georgi Gerganov ggerganov@gmail.com 2023-12-27T11:02:13+02:00 GitHub noreply@github.com 2023-12-27T11:02:13+02:00 ggml : fix dot product for ARM (#4630)
f56d6077d0c37e6606ac0a4fa3169de70593acfe dc68f0054cd279cddddb0cae0c9ef4f9cbaa512a wonjun Jang strutive07@gmail.com 2023-12-27T17:37:25+09:00 GitHub noreply@github.com 2023-12-27T17:37:25+09:00 Add byte token type when tokenizer.model is not exists (#4641)
dc68f0054cd279cddddb0cae0c9ef4f9cbaa512a de8e496437c59e7d1cc84109e3e49a3478aee25a slaren slarengh@gmail.com 2023-12-26T21:23:59+01:00 GitHub noreply@github.com 2023-12-26T21:23:59+01:00 cuda : fix vmm pool with multi GPU (#4620)
de8e496437c59e7d1cc84109e3e49a3478aee25a 77465dad48d7c945c367ab46b6f2ea98ae9b7b15 WillCorticesAI 150854901+WillCorticesAI@users.noreply.github.com 2023-12-26T05:42:08-05:00 GitHub noreply@github.com 2023-12-26T11:42:08+01:00 Update comment for AdamW implementation reference. (#4604)
77465dad48d7c945c367ab46b6f2ea98ae9b7b15 a206137f927daef1752753cf5e281220b449a468 FantasyGmm 16450052+FantasyGmm@users.noreply.github.com 2023-12-26T18:38:36+08:00 GitHub noreply@github.com 2023-12-26T11:38:36+01:00 Fix new CUDA10 compilation errors (#4635)
a206137f927daef1752753cf5e281220b449a468 b9f47952ffae4e0d3420905526003c23333f6c98 Paul Tsochantaris ptsochantaris@icloud.com 2023-12-25T16:09:53Z GitHub noreply@github.com 2023-12-25T18:09:53+02:00 Adding Emeltal reference to UI list (#4629)
b9f47952ffae4e0d3420905526003c23333f6c98 753be377b69bda2d65a7e089f2b7f0c53ef3495e slaren slarengh@gmail.com 2023-12-24T21:01:12+01:00 GitHub noreply@github.com 2023-12-24T22:01:12+02:00 simplify bug issue template (#4623)
753be377b69bda2d65a7e089f2b7f0c53ef3495e 5bf3953d7e9831ea22b0bc017ce97409b801ccf1 Shintarou Okada kokuzen@gmail.com 2023-12-24T22:35:49+09:00 GitHub noreply@github.com 2023-12-24T15:35:49+02:00 llama : add PLaMo model (#3557)
5bf3953d7e9831ea22b0bc017ce97409b801ccf1 708e179e8562c2604240df95a2241dea17fd808b slaren slarengh@gmail.com 2023-12-24T14:34:22+01:00 GitHub noreply@github.com 2023-12-24T14:34:22+01:00 cuda : improve cuda pool efficiency using virtual memory (#4606)
708e179e8562c2604240df95a2241dea17fd808b 925e5584a058afb612f9c20bc472c130f5d0f891 slaren slarengh@gmail.com 2023-12-23T16:10:51+01:00 GitHub noreply@github.com 2023-12-23T16:10:51+01:00 fallback to CPU buffer if host buffer alloc fails (#4610)
925e5584a058afb612f9c20bc472c130f5d0f891 6123979952385847d8348e295d77d6e01da8aa84 Samuel Maynard samwmaynard@gmail.com 2023-12-23T11:35:55+02:00 GitHub noreply@github.com 2023-12-23T11:35:55+02:00 ci(docker): fix tags in "Build and push docker image (tagged)" (#4603)
6123979952385847d8348e295d77d6e01da8aa84 b9ec82d262cb20d7f0a8a1157bfa9aace40e2625 Alexey Parfenov zxed@alkatrazstudio.net 2023-12-23T09:31:49Z GitHub noreply@github.com 2023-12-23T11:31:49+02:00 server : allow to specify custom prompt for penalty calculation (#3727)
b9ec82d262cb20d7f0a8a1157bfa9aace40e2625 e0a4002273907b2c414b6b5442d99e08bfe2df35 kalomaze 66376113+kalomaze@users.noreply.github.com 2023-12-23T03:27:07-06:00 GitHub noreply@github.com 2023-12-23T11:27:07+02:00 grammar : check the full vocab only if necessary (opt) (#4306)
e0a4002273907b2c414b6b5442d99e08bfe2df35 7082d24cec35e9ce9147535a2224dfc67ee0a78c Johannes Gäßler johannesg@5d6.de 2023-12-23T09:16:33+01:00 GitHub noreply@github.com 2023-12-23T09:16:33+01:00 CUDA: fixed row rounding for 0 tensor splits (#4594)
7082d24cec35e9ce9147535a2224dfc67ee0a78c ba661751322a7c201fd3bef71af077c5aebfaa2a LeonEricsson 70749762+LeonEricsson@users.noreply.github.com 2023-12-22T17:05:56+01:00 GitHub noreply@github.com 2023-12-22T18:05:56+02:00 lookup : add prompt lookup decoding example (#4484)
ba661751322a7c201fd3bef71af077c5aebfaa2a a55876955b1a83464171de8d578d3ab062a7b62d Georgi Gerganov ggerganov@gmail.com 2023-12-22T17:53:43+02:00 GitHub noreply@github.com 2023-12-22T17:53:43+02:00 sync : ggml (fix im2col) (#4591)
a55876955b1a83464171de8d578d3ab062a7b62d 6724ef16573ec7ecce620be56cbbff145856b2fb FantasyGmm 16450052+FantasyGmm@users.noreply.github.com 2023-12-22T23:11:12+08:00 GitHub noreply@github.com 2023-12-22T17:11:12+02:00 cuda : fix jetson compile error (#4560)
6724ef16573ec7ecce620be56cbbff145856b2fb 48b7ff193e64c97ab174280ba0eb8d14b47c49ba Henrik Forstén henrik.forsten@gmail.com 2023-12-22T15:34:05+02:00 GitHub noreply@github.com 2023-12-22T14:34:05+01:00 Fix CudaMemcpy direction (#4599)
48b7ff193e64c97ab174280ba0eb8d14b47c49ba 48b24b170e3b4f9dc28200306840cb07d1c123df slaren slarengh@gmail.com 2023-12-22T12:12:53+01:00 GitHub noreply@github.com 2023-12-22T13:12:53+02:00 llama : fix platforms without mmap (#4578)
48b24b170e3b4f9dc28200306840cb07d1c123df 28cb35a0ecb9852adc3494aa51dde60141939d64 Herman Semenov GermanAizek@yandex.ru 2023-12-22T09:26:49Z GitHub noreply@github.com 2023-12-22T11:26:49+02:00 ggml : add comment about backward GGML_OP_DIAG_MASK_INF (#4203)
28cb35a0ecb9852adc3494aa51dde60141939d64 f31b98489824a86c937fa62ccf5dfd4bb0327b86 Michael Kesper mkesper@schokokeks.org 2023-12-22T09:03:25+01:00 GitHub noreply@github.com 2023-12-22T10:03:25+02:00 make : add LLAMA_HIP_UMA option (#4587)
f31b98489824a86c937fa62ccf5dfd4bb0327b86 2bb98279c5a087d62949972b35cf63ff974ffe6a rhuddleston ryan.huddleston@percona.com 2023-12-21T23:56:34-07:00 GitHub noreply@github.com 2023-12-22T08:56:34+02:00 ci : tag docker image with build number (#4584)
2bb98279c5a087d62949972b35cf63ff974ffe6a 0137ef88ea9f8fd837a065700814329d24adeec3 Deins deinsegle@gmail.com 2023-12-22T08:49:54+02:00 GitHub noreply@github.com 2023-12-22T08:49:54+02:00 readme : add zig bindings (#4581)
0137ef88ea9f8fd837a065700814329d24adeec3 c7e9701f86564088350209d2f9d71c96ea00527f bobqianic 129547291+bobqianic@users.noreply.github.com 2023-12-22T06:47:01Z GitHub noreply@github.com 2023-12-22T08:47:01+02:00 ggml : extend `enum ggml_log_level` with `GGML_LOG_LEVEL_DEBUG` (#4579)
c7e9701f86564088350209d2f9d71c96ea00527f afefa319f1f59b002dfa0d1ef407a2c74bd9770b crasm crasm@git.vczf.us 2023-12-22T01:19:36-05:00 GitHub noreply@github.com 2023-12-22T08:19:36+02:00 llama : add ability to cancel model loading (#4462)
afefa319f1f59b002dfa0d1ef407a2c74bd9770b 769a7bc85eaa44e3d7eadf39abfeff7bb0b9cc2f Georgi Gerganov ggerganov@gmail.com 2023-12-21T23:20:49+02:00 GitHub noreply@github.com 2023-12-21T23:20:49+02:00 ggml : change ggml_scale to take a float instead of tensor (#4573)
769a7bc85eaa44e3d7eadf39abfeff7bb0b9cc2f 32259b2dade6f6856739bf7ba0a4ff7b474dc760 Georgi Gerganov ggerganov@gmail.com 2023-12-21T23:20:36+02:00 GitHub noreply@github.com 2023-12-21T23:20:36+02:00 gguf-py : fix broken link
32259b2dade6f6856739bf7ba0a4ff7b474dc760 4a5f9d629ecfd0a53afdddbaf54a4fa02d9a9ce9 Georgi Gerganov ggerganov@gmail.com 2023-12-21T23:07:58+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-21T23:08:14+02:00 gguf : simplify example dependencies
4a5f9d629ecfd0a53afdddbaf54a4fa02d9a9ce9 d232aca5a73b290e218a2e48b91023d5e994203f Samuel Maynard samwmaynard@gmail.com 2023-12-21T22:36:26+02:00 GitHub noreply@github.com 2023-12-21T22:36:26+02:00 ci : add `jlumbroso/free-disk-space` to docker workflow (#4150)
d232aca5a73b290e218a2e48b91023d5e994203f 31f27758faf4a4bd08101a57c7ec3a473f771f86 slaren slarengh@gmail.com 2023-12-21T21:07:46+01:00 GitHub noreply@github.com 2023-12-21T21:07:46+01:00 llama : initial ggml-backend integration (#4520)
31f27758faf4a4bd08101a57c7ec3a473f771f86 56fa50819f7a3ca2128f63b81c17c08a4454479e Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2023-12-21T11:57:48-08:00 GitHub noreply@github.com 2023-12-21T21:57:48+02:00 llama : allow getting n_batch from llama_context in c api (#4540)
56fa50819f7a3ca2128f63b81c17c08a4454479e 0f630fbc924aaabeea6eaf466bb4b47d13015c3e Finn Voorhees finnvoorhees@gmail.com 2023-12-21T14:55:02-05:00 GitHub noreply@github.com 2023-12-21T21:55:02+02:00 metal : fix `ggml_metal_log` vargs (#4373)
0f630fbc924aaabeea6eaf466bb4b47d13015c3e 562cf222b5129e40b312877e928eac3a02e4ec33 Erik Garrison erik.garrison@gmail.com 2023-12-21T13:45:32-06:00 GitHub noreply@github.com 2023-12-21T21:45:32+02:00 cuda : ROCm AMD Unified Memory Architecture (UMA) handling (#4449)
562cf222b5129e40b312877e928eac3a02e4ec33 8fe03ffddaaa0ab5d48feaafe398151c9f22d4f6 arlo-phoenix 140345165+arlo-phoenix@users.noreply.github.com 2023-12-21T20:13:25+01:00 GitHub noreply@github.com 2023-12-21T20:13:25+01:00 ggml-cuda: Fix HIP build by adding define for __trap (#4569)
8fe03ffddaaa0ab5d48feaafe398151c9f22d4f6 9154494808dc865475c59022c29060b4947a803b Jared Van Bortel jared@nomic.ai 2023-12-21T12:55:34-05:00 GitHub noreply@github.com 2023-12-21T19:55:34+02:00 common : remove incorrect --model-draft default (#4568)
9154494808dc865475c59022c29060b4947a803b c083718c895b7c8c7fb2a4660643fb78d0c64dfd Johannes Gäßler johannesg@5d6.de 2023-12-21T18:42:59+01:00 GitHub noreply@github.com 2023-12-21T18:42:59+01:00 CUDA: mul_mat_id always on GPU for batches >= 32 (#4553)
c083718c895b7c8c7fb2a4660643fb78d0c64dfd 880e352277fc017df4d5794f0c21c44e1eae2b84 Georgi Gerganov ggerganov@gmail.com 2023-12-21T19:27:14+02:00 GitHub noreply@github.com 2023-12-21T19:27:14+02:00 readme : update coding guidelines
880e352277fc017df4d5794f0c21c44e1eae2b84 66f35a2f48e1965a13835a523e677223dbf148be howlger github@voormann.de 2023-12-21T18:07:34+01:00 GitHub noreply@github.com 2023-12-21T19:07:34+02:00 py : open merges file as 'utf-8' (#4566)
66f35a2f48e1965a13835a523e677223dbf148be 139882392258671ffe5acdfcadc0bc08572d6eef bobqianic 129547291+bobqianic@users.noreply.github.com 2023-12-21T17:06:44Z GitHub noreply@github.com 2023-12-21T19:06:44+02:00 cuda : better error message for ggml_get_rows (#4561)
139882392258671ffe5acdfcadc0bc08572d6eef d3223afdad0ed2821a8ddf739c291cd410c92a11 slaren slarengh@gmail.com 2023-12-21T18:02:30+01:00 GitHub noreply@github.com 2023-12-21T18:02:30+01:00 cuda : replace asserts in wrong architecture checks with __trap (#4556)
d3223afdad0ed2821a8ddf739c291cd410c92a11 1d7a1912cea2227f9a1a449758ed622c560542f9 Johannes Gäßler johannesg@5d6.de 2023-12-21T17:34:17+01:00 GitHub noreply@github.com 2023-12-21T18:34:17+02:00 llama : disable per-tensor info prints on model load (#4562)
1d7a1912cea2227f9a1a449758ed622c560542f9 799fc2268989482054944c902874cca76337580f LoganDark github@logandark.mozmail.com 2023-12-21T01:59:27-08:00 GitHub noreply@github.com 2023-12-21T10:59:27+01:00 Fix access violation in ggml_cuda_free_data if tensor->extra is NULL (#4554)
799fc2268989482054944c902874cca76337580f 328b83de23b33240e28f4e74900d1d06726f5eb1 Johannes Gäßler johannesg@5d6.de 2023-12-20T15:41:22+01:00 GitHub noreply@github.com 2023-12-20T15:41:22+01:00 CUDA: Faster Mixtral prompt processing (#4538)
328b83de23b33240e28f4e74900d1d06726f5eb1 a7aee47b98e45539d491071b25778b833b77e387 Eric Sommerlade es0m@users.noreply.github.com 2023-12-19T16:17:01Z GitHub noreply@github.com 2023-12-19T18:17:01+02:00 ggml : fixed check for _MSC_VER (#4535)
a7aee47b98e45539d491071b25778b833b77e387 0e18b2e7d0b5c0a509ea40098def234b8d4a938a arlo-phoenix 140345165+arlo-phoenix@users.noreply.github.com 2023-12-18T22:33:45+01:00 GitHub noreply@github.com 2023-12-18T22:33:45+01:00 ggml-cuda: Fix HIP build (#4528)
0e18b2e7d0b5c0a509ea40098def234b8d4a938a 6ff39b129d0281d045f83d515e51b7197b44b253 Georgi Gerganov ggerganov@gmail.com 2023-12-18T20:17:43+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-18T20:17:43+02:00 llama.swiftui : add tinyllama 1.1B F16
6ff39b129d0281d045f83d515e51b7197b44b253 b9e74f9bca5fdf7d0a22ed25e7a9626335fdfa48 Georgi Gerganov ggerganov@gmail.com 2023-12-18T20:05:12+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-18T20:05:12+02:00 llama.swiftui : add more models
b9e74f9bca5fdf7d0a22ed25e7a9626335fdfa48 3c04bf6da89eaf4c7d317e0518f0687dfcbf2de7 Ebey Abraham ebey97@gmail.com 2023-12-18T17:27:47Z GitHub noreply@github.com 2023-12-18T19:27:47+02:00 llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490)
3c04bf6da89eaf4c7d317e0518f0687dfcbf2de7 2994f0c5a2e8c96955b422dedc93ec2595d16b82 hankcs cnhankmc@gmail.com 2023-12-18T05:14:58-08:00 GitHub noreply@github.com 2023-12-18T15:14:58+02:00 llama : fix try_override for bool_value which always return true (#4519)
2994f0c5a2e8c96955b422dedc93ec2595d16b82 b1306c439490c7fa4ec33594500d980d1e9e15e6 Jared Van Bortel jared@nomic.ai 2023-12-17T19:39:02-05:00 GitHub noreply@github.com 2023-12-17T19:39:02-05:00 decode : fix logits_valid for legacy API (#4516)
b1306c439490c7fa4ec33594500d980d1e9e15e6 800a489e4a8be199122259a995b1ee9dd7fae320 Georgi Gerganov ggerganov@gmail.com 2023-12-17T20:16:23+02:00 GitHub noreply@github.com 2023-12-17T20:16:23+02:00 readme : update hot topics
800a489e4a8be199122259a995b1ee9dd7fae320 f7f468a97dceec2f8fe8b1ed7a2091083446ebc7 Georgi Gerganov ggerganov@gmail.com 2023-12-17T19:38:41+02:00 GitHub noreply@github.com 2023-12-17T19:38:41+02:00 llama.swiftui : add bench functionality (#4483)
f7f468a97dceec2f8fe8b1ed7a2091083446ebc7 919c40660fd27157b391b5832d2a577d5afef4cb Jared Van Bortel jared@nomic.ai 2023-12-17T10:45:46-05:00 GitHub noreply@github.com 2023-12-17T10:45:46-05:00 gguf-py : fail fast on nonsensical special token IDs (#4489)
919c40660fd27157b391b5832d2a577d5afef4cb 45668633fdb522a925c3dafc1ecf426f539efb27 Matheus Gabriel Alves Silva matheusgasource@gmail.com 2023-12-17T12:23:33-03:00 GitHub noreply@github.com 2023-12-17T17:23:33+02:00 build : Check the ROCm installation location (#4485)
45668633fdb522a925c3dafc1ecf426f539efb27 0ffc92d2d23a789625f018840469af045be1e3c0 slaren slarengh@gmail.com 2023-12-17T16:05:56+01:00 GitHub noreply@github.com 2023-12-17T16:05:56+01:00 finetune : keep allocs alive until all allocations are done (#4486)
0ffc92d2d23a789625f018840469af045be1e3c0 8edd2b40fdbcafbf630f2cf29306b29d5cb48c42 olexiyb olexiyb@gmail.com 2023-12-17T17:02:16+02:00 GitHub noreply@github.com 2023-12-17T17:02:16+02:00 server : disable llm logs if SERVER_VERBOSE is off (#3792)
8edd2b40fdbcafbf630f2cf29306b29d5cb48c42 eb16dae7e70ca97396190698b29c0f9ee3388e88 AdithyanI adithyan.i4internet@gmail.com 2023-12-17T15:57:56+01:00 GitHub noreply@github.com 2023-12-17T16:57:56+02:00 server : fix grammar being ignored (#4494)
eb16dae7e70ca97396190698b29c0f9ee3388e88 62bd52b7bf90819e75f427a95a484cd5eee0b3c7 Alexey Parfenov zxed@alkatrazstudio.net 2023-12-17T14:56:09Z GitHub noreply@github.com 2023-12-17T16:56:09+02:00 server : fix possible ambiguity in content type charset (#4501)
62bd52b7bf90819e75f427a95a484cd5eee0b3c7 5daa5f54fdcd2b5228add1a4c43a1897b2168f35 mzcu milos.cubrilo@gmail.com 2023-12-17T15:54:37+01:00 GitHub noreply@github.com 2023-12-17T16:54:37+02:00 server : allow requests larger than 8K (#4500)
5daa5f54fdcd2b5228add1a4c43a1897b2168f35 c6c4fc081c1df1c60a9bfe3e6a3fd086f1a29ec7 Bach Le bach@bullno1.com 2023-12-17T18:57:33+08:00 GitHub noreply@github.com 2023-12-17T11:57:33+01:00 Link to cublas dynamically on Windows even with LLAMA_STATIC (#4506)
c6c4fc081c1df1c60a9bfe3e6a3fd086f1a29ec7 8a5be3bd5885d79ad84aadf32bb8c1a67bd43c19 slaren slarengh@gmail.com 2023-12-16T18:58:46+01:00 GitHub noreply@github.com 2023-12-16T18:58:46+01:00 lora : add support for non-llama models (#3333)
8a5be3bd5885d79ad84aadf32bb8c1a67bd43c19 88ae8952b65cbf32eb1f5703681ea592e510e570 Jared Van Bortel jared@nomic.ai 2023-12-15T22:16:15-05:00 GitHub noreply@github.com 2023-12-15T22:16:15-05:00 llama : sanity checks for access to logits (#4274)
88ae8952b65cbf32eb1f5703681ea592e510e570 ee4725a686643669a8587142fa068cbf29de3ce2 ShadovvBeast ShadovvBeast@gmail.com 2023-12-15T13:49:01+02:00 GitHub noreply@github.com 2023-12-15T13:49:01+02:00 server : add optional API Key Authentication example (#4441)
ee4725a686643669a8587142fa068cbf29de3ce2 6744dbe924a317e3e2a5a2a4a2037061b2223449 slaren slarengh@gmail.com 2023-12-15T12:45:50+01:00 GitHub noreply@github.com 2023-12-15T12:45:50+01:00 ggml : group mul_mat_id rows by matrix (cpu only) (#4480)
6744dbe924a317e3e2a5a2a4a2037061b2223449 cafcd4f89500b8afef722cdb08088eceb8a22572 slaren slarengh@gmail.com 2023-12-14T20:05:21+01:00 GitHub noreply@github.com 2023-12-14T20:05:21+01:00 ggml : use ggml_row_size where possible (#4472)
cafcd4f89500b8afef722cdb08088eceb8a22572 c50e40016394f124b97ce39da48148b1f6c01833 slaren slarengh@gmail.com 2023-12-14T16:52:08+01:00 GitHub noreply@github.com 2023-12-14T16:52:08+01:00 ggml : remove n_dims from ggml_tensor (#4469)
c50e40016394f124b97ce39da48148b1f6c01833 20a68a7030ee06e8eb7eb8e24ae4ac52dc17803f wonjun Jang strutive07@gmail.com 2023-12-14T21:44:49+09:00 GitHub noreply@github.com 2023-12-14T14:44:49+02:00 py : add protobuf dependency (#4466)
20a68a7030ee06e8eb7eb8e24ae4ac52dc17803f 55e87c3749cb4985c3b316984d40e00e4df4a5d0 LostRuins 39025047+LostRuins@users.noreply.github.com 2023-12-14T20:13:33+08:00 GitHub noreply@github.com 2023-12-14T14:13:33+02:00 ggml : add ggml_row_size() (fixes llama out of space) (#4461)
55e87c3749cb4985c3b316984d40e00e4df4a5d0 873637afc7924f435ac44c067630a28e82eefa7b Georgi Gerganov ggerganov@gmail.com 2023-12-14T10:35:29+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-14T10:35:29+02:00 ggml : fix OpenCL broadcast requirement for ggml_mul (close #4453)
873637afc7924f435ac44c067630a28e82eefa7b 0353a1840134b24b07ab61fd4490192f28c4db6b wonjun Jang strutive07@gmail.com 2023-12-14T17:09:34+09:00 GitHub noreply@github.com 2023-12-14T10:09:34+02:00 convert : support loading vocab from fast tokenizer config (#3633)
0353a1840134b24b07ab61fd4490192f28c4db6b 948ff137ec37f1ec74c02905917fa0afc9b97514 BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2023-12-14T02:38:49-05:00 GitHub noreply@github.com 2023-12-14T09:38:49+02:00 readme : update supported model list (#4457)
948ff137ec37f1ec74c02905917fa0afc9b97514 4d98d9a65665eee3838cef936641f640e3f5b649 shibe2 shibe@tuta.io 2023-12-13T23:57:15+04:00 GitHub noreply@github.com 2023-12-13T21:57:15+02:00 server : fix handling of characters that span multiple tokens when streaming (#4446)
4d98d9a65665eee3838cef936641f640e3f5b649 70f806b821f603cafb6f634c93a6729dc21bb354 Georgi Gerganov ggerganov@gmail.com 2023-12-13T21:54:54+02:00 GitHub noreply@github.com 2023-12-13T21:54:54+02:00 sync : ggml (SD ops, tests, kernels) (#4444)
70f806b821f603cafb6f634c93a6729dc21bb354 9fb13f95840c722ad419f390dc8a9c86080a3700 Jared Van Bortel jared@nomic.ai 2023-12-13T12:10:10-05:00 GitHub noreply@github.com 2023-12-13T12:10:10-05:00 build : detect host compiler and cuda compiler separately (#4414)
9fb13f95840c722ad419f390dc8a9c86080a3700 113f9942fc73a262c85e9dcf7c2ea7336250bba0 Siwen Yu yusiwen@gmail.com 2023-12-13T20:50:14+08:00 GitHub noreply@github.com 2023-12-13T14:50:14+02:00 common : add `--version` option to show build info in CLI (#4433)
113f9942fc73a262c85e9dcf7c2ea7336250bba0 799a1cb13b0b1b560ab0ceff485caed68faa8f1f Georgi Gerganov ggerganov@gmail.com 2023-12-13T14:05:38+02:00 GitHub noreply@github.com 2023-12-13T14:05:38+02:00 readme : update hot topics
799a1cb13b0b1b560ab0ceff485caed68faa8f1f fecac45658a99eddc4d6e36ba0310ca8f87a77f0 slaren slarengh@gmail.com 2023-12-13T13:04:25+01:00 GitHub noreply@github.com 2023-12-13T14:04:25+02:00 llama : add Mixtral support (#4406)
fecac45658a99eddc4d6e36ba0310ca8f87a77f0 9494d7c4774ab745490b5a19570ff7747a194143 kalomaze 66376113+kalomaze@users.noreply.github.com 2023-12-12T04:12:35-06:00 GitHub noreply@github.com 2023-12-12T12:12:35+02:00 server : tweak default sampling parameters (#4367)
9494d7c4774ab745490b5a19570ff7747a194143 6138963fb232cbae70c9d181db0ba125708f473d Richard Kiss him@richardkiss.com 2023-12-12T01:53:36-08:00 GitHub noreply@github.com 2023-12-12T11:53:36+02:00 english : use `typos` to fix comments and logs (#4354)
6138963fb232cbae70c9d181db0ba125708f473d 6391817cd19a4507c6c941a1fd08756268662b2d Jared Van Bortel jared@nomic.ai 2023-12-12T04:27:26-05:00 GitHub noreply@github.com 2023-12-12T11:27:26+02:00 build : target Windows 8 for standard mingw-w64 (#4405)
6391817cd19a4507c6c941a1fd08756268662b2d d9d4cfef64ea416dd66632173787d03ffb180cc7 crasm crasm@git.vczf.us 2023-12-12T04:25:57-05:00 GitHub noreply@github.com 2023-12-12T11:25:57+02:00 llama : document logits_all deprecation (#4418)
d9d4cfef64ea416dd66632173787d03ffb180cc7 41a11aaf99feff4901e4c8dc48ad00766c5da4e9 Vladimir Zorin vladimir@deviant.guru 2023-12-12T11:25:29+02:00 GitHub noreply@github.com 2023-12-12T11:25:29+02:00 server : fix local model name in server (#4420)
41a11aaf99feff4901e4c8dc48ad00766c5da4e9 8a7b2fa528f130631a5f43648481596ab320ed5a Taikono-Himazin kazu@po.harenet.ne.jp 2023-12-12T18:24:32+09:00 GitHub noreply@github.com 2023-12-12T11:24:32+02:00 ggml : increased GGML_MAX_PARAMS to allow finetuning of 70b models (#4424)
8a7b2fa528f130631a5f43648481596ab320ed5a e18f7345a300920e234f732077bda660cc6cda9c Yueh-Po Peng 94939112+y10ab1@users.noreply.github.com 2023-12-11T06:27:38+08:00 GitHub noreply@github.com 2023-12-10T23:27:38+01:00 Update README.md (#4388)
e18f7345a300920e234f732077bda660cc6cda9c fe680e3d1080a765e5d3150ffd7bab189742898d Xiang (Kevin) Li kevinli020508@gmail.com 2023-12-09T16:29:27-05:00 GitHub noreply@github.com 2023-12-09T23:29:27+02:00 grammar : revert the replacement of llama_token_to_piece with id_to_token (#4396)
fe680e3d1080a765e5d3150ffd7bab189742898d bcc0eb4591bec5ec02fad3f2bdcb1b265052ea56 Georgi Gerganov ggerganov@gmail.com 2023-12-07T22:26:54+02:00 GitHub noreply@github.com 2023-12-07T22:26:54+02:00 sync : ggml (new ops, tests, backend, etc.) (#4359)
bcc0eb4591bec5ec02fad3f2bdcb1b265052ea56 81bc9214a389362010f7a57f4cbc30e5f83a2d28 Georgi Gerganov ggerganov@gmail.com 2023-12-07T13:03:17+02:00 GitHub noreply@github.com 2023-12-07T13:03:17+02:00 llama : per-layer KV cache + quantum K cache (#4309)
81bc9214a389362010f7a57f4cbc30e5f83a2d28 05cd6e5036d72d0930de4d8f6be7bce09e8dda24 Hongyu Ouyang 96765450+casavaca@users.noreply.github.com 2023-12-07T02:25:22-08:00 GitHub noreply@github.com 2023-12-07T12:25:22+02:00 train : fix #4227 (double free in examples/train-text-from-scratch/train-text-from-scratch.cpp) (#4351)
05cd6e5036d72d0930de4d8f6be7bce09e8dda24 caa9249217c5fd524b900add5ddcbeaa20cbcb12 Georgi Gerganov ggerganov@gmail.com 2023-12-06T20:21:59+02:00 GitHub noreply@github.com 2023-12-06T20:21:59+02:00 server : recognize cache_prompt parameter in OAI API (#4347)
caa9249217c5fd524b900add5ddcbeaa20cbcb12 da5eaef1f34d0a1f584cd4a092e7691ea46a9d91 Georgi Gerganov ggerganov@gmail.com 2023-12-06T10:41:03+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-06T10:41:03+02:00 common : fix compile warning
da5eaef1f34d0a1f584cd4a092e7691ea46a9d91 5f6e0c0dff1e7a89331e6b25eca9a9fd71324069 stduhpf stephduh@live.fr 2023-12-06T09:08:17+01:00 GitHub noreply@github.com 2023-12-06T10:08:17+02:00 speculative : support `--color` (#4343)
5f6e0c0dff1e7a89331e6b25eca9a9fd71324069 5aa365d88fdb8fdd430ef3fc141c7a5fd37c3502 Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2023-12-05T10:55:12-10:00 GitHub noreply@github.com 2023-12-05T22:55:12+02:00 grammar : pre-computed pieces + reserve mem + less string copies (#4330)
5aa365d88fdb8fdd430ef3fc141c7a5fd37c3502 52c8bc3cf312e1caf02d37bfb9d9d865cbe33594 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-12-05T10:19:18-07:00 GitHub noreply@github.com 2023-12-05T19:19:18+02:00 llama : allow overriding GGUF metadata when loading model (#4092)
52c8bc3cf312e1caf02d37bfb9d9d865cbe33594 e4b76bbe316ee50fb17d9ac29e654c0edf830eba MaggotHATE clay1326@gmail.com 2023-12-05T15:05:51+05:00 GitHub noreply@github.com 2023-12-05T12:05:51+02:00 sampling : custom samplers order (#4285)
e4b76bbe316ee50fb17d9ac29e654c0edf830eba 23b5e12eb5a76489b4c3ee22213a081da68b1809 kchro3 62481661+kchro3@users.noreply.github.com 2023-12-04T23:29:46-08:00 GitHub noreply@github.com 2023-12-05T09:29:46+02:00 swift : revert compiler checks for swift package (#4332)
23b5e12eb5a76489b4c3ee22213a081da68b1809 d208995c6da66f252d4054c1c5a90eb8ccb7a2f7 Daniel Bevenius daniel.bevenius@gmail.com 2023-12-04T17:04:21+01:00 GitHub noreply@github.com 2023-12-04T18:04:21+02:00 simple : update error message for KV cache check (#4324)
d208995c6da66f252d4054c1c5a90eb8ccb7a2f7 5c9f90cba1cc6b0a2a7d19ee5dcb73cad6331d30 Miwa / Ensan 63481257+ensan-hcl@users.noreply.github.com 2023-12-05T01:03:49+09:00 GitHub noreply@github.com 2023-12-04T18:03:49+02:00 swift : fix concatenation method to avoid invalid UTF8 stringfication (#4325)
5c9f90cba1cc6b0a2a7d19ee5dcb73cad6331d30 4fa44e84adb4c78e1885694cc3513982d4af2b08 Miwa / Ensan 63481257+ensan-hcl@users.noreply.github.com 2023-12-04T22:43:45+09:00 GitHub noreply@github.com 2023-12-04T15:43:45+02:00 swift : fix prompt tokenization logic (#4321)
4fa44e84adb4c78e1885694cc3513982d4af2b08 fbbc42827b2949b95bcde23ce47bb47d006c895d Ikko Eltociear Ashimine eltociear@gmail.com 2023-12-04T16:57:35+09:00 GitHub noreply@github.com 2023-12-04T09:57:35+02:00 grammar-parser : fix typo (#4318)
fbbc42827b2949b95bcde23ce47bb47d006c895d adf3de4f69ff7e44131222f05f9c7447ac0be3cb Georgi Gerganov ggerganov@gmail.com 2023-12-03T15:56:35+02:00 GitHub noreply@github.com 2023-12-03T15:56:35+02:00 ggml : reuse ggml_get_n_tasks() in ggml_graph_plan() (#4308)
adf3de4f69ff7e44131222f05f9c7447ac0be3cb 33e171d1e9fc4903f9314b490d77fb8d58331b63 Georgi Gerganov ggerganov@gmail.com 2023-12-03T15:56:22+02:00 GitHub noreply@github.com 2023-12-03T15:56:22+02:00 ggml : fix soft max out-of-bounds access (#4307)
33e171d1e9fc4903f9314b490d77fb8d58331b63 6949b50df56ee58a2d76d45487942cb211c08629 Ed Lee edilee@mozilla.com 2023-12-03T01:10:43-08:00 GitHub noreply@github.com 2023-12-03T11:10:43+02:00 server : fix OpenAI API `stop` field to be optional (#4299)
6949b50df56ee58a2d76d45487942cb211c08629 d7b800b8bc490a221acbd83c575206a907f2f6e2 Rickard Edén rickardeden@gmail.com 2023-12-03T10:03:25+01:00 GitHub noreply@github.com 2023-12-03T11:03:25+02:00 py : add grammar to oai like api (#4294)
d7b800b8bc490a221acbd83c575206a907f2f6e2 5a7d3125e7c24f223659b7f0b7aa7736986e92c0 Georgi Gerganov ggerganov@gmail.com 2023-12-03T10:58:16+02:00 GitHub noreply@github.com 2023-12-03T10:58:16+02:00 llama : pad KV cache size (#4280)
5a7d3125e7c24f223659b7f0b7aa7736986e92c0 d5a1cbde60531d02ac74da27ea355182e3a4d516 Georgi Gerganov ggerganov@gmail.com 2023-12-01T20:39:12+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-01T20:39:12+02:00 llama : avoid using "optional" keyword (#4283)
d5a1cbde60531d02ac74da27ea355182e3a4d516 b220222a64ce760bfbec9c770f11db3ec6a6abb6 Georgi Gerganov ggerganov@gmail.com 2023-12-01T20:35:03+02:00 Georgi Gerganov ggerganov@gmail.com 2023-12-01T20:35:47+02:00 llama : support optional tensors (#4283)
b220222a64ce760bfbec9c770f11db3ec6a6abb6 511f52c334e37033f9c9de07b98fca4abc9470bd Miwa / Ensan 63481257+ensan-hcl@users.noreply.github.com 2023-12-02T03:19:45+09:00 GitHub noreply@github.com 2023-12-01T20:19:45+02:00 swift : fix token_to_piece implementation (#4278)
511f52c334e37033f9c9de07b98fca4abc9470bd 03562f3a86d6706eea9f4fc09b532946c191b34e Jared Van Bortel jared@nomic.ai 2023-12-01T13:18:35-05:00 GitHub noreply@github.com 2023-12-01T20:18:35+02:00 build : enable libstdc++ assertions for debug builds (#4275)
03562f3a86d6706eea9f4fc09b532946c191b34e 37c746d687d877bc11803e96b4dc5f378b83c0a0 CausalLM 148736309+CausalLM@users.noreply.github.com 2023-12-02T02:17:06+08:00 GitHub noreply@github.com 2023-12-01T20:17:06+02:00 llama : support attention bias on LLaMA architecture (#4283)
37c746d687d877bc11803e96b4dc5f378b83c0a0 880f57973b8e0091d0f9f50eb5ab4cd4e31582ca Shijie 821898965@qq.com 2023-12-02T02:16:31+08:00 GitHub noreply@github.com 2023-12-01T20:16:31+02:00 llama : add Qwen support (#4281)
880f57973b8e0091d0f9f50eb5ab4cd4e31582ca 8d6d9f033b8101f929e445cf45b39e1557ca7934 Georgi Gerganov ggerganov@gmail.com 2023-12-01T18:42:11+02:00 GitHub noreply@github.com 2023-12-01T18:42:11+02:00 llama : fix integer overflow during quantization (#4284)
8d6d9f033b8101f929e445cf45b39e1557ca7934 ef47ec18da469423c276b683dd9b5741cee7023e Daniel Bevenius daniel.bevenius@gmail.com 2023-12-01T10:41:56+01:00 GitHub noreply@github.com 2023-12-01T11:41:56+02:00 py : add requirements file for convert-hf-to-gguf.py (#4277)
ef47ec18da469423c276b683dd9b5741cee7023e 1d144112c0fbbb4ecc07dbcf4f05a380148bd6de Georgi Gerganov ggerganov@gmail.com 2023-12-01T10:51:24+02:00 GitHub noreply@github.com 2023-12-01T10:51:24+02:00 ggml : add ggml_soft_max_ext (#4256)
1d144112c0fbbb4ecc07dbcf4f05a380148bd6de f43f09366dfd018e4568e23a232aaa8c4f7cfc78 Ziad Ben Hadj-Alouane zied.benhadjalouane@gmail.com 2023-11-30T17:25:49-05:00 GitHub noreply@github.com 2023-12-01T00:25:49+02:00 server : add --log-disable to disable logging to file (#4260)
f43f09366dfd018e4568e23a232aaa8c4f7cfc78 d2809a3ba2780e00fce5a6149a7eda09f1c0e906 Ziad Ben Hadj-Alouane zied.benhadjalouane@gmail.com 2023-11-30T17:25:04-05:00 GitHub noreply@github.com 2023-12-01T00:25:04+02:00 server : add single-client multi-prompt support (#4232)
d2809a3ba2780e00fce5a6149a7eda09f1c0e906 15f5d96037e597523b721aa39c874d69de2acf85 WillCorticesAI 150854901+WillCorticesAI@users.noreply.github.com 2023-11-30T17:23:44-05:00 GitHub noreply@github.com 2023-12-01T00:23:44+02:00 make : fix Apple clang determination bug (#4272)
15f5d96037e597523b721aa39c874d69de2acf85 33c9892af58b7b161f2a532935dcccff8c8048c6 Jared Van Bortel jared@nomic.ai 2023-11-30T17:23:08-05:00 GitHub noreply@github.com 2023-12-01T00:23:08+02:00 build : fix build info generation and cleanup Makefile (#3920)
33c9892af58b7b161f2a532935dcccff8c8048c6 8efa0f6ebed53c9453e6721da86fb294e5015909 John 78893154+cmp-nct@users.noreply.github.com 2023-11-30T23:11:14+01:00 GitHub noreply@github.com 2023-11-30T23:11:14+01:00 llava : ShareGPT4V compatibility (vision encoder only loading) (#4172)
8efa0f6ebed53c9453e6721da86fb294e5015909 524907aa768a26cbf83d8e2eb30547e2ee1d1b1a Andrew Godfrey AndrewGodfrey@users.noreply.github.com 2023-11-30T13:56:19-08:00 GitHub noreply@github.com 2023-11-30T23:56:19+02:00 main : pass LOG_TEE callback to llama.cpp log (#4033)
524907aa768a26cbf83d8e2eb30547e2ee1d1b1a 3bd2c7ce1b752973cf937482a0333e85d1681e2b vodkaslime 646329483@qq.com 2023-12-01T05:49:21+08:00 GitHub noreply@github.com 2023-11-30T23:49:21+02:00 readme : fix (#4135)
3bd2c7ce1b752973cf937482a0333e85d1681e2b bde629bb53b85886ee0fe83524c1efe2689bc618 Juraj Bednar juraj@bednar.io 2023-11-30T22:46:01+01:00 GitHub noreply@github.com 2023-11-30T23:46:01+02:00 docker : add finetune option (#4211)
bde629bb53b85886ee0fe83524c1efe2689bc618 f7f9e06212d44530b3200033286049dbdf84b3d3 Miwa / Ensan 63481257+ensan-hcl@users.noreply.github.com 2023-12-01T06:45:17+09:00 GitHub noreply@github.com 2023-11-30T23:45:17+02:00 batched.swift : update README.md (#4214)
f7f9e06212d44530b3200033286049dbdf84b3d3 74daabae6927b99e7333d6126dee35193c418457 Li Tan tanliboy@gmail.com 2023-11-30T13:44:11-08:00 GitHub noreply@github.com 2023-11-30T23:44:11+02:00 cmake : fix the metal file foder path (#4217)
74daabae6927b99e7333d6126dee35193c418457 b18c66ca6eee4fe0465cff5042daf05005dc9ab2 Dawid Wysocki 62249621+TortillaZHawaii@users.noreply.github.com 2023-11-30T22:43:32+01:00 GitHub noreply@github.com 2023-11-30T23:43:32+02:00 readme : fix typo (#4253)
b18c66ca6eee4fe0465cff5042daf05005dc9ab2 f4d973cecb7368c985720ba9100ae6abba14806d Daniel Bevenius daniel.bevenius@gmail.com 2023-11-30T22:43:08+01:00 GitHub noreply@github.com 2023-11-30T23:43:08+02:00 llama : fix alignment of general.name in print meta (#4254)
f4d973cecb7368c985720ba9100ae6abba14806d 954e22858c5cea1dc03e9172d3879402af2b5990 slaren slarengh@gmail.com 2023-11-30T22:42:23+01:00 GitHub noreply@github.com 2023-11-30T23:42:23+02:00 convert.py : fix llama/llama2 conversion due to vocab_size=-1 (#4258)
954e22858c5cea1dc03e9172d3879402af2b5990 e2bd725f4b39bc5c6234858d158e01248f5ab5bd tarcey cey.tarik@gmail.com 2023-11-30T22:40:23+01:00 GitHub noreply@github.com 2023-11-30T23:40:23+02:00 llama : fix typical sampling (#4261)
e2bd725f4b39bc5c6234858d158e01248f5ab5bd 1f5cd83275fabb43f2ae92c30033b384a3eb37b4 rhjdvsgsgks 26178113+rhjdvsgsgks@users.noreply.github.com 2023-11-30T20:50:40Z GitHub noreply@github.com 2023-11-30T22:50:40+02:00 py : fix oai proxy (#3972)
1f5cd83275fabb43f2ae92c30033b384a3eb37b4 4fea3420ee3918d125d74c94d962a6ea82875351 Georgi Gerganov ggerganov@gmail.com 2023-11-29T11:00:17+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-29T11:00:17+02:00 examples : add readme files
4fea3420ee3918d125d74c94d962a6ea82875351 64e64aa2557d97490b2fe1262b313e2f4a1607e3 Peter Sugihara peter@campsh.com 2023-11-28T23:16:34-08:00 GitHub noreply@github.com 2023-11-29T09:16:34+02:00 readme : add FreeChat (#4248)
64e64aa2557d97490b2fe1262b313e2f4a1607e3 8406b0924bf323f37d536dee8b8165c1f3d9d11d Jared Van Bortel jared@nomic.ai 2023-11-28T04:51:11-05:00 GitHub noreply@github.com 2023-11-28T11:51:11+02:00 ggml : restore abort() in GGML_ASSERT (#4242)
8406b0924bf323f37d536dee8b8165c1f3d9d11d b38a16dfcff88d547f78f52d1bea31b84a05aff7 Georgi Gerganov ggerganov@gmail.com 2023-11-28T10:32:03+02:00 GitHub noreply@github.com 2023-11-28T10:32:03+02:00 ggml : re-enable BLAS for CPU when src0 != F32 + remove redundant full offload checks in llama.cpp (#4240)
b38a16dfcff88d547f78f52d1bea31b84a05aff7 0dab8cd7cca7e1bc3550dcb4797b9062cdbb1ebd bandoti 141645996+bandoti@users.noreply.github.com 2023-11-27T15:25:42-04:00 GitHub noreply@github.com 2023-11-27T21:25:42+02:00 cmake : fix issue with version info not getting baked into LlamaConfig.cmake (#3970)
0dab8cd7cca7e1bc3550dcb4797b9062cdbb1ebd bb03290c17540768a16000a2b01ee4f22440aba1 Kasumi 90275229+kasumi-1@users.noreply.github.com 2023-11-28T01:39:42+08:00 GitHub noreply@github.com 2023-11-27T19:39:42+02:00 readme : add Amica to UI list (#4230)
bb03290c17540768a16000a2b01ee4f22440aba1 f3b269813f6147c5b5cda082e6b45cf04a932e0d Bailey Chittle 39804642+bachittle@users.noreply.github.com 2023-11-27T09:56:52-05:00 GitHub noreply@github.com 2023-11-27T16:56:52+02:00 examples : iOS example with swift ui (#4159)
f3b269813f6147c5b5cda082e6b45cf04a932e0d 3e73d31d9cc0232882ce61c64742aff3ecfec416 Jared Van Bortel jared@nomic.ai 2023-11-26T22:58:43-05:00 GitHub noreply@github.com 2023-11-26T22:58:43-05:00 ggml : fix -Warray-bounds warning with gcc (#4231)
3e73d31d9cc0232882ce61c64742aff3ecfec416 9656026b53236ed7328458269c4c798dd50ac8d1 Georgi Gerganov ggerganov@gmail.com 2023-11-26T21:51:46+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-26T21:52:23+02:00 lookahead : support `-n -1` infinite generation
9656026b53236ed7328458269c4c798dd50ac8d1 922754a8d60080e956891f6cee1fb03aa48d57c6 Georgi Gerganov ggerganov@gmail.com 2023-11-26T20:42:51+02:00 GitHub noreply@github.com 2023-11-26T20:42:51+02:00 readme : update hot topics
922754a8d60080e956891f6cee1fb03aa48d57c6 22da05536ff4ad963080773bef1fb839fdab95d3 Georgi Gerganov ggerganov@gmail.com 2023-11-26T20:33:07+02:00 GitHub noreply@github.com 2023-11-26T20:33:07+02:00 lookahead : add example for lookahead decoding (#4207)
22da05536ff4ad963080773bef1fb839fdab95d3 1ddb52ec38f9931925a587f45a23b1c37152c028 Xiao-Yong Jin jinxiaoyong@gmail.com 2023-11-26T02:30:02-06:00 GitHub noreply@github.com 2023-11-26T10:30:02+02:00 metal : fix yarn (#4220)
1ddb52ec38f9931925a587f45a23b1c37152c028 f837c3a992b2b6146936cb120871a8cf9d0e3857 Galunid karolek1231456@gmail.com 2023-11-25T22:45:02+01:00 GitHub noreply@github.com 2023-11-25T22:45:02+01:00 scripts : Use mmap in torch load (#4202)
f837c3a992b2b6146936cb120871a8cf9d0e3857 3014b5415d08e3dff961da6eea835b9760a701b8 Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2023-11-25T08:58:23-08:00 GitHub noreply@github.com 2023-11-25T18:58:23+02:00 llama : grammar `reserve` space in `decode_utf8` (#4210)
3014b5415d08e3dff961da6eea835b9760a701b8 04814e718edb13bdf8cca861dc2e5ab4e1995c30 crasm crasm@git.vczf.us 2023-11-25T10:47:07-05:00 GitHub noreply@github.com 2023-11-25T10:47:07-05:00 Update docs for yarn_ext_factor <0.0 as unspecified instead of NaN (#4189)
04814e718edb13bdf8cca861dc2e5ab4e1995c30 af19d3573481d409b3c4e55494810eb1f65a9aae Georgi Gerganov ggerganov@gmail.com 2023-11-25T12:02:13+02:00 GitHub noreply@github.com 2023-11-25T12:02:13+02:00 readme : update hot topics
af19d3573481d409b3c4e55494810eb1f65a9aae e9c13ff78114af6fc6a4f27cc8dcdda0f3d389fb Georgi Gerganov ggerganov@gmail.com 2023-11-25T11:29:06+02:00 GitHub noreply@github.com 2023-11-25T11:29:06+02:00 server : OAI API compatibility (#4198)
e9c13ff78114af6fc6a4f27cc8dcdda0f3d389fb 8a052c131ed3525313cdb84e5ae4e2b6cf8d2e24 slaren slarengh@gmail.com 2023-11-24T18:10:01+01:00 GitHub noreply@github.com 2023-11-24T18:10:01+01:00 llama : set metal log callback correctly (#4204)
8a052c131ed3525313cdb84e5ae4e2b6cf8d2e24 189d68446e7ef21e8f3af3c0a3d91c35a39aec89 slaren slarengh@gmail.com 2023-11-24T18:04:31+01:00 GitHub noreply@github.com 2023-11-24T18:04:31+01:00 ggml-cuda : support stablelm rope (#4156)
189d68446e7ef21e8f3af3c0a3d91c35a39aec89 2568a4bf548d7392e9c78c008b33b4c11d53fe95 Galunid karolek1231456@gmail.com 2023-11-24T15:02:49+01:00 GitHub noreply@github.com 2023-11-24T15:02:49+01:00 convert : fix tensors using grad in some models (#4173)
2568a4bf548d7392e9c78c008b33b4c11d53fe95 b35f3d0def3efde92ed465d92a267430d957e87d eastriver lee@eastriver.dev 2023-11-24T18:25:10+09:00 GitHub noreply@github.com 2023-11-24T11:25:10+02:00 main.swift : fix eos checking (#4197)
b35f3d0def3efde92ed465d92a267430d957e87d 55978ce09b69d3987d17d08d92d8cc27193e0773 Aaryaman Vasishta aaryaman.vasishta@amd.com 2023-11-24T16:52:39+09:00 GitHub noreply@github.com 2023-11-24T09:52:39+02:00 readme : use PATH for Windows ROCm (#4195)
55978ce09b69d3987d17d08d92d8cc27193e0773 6b0a7420d03b9d13cb0e9439a01ce8476d8bf093 Haohui Mai ricetons@gmail.com 2023-11-23T13:56:53-08:00 GitHub noreply@github.com 2023-11-23T22:56:53+01:00 Fix incorrect format strings and uninitialized variables. (#4133)
6b0a7420d03b9d13cb0e9439a01ce8476d8bf093 d103d935c0e75769a6a597f7a64cab72c6cc3e79 Georgi Gerganov ggerganov@gmail.com 2023-11-23T19:07:56+02:00 GitHub noreply@github.com 2023-11-23T19:07:56+02:00 llama : KV cache view API + better KV cache management (#4170)
d103d935c0e75769a6a597f7a64cab72c6cc3e79 9d5949f04b1f8b76184818abbd99938c2020803f Georgi Gerganov ggerganov@gmail.com 2023-11-23T13:51:22+02:00 GitHub noreply@github.com 2023-11-23T13:51:22+02:00 readme : update hot topics
9d5949f04b1f8b76184818abbd99938c2020803f ff8238f71d56245f4a6dbea693f4ebd81263464d Daniel Bevenius daniel.bevenius@gmail.com 2023-11-23T12:34:20+01:00 GitHub noreply@github.com 2023-11-23T13:34:20+02:00 examples : fix typo in parallel example doc comment (#4181)
ff8238f71d56245f4a6dbea693f4ebd81263464d 8e672efe632bb6a7333964a255c4b96f018b9a65 Georgi Gerganov ggerganov@gmail.com 2023-11-23T11:35:04+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-23T11:35:04+02:00 docs : add llama-star arch idea
8e672efe632bb6a7333964a255c4b96f018b9a65 0b871f1a04ef60e114bbe43004fd9c21114e802d Galunid karolek1231456@gmail.com 2023-11-21T16:22:30+01:00 GitHub noreply@github.com 2023-11-21T16:22:30+01:00 stablelm : simplify + speedup generation (#4153)
0b871f1a04ef60e114bbe43004fd9c21114e802d dfc7cd48b1cc31d759c093e917a18c0efe03d0e8 Galunid karolek1231456@gmail.com 2023-11-20T19:30:00+01:00 GitHub noreply@github.com 2023-11-20T19:30:00+01:00 finetune - update readme to mention llama support only (#4148)
dfc7cd48b1cc31d759c093e917a18c0efe03d0e8 881800d1f083c39431cef288347082be516d1c80 Aaryaman Vasishta aaryaman.vasishta@amd.com 2023-11-21T00:02:46+09:00 GitHub noreply@github.com 2023-11-20T17:02:46+02:00 readme : update ROCm Windows instructions (#4122)
881800d1f083c39431cef288347082be516d1c80 f23c0359a32871947169a044eb1dc4dbffd0f405 Seb C 47074056+Sebby37@users.noreply.github.com 2023-11-21T00:26:59+10:30 GitHub noreply@github.com 2023-11-20T14:56:59+01:00 main : Add ChatML functionality to main example (#4046)
f23c0359a32871947169a044eb1dc4dbffd0f405 40a34fe8d034bd484efd79ccbb95059ca6308dcb Galunid karolek1231456@gmail.com 2023-11-20T11:35:47+01:00 GitHub noreply@github.com 2023-11-20T11:35:47+01:00 ci : add flake8 to github actions (python linting) (#4129)
40a34fe8d034bd484efd79ccbb95059ca6308dcb dae06c06e5c6232ae2be4d567dd5101e1e96c814 Branden Butler bwtbutler@hotmail.com 2023-11-20T03:50:04-06:00 GitHub noreply@github.com 2023-11-20T11:50:04+02:00 speculative : fix prompt tokenization in speculative example (#4025)
dae06c06e5c6232ae2be4d567dd5101e1e96c814 05e8301e4593e2a67b4bae24f093dd12ce5cc7c2 Georgi Gerganov ggerganov@gmail.com 2023-11-19T19:16:07+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-19T19:16:07+02:00 Revert "finetune : add --n-gpu-layers flag info to --help (#4128)"
05e8301e4593e2a67b4bae24f093dd12ce5cc7c2 936c79b2275a8f15f3512e63de615c676904d650 Clark Saben 76020733+csaben@users.noreply.github.com 2023-11-19T11:56:38-05:00 GitHub noreply@github.com 2023-11-19T18:56:38+02:00 finetune : add --n-gpu-layers flag info to --help (#4128)
936c79b2275a8f15f3512e63de615c676904d650 262005ad9ded375e9c544a02c18ef6254fe185a2 SoftwareRenderer 138734813+SoftwareRenderer@users.noreply.github.com 2023-11-19T11:54:10-05:00 GitHub noreply@github.com 2023-11-19T18:54:10+02:00 server : relay error messages (#4131)
262005ad9ded375e9c544a02c18ef6254fe185a2 35985acffaab1eb4ffcbc22c86063bc630f24a89 kchro3 62481661+kchro3@users.noreply.github.com 2023-11-19T08:52:57-08:00 GitHub noreply@github.com 2023-11-19T18:52:57+02:00 common : comma should be semicolon (#4137)
35985acffaab1eb4ffcbc22c86063bc630f24a89 e937066420b79a757bf80e9836eb12b88420a218 Georgi Gerganov ggerganov@gmail.com 2023-11-19T18:50:49+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-19T18:50:49+02:00 gitignore : tokenize
e937066420b79a757bf80e9836eb12b88420a218 28a2e6e7d476717881be6eb9e2d3331342cec57b slaren slarengh@gmail.com 2023-11-19T11:10:52+01:00 GitHub noreply@github.com 2023-11-19T11:10:52+01:00 gguf-py : export chat templates (#4125)
28a2e6e7d476717881be6eb9e2d3331342cec57b 0b5c3b04572a05f80163a365070fb377a837ac27 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-18T14:48:17-07:00 GitHub noreply@github.com 2023-11-18T14:48:17-07:00 tokenize example: Respect normal add BOS token behavior (#4126)
0b5c3b04572a05f80163a365070fb377a837ac27 2923f17f6fec049a71186636c3c4d96408856194 Galunid karolek1231456@gmail.com 2023-11-18T21:08:33+01:00 GitHub noreply@github.com 2023-11-18T21:08:33+01:00 scripts : Remove missed baichuan convert script (#4127)
2923f17f6fec049a71186636c3c4d96408856194 bbecf3f415797f812893947998bda4f866fa900e Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-18T08:11:18-07:00 GitHub noreply@github.com 2023-11-18T08:11:18-07:00 Clean up ggml-cuda.cu warnings when compiling with clang (for ROCM) (#4124)
bbecf3f415797f812893947998bda4f866fa900e 8e9361089dd31ae9ae59452a8ee409fd51a16371 slaren slarengh@gmail.com 2023-11-17T20:39:11+01:00 GitHub noreply@github.com 2023-11-17T21:39:11+02:00 llama : increase max nodes (#4115)
8e9361089dd31ae9ae59452a8ee409fd51a16371 5ad387e994dde77a47ec547a4a65f7611dc325f4 Roger Meier r.meier@siemens.com 2023-11-17T17:11:23+01:00 GitHub noreply@github.com 2023-11-17T18:11:23+02:00 build : support ppc64le build for make and CMake (#3963)
5ad387e994dde77a47ec547a4a65f7611dc325f4 2fa02b4b3d86182381311c98b75065ee1b7c2930 Georgi Gerganov ggerganov@gmail.com 2023-11-17T18:01:38+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-17T18:01:38+02:00 tokenize : fix trailing whitespace
2fa02b4b3d86182381311c98b75065ee1b7c2930 2ab0707acbf3a3ca9e5bc5959c7920c22eba2257 zakkor edward.partenie@gmail.com 2023-11-17T17:36:44+02:00 GitHub noreply@github.com 2023-11-17T17:36:44+02:00 examples : add tokenize (#4039)
2ab0707acbf3a3ca9e5bc5959c7920c22eba2257 11173c92d6eaa2bd1308c2389f44f838480836ac Don Mahurin dmahurin@users.noreply.github.com 2023-11-17T07:32:34-08:00 GitHub noreply@github.com 2023-11-17T17:32:34+02:00 convert : use 'model' value if it exists. This allows karpathy/tinyllamas to load (#4089)
11173c92d6eaa2bd1308c2389f44f838480836ac 9e87ef60e18d69338c5efea314aa7e718bf2040a John 78893154+cmp-nct@users.noreply.github.com 2023-11-17T16:24:30+01:00 GitHub noreply@github.com 2023-11-17T17:24:30+02:00 py : Falcon HF compatibility (#4104)
9e87ef60e18d69338c5efea314aa7e718bf2040a c7cce1246e248124117ae5bc058923e3ade95f11 Jannis Schönleber joennlae@gmail.com 2023-11-17T16:24:07+01:00 GitHub noreply@github.com 2023-11-17T17:24:07+02:00 common : improve yaml log escaping (#4080)
c7cce1246e248124117ae5bc058923e3ade95f11 f7d5e975424ff0eea55ca5a9181ac8e15553c1fc Huawei Lin huaweilin.cs@gmail.com 2023-11-17T10:22:56-05:00 GitHub noreply@github.com 2023-11-17T17:22:56+02:00 llava : fix compilation warning that fread return value is not used (#4069)
f7d5e975424ff0eea55ca5a9181ac8e15553c1fc ba4cf5c0bf37a729d29e899dadf14541cddd23d4 Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2023-11-17T16:20:53+01:00 GitHub noreply@github.com 2023-11-17T17:20:53+02:00 py : remove superfluous import statements (#4076)
ba4cf5c0bf37a729d29e899dadf14541cddd23d4 e85bb1a8e736228a1f0d965777de5f77f22834b8 Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2023-11-17T16:19:16+01:00 GitHub noreply@github.com 2023-11-17T17:19:16+02:00 train : move number of gpu layers argument parsing to common/train.cpp (#4074)
e85bb1a8e736228a1f0d965777de5f77f22834b8 3e916a07ac093045d88ef0c4fa78647ae0efc010 slaren slarengh@gmail.com 2023-11-17T16:17:37+01:00 GitHub noreply@github.com 2023-11-17T17:17:37+02:00 llama : add functions to get the model's metadata (#4013)
3e916a07ac093045d88ef0c4fa78647ae0efc010 947f64f1630bb8b0b363a3bb5e29e11425312d57 gwjr 502526+gwjr@users.noreply.github.com 2023-11-17T14:48:19Z GitHub noreply@github.com 2023-11-17T16:48:19+02:00 finetune : speed-up ggml_compute_forward_out_prod_f32 via BLAS (#4079)
947f64f1630bb8b0b363a3bb5e29e11425312d57 b83e149ec6264d078e6a47412e7347bf5c2bfcc9 Andrew Godfrey AndrewGodfrey@users.noreply.github.com 2023-11-17T02:23:11-08:00 GitHub noreply@github.com 2023-11-17T11:23:11+01:00 finetune : zero the loraB initial vectors (#4082)
b83e149ec6264d078e6a47412e7347bf5c2bfcc9 4f447a48339977073a1af4f33ae873465ff64994 Andrew Godfrey AndrewGodfrey@users.noreply.github.com 2023-11-17T00:01:15-08:00 GitHub noreply@github.com 2023-11-17T10:01:15+02:00 cuda : get_row_rounding F32 (#4095)
4f447a48339977073a1af4f33ae873465ff64994 91f6499393d2d999331fbfdba47a7f8b9f913f0d Georgi Gerganov ggerganov@gmail.com 2023-11-17T10:00:15+02:00 GitHub noreply@github.com 2023-11-17T10:00:15+02:00 llama : fix data units (#4101)
91f6499393d2d999331fbfdba47a7f8b9f913f0d 8da46278e1a57107591653275f8e03a281de94f0 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-16T19:14:37-07:00 GitHub noreply@github.com 2023-11-16T19:14:37-07:00 Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040)
8da46278e1a57107591653275f8e03a281de94f0 a6fc554e268634494f33b0de76f9dde650dd292f texmex76 40733439+texmex76@users.noreply.github.com 2023-11-16T16:01:48+01:00 GitHub noreply@github.com 2023-11-16T17:01:48+02:00 gguf : fix potential infinite loops while parsing (#4100)
a6fc554e268634494f33b0de76f9dde650dd292f 1cf2850d52bb027aa215e039ed9a0c61beeef8d3 Jared Van Bortel jared@nomic.ai 2023-11-15T11:34:47-05:00 GitHub noreply@github.com 2023-11-15T11:34:47-05:00 llama : restore prefix space in llama tokenizer (#4081)
1cf2850d52bb027aa215e039ed9a0c61beeef8d3 6bb4908a17150b49373b5f977685b2e180a04f6f slaren slarengh@gmail.com 2023-11-15T13:58:13+01:00 GitHub noreply@github.com 2023-11-15T14:58:13+02:00 ggml-cuda : increase max graph size (#4084)
6bb4908a17150b49373b5f977685b2e180a04f6f 36eed0c42c5b0bf74af81fb9243d262014f9382f Michael Potter NanoTekGuy@Gmail.com 2023-11-14T09:34:41-08:00 GitHub noreply@github.com 2023-11-14T12:34:41-05:00 Fix MacOS Sonoma model quantization (#4052)
36eed0c42c5b0bf74af81fb9243d262014f9382f b46d12f86d56bef3dc8b596dfb3d22f3b08102be Galunid karolek1231456@gmail.com 2023-11-14T11:17:12+01:00 GitHub noreply@github.com 2023-11-14T11:17:12+01:00 stablelm : StableLM support (#3586)
b46d12f86d56bef3dc8b596dfb3d22f3b08102be bd90eca237b498dd106d315dcb9ad3e6fae3906f afrideva 95653597+afrideva@users.noreply.github.com 2023-11-13T17:03:40-08:00 GitHub noreply@github.com 2023-11-13T18:03:40-07:00 convert.py: also look for plain model.safetensors (#4043)
bd90eca237b498dd106d315dcb9ad3e6fae3906f 3d68f364f15778dc326f5024f2e5af1ad6dfddef M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-11-13T18:20:52+03:00 GitHub noreply@github.com 2023-11-13T18:20:52+03:00 llava : fix regression for square images in #3613 (#4056)
3d68f364f15778dc326f5024f2e5af1ad6dfddef c049b37d7baf558944501705b91ac89b26ee3e41 Georgi Gerganov ggerganov@gmail.com 2023-11-13T16:55:52+02:00 GitHub noreply@github.com 2023-11-13T16:55:52+02:00 ggml : sync (im2col, GPU conv, 32-bit arm compat) (#4060)
c049b37d7baf558944501705b91ac89b26ee3e41 4760e7cc0b68570d58f55e8dda469805d1759d0d Georgi Gerganov ggerganov@gmail.com 2023-11-13T14:18:08+02:00 GitHub noreply@github.com 2023-11-13T14:18:08+02:00 readme : update hot topics
4760e7cc0b68570d58f55e8dda469805d1759d0d bb50a792ec2a49944470c82694fa364345e95170 Georgi Gerganov ggerganov@gmail.com 2023-11-13T14:16:23+02:00 GitHub noreply@github.com 2023-11-13T14:16:23+02:00 sync : ggml (backend v2) (#3912)
bb50a792ec2a49944470c82694fa364345e95170 21fd874c8d2a14dea2d56724e4357c0824aee6a8 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-13T01:58:15-07:00 GitHub noreply@github.com 2023-11-13T01:58:15-07:00 Add ReLU and SQR CUDA ops to (partially) fix Persimmon offloading (#4041)
21fd874c8d2a14dea2d56724e4357c0824aee6a8 532dd74e38c29e16ea1cfc4e7eedb4f2fab3f3cd Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-12T16:39:37-07:00 GitHub noreply@github.com 2023-11-12T16:39:37-07:00 gguf-py: gguf_writer: Use bytearray to build metadata (#4051)
532dd74e38c29e16ea1cfc4e7eedb4f2fab3f3cd e86fc56f7521ca4b18d1d9939e82abd40c2f1c01 Richard Kiss him@richardkiss.com 2023-11-11T22:04:58-08:00 GitHub noreply@github.com 2023-11-11T23:04:58-07:00 Fix some documentation typos/grammar mistakes (#4032)
e86fc56f7521ca4b18d1d9939e82abd40c2f1c01 d96ca7ded77df764db797b68b4a29e34c5b56285 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-11-11T18:35:31+03:00 GitHub noreply@github.com 2023-11-11T08:35:31-07:00 Fix gguf-convert-endian script (#4037)
d96ca7ded77df764db797b68b4a29e34c5b56285 34b0a082074b073eb14c2bd93c0c070e20ddcd16 Alexey Parfenov zxed@alkatrazstudio.net 2023-11-11T05:48:21Z GitHub noreply@github.com 2023-11-10T23:48:21-06:00 server : fix crash when prompt exceeds context size (#3996)
34b0a082074b073eb14c2bd93c0c070e20ddcd16 4a4fd3eefad5bd17ab6bcd8e2181b4f62eae76cf Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-10T22:04:50-07:00 GitHub noreply@github.com 2023-11-11T08:04:50+03:00 gguf-py: Refactor and allow reading/modifying existing GGUF files (#3981)
4a4fd3eefad5bd17ab6bcd8e2181b4f62eae76cf df9d1293defe783f42bc83af732d3c670552c541 Jhen-Jie Hong iainst0409@gmail.com 2023-11-11T06:49:33+08:00 GitHub noreply@github.com 2023-11-10T16:49:33-06:00 server : allow continue edit on completion mode (#3950)
df9d1293defe783f42bc83af732d3c670552c541 a75fa576abba9d37f463580c379e4bbf1e1ad03c Galunid karolek1231456@gmail.com 2023-11-10T14:24:54+01:00 GitHub noreply@github.com 2023-11-10T14:24:54+01:00 Unbreak persimmon after #3837 (#4010)
a75fa576abba9d37f463580c379e4bbf1e1ad03c 57ad015dc3011b046ed5a23186c86ea55f987c54 Galunid karolek1231456@gmail.com 2023-11-09T11:09:29+01:00 GitHub noreply@github.com 2023-11-09T11:09:29+01:00 scripts: Generalize convert scripts (#3838)
57ad015dc3011b046ed5a23186c86ea55f987c54 875fb42871a0f5a88fbe31a0b5edd697b84038e4 Mihai mihai.chirculescu@yahoo.com 2023-11-09T04:00:34+02:00 GitHub noreply@github.com 2023-11-08T20:00:34-06:00 server : add min_p param (#3877)
875fb42871a0f5a88fbe31a0b5edd697b84038e4 0a7c980b6f94a049cb804573df2d8092a34df8e4 slaren slarengh@gmail.com 2023-11-08T13:15:14+01:00 GitHub noreply@github.com 2023-11-08T13:15:14+01:00 ggml-alloc : fix backend assignments of views (#3982)
0a7c980b6f94a049cb804573df2d8092a34df8e4 413503d4b92500d82b002d03c580a71a54747138 Jared Van Bortel cebtenzzre@gmail.com 2023-11-07T12:43:04-05:00 GitHub noreply@github.com 2023-11-07T12:43:04-05:00 gguf : track writer state, free unneeded tensors, cleanup (#3871)
413503d4b92500d82b002d03c580a71a54747138 e9c1cecb9d7d743d30b4a29ecd56a411437def0a Georgi Gerganov ggerganov@gmail.com 2023-11-07T19:25:32+02:00 GitHub noreply@github.com 2023-11-07T20:25:32+03:00 make : do not add linker flags when compiling static llava lib (#3977)
e9c1cecb9d7d743d30b4a29ecd56a411437def0a 54b4df8886103b436a4bb3b60f4d84824f9e8868 xaedes xaedes@gmail.com 2023-11-07T09:04:51+01:00 GitHub noreply@github.com 2023-11-07T10:04:51+02:00 ggml : fix backward rope after YaRN (#3974)
54b4df8886103b436a4bb3b60f4d84824f9e8868 46876d2a2c92e60579dc732cdb8cbd243b06f317 Matthew Tejo matthew.tejo@gmail.com 2023-11-06T23:43:59-08:00 GitHub noreply@github.com 2023-11-07T10:43:59+03:00 Use params when loading models in llava-cli (#3976)
46876d2a2c92e60579dc732cdb8cbd243b06f317 381efbf480959bb6d1e247a8b0c2328f22e350f8 Meng Zhang meng@tabbyml.com 2023-11-06T22:49:08-08:00 GitHub noreply@github.com 2023-11-07T08:49:08+02:00 cuda : supports running on CPU for GGML_USE_CUBLAS=ON build (#3946)
381efbf480959bb6d1e247a8b0c2328f22e350f8 2833a6f63c1b87c7f4ac574bcf7a15a2f3bf3ede Damian Stewart d@damianstewart.com 2023-11-06T22:36:23+01:00 GitHub noreply@github.com 2023-11-07T00:36:23+03:00 llava : expose as a shared library for downstream projects (#3613)
2833a6f63c1b87c7f4ac574bcf7a15a2f3bf3ede d9ccce2e339ca0396560d18b8637f2c848d72a08 slaren slarengh@gmail.com 2023-11-05T18:45:16+01:00 GitHub noreply@github.com 2023-11-05T18:45:16+01:00 ggml-cuda : fix f16 mul mat (#3961)
d9ccce2e339ca0396560d18b8637f2c848d72a08 bb60fd0bf6bb270744d86dd45b3a95af01b7de45 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-05T10:06:06-07:00 GitHub noreply@github.com 2023-11-05T10:06:06-07:00 Allow common process_escapes to handle \x sequences (#3928)
bb60fd0bf6bb270744d86dd45b3a95af01b7de45 132d25b8a62ea084447e0014a0112c1b371fb3f8 Thái Hoàng Tâm 75922889+RoyalHeart@users.noreply.github.com 2023-11-05T23:15:27+07:00 GitHub noreply@github.com 2023-11-05T18:15:27+02:00 server : fix typo for --alias shortcut from -m to -a (#3958)
132d25b8a62ea084447e0014a0112c1b371fb3f8 3d48f42efcd05381221654376e9f6f69d76af739 Jared Van Bortel cebtenzzre@gmail.com 2023-11-05T10:08:57-05:00 GitHub noreply@github.com 2023-11-05T10:08:57-05:00 cuda : fix disabling device with --tensor-split 1,0 (#3951)
3d48f42efcd05381221654376e9f6f69d76af739 c41ea36eaa3548776de4cb3d5d49b925cd3fc0f2 Meng Zhang meng@tabbyml.com 2023-11-05T04:40:08-08:00 GitHub noreply@github.com 2023-11-05T14:40:08+02:00 llama : mark LLM_ARCH_STARCODER as full offload supported (#3945)
c41ea36eaa3548776de4cb3d5d49b925cd3fc0f2 a7fac013cf1cc7bbc0160a226aa2412e9f22e78a Eve 139727413+netrunnereve@users.noreply.github.com 2023-11-05T08:03:09Z GitHub noreply@github.com 2023-11-05T10:03:09+02:00 cmake : MSVC instruction detection (fixed up #809) (#3923)
a7fac013cf1cc7bbc0160a226aa2412e9f22e78a 48ade94538fa509465d71023e49d07aab0ec8cd5 Eve 139727413+netrunnereve@users.noreply.github.com 2023-11-05T07:46:44Z GitHub noreply@github.com 2023-11-05T09:46:44+02:00 ci : use intel sde when ci cpu doesn't support avx512 (#3949)
48ade94538fa509465d71023e49d07aab0ec8cd5 f28af0d81aa1010afa5de74cf627dcb04bea3157 slaren slarengh@gmail.com 2023-11-05T08:12:13+01:00 GitHub noreply@github.com 2023-11-05T09:12:13+02:00 cuda : revert CUDA pool stuff (#3944)
f28af0d81aa1010afa5de74cf627dcb04bea3157 d9b33fe95bd257b36c84ee5769cc048230067d6f Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-04T16:20:34-06:00 GitHub noreply@github.com 2023-11-04T16:20:34-06:00 gguf-py: Support 01.AI Yi models (#3943)
d9b33fe95bd257b36c84ee5769cc048230067d6f 5ba37461711095c0284233dbd14f0d9010cdbf56 Peter Sugihara peter@campsh.com 2023-11-03T12:18:18-07:00 GitHub noreply@github.com 2023-11-03T21:18:18+02:00 metal : round up to 16 to fix MTLDebugComputeCommandEncoder assertion (#3938)
5ba37461711095c0284233dbd14f0d9010cdbf56 abb77e7319aabc0b5cfb7c22da690a692489b6b7 Xiao-Yong Jin jinxiaoyong@gmail.com 2023-11-03T13:00:31-05:00 GitHub noreply@github.com 2023-11-03T14:00:31-04:00 ggml-metal: fix yarn rope (#3937)
abb77e7319aabc0b5cfb7c22da690a692489b6b7 8f961abdc4e134c83bf8c2ad618ab256b4cae0f9 slaren slarengh@gmail.com 2023-11-03T12:13:09+01:00 GitHub noreply@github.com 2023-11-03T12:13:09+01:00 ggml-cuda : move row numbers to x grid dim in mmv kernels (#3921)
8f961abdc4e134c83bf8c2ad618ab256b4cae0f9 05816027d649f977468fc804cdb54e99eac246d1 Georgi Gerganov ggerganov@gmail.com 2023-11-03T09:41:17+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-03T09:41:56+02:00 speculative : change default p_accept to 0.5 + CLI args (#3919)
05816027d649f977468fc804cdb54e99eac246d1 3fdbe6b66b7b5c6ad3b2f245cbad1517c27ff776 Georgi Gerganov ggerganov@gmail.com 2023-11-03T09:24:00+02:00 GitHub noreply@github.com 2023-11-03T09:24:00+02:00 common : YAYF (yet another YARN fix) (#3925)
3fdbe6b66b7b5c6ad3b2f245cbad1517c27ff776 629f917cd6b96ba1274c49a8aab163b1b189229d cebtenzzre cebtenzzre@gmail.com 2023-11-03T02:31:58-04:00 GitHub noreply@github.com 2023-11-03T08:31:58+02:00 llama : change yarn_ext_factor placeholder to -1 (#3922)
629f917cd6b96ba1274c49a8aab163b1b189229d 51b2fc11f7f605fff49725a4540e9a6ef7b51b70 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-11-02T13:58:22-06:00 GitHub noreply@github.com 2023-11-02T21:58:22+02:00 cuda : add ROCM aliases for CUDA pool stuff (#3918)
51b2fc11f7f605fff49725a4540e9a6ef7b51b70 224e7d5b14cbabab7ae45c64db2cfde979c8455d Andrei abetlen@gmail.com 2023-11-02T15:40:31-04:00 GitHub noreply@github.com 2023-11-02T21:40:31+02:00 cmake : fix relative path to git submodule index (#3915)
224e7d5b14cbabab7ae45c64db2cfde979c8455d c7743fe1c1cbda5a886362aa371480360580fdf0 Georgi Gerganov ggerganov@gmail.com 2023-11-02T20:44:12+02:00 GitHub noreply@github.com 2023-11-02T20:44:12+02:00 readme : add notice about #3912
c7743fe1c1cbda5a886362aa371480360580fdf0 d6069051de7165a4e06662c89257f5d2905bb156 Georgi Gerganov ggerganov@gmail.com 2023-11-02T20:32:11+02:00 GitHub noreply@github.com 2023-11-02T20:32:11+02:00 cuda : fix const ptrs warning causing ROCm build issues (#3913)
d6069051de7165a4e06662c89257f5d2905bb156 4ff1046d75e64f0e556d8dcd930ea25c23eb8b18 Oleksii Maryshchenko oleksii.maryshchenko@gmail.com 2023-11-02T18:10:39+01:00 GitHub noreply@github.com 2023-11-02T19:10:39+02:00 cuda : use CUDA memory pool with async memory allocation/deallocation when available (#3903)
4ff1046d75e64f0e556d8dcd930ea25c23eb8b18 21958bb393a654591ed26f339791b752d58f5c8b Georgi Gerganov ggerganov@gmail.com 2023-11-02T16:22:30+02:00 GitHub noreply@github.com 2023-11-02T16:22:30+02:00 gguf : print error for GGUFv1 files (#3908)
21958bb393a654591ed26f339791b752d58f5c8b 2756c4fbffab097736d5116007872d86456a544a slaren slarengh@gmail.com 2023-11-02T13:10:33+01:00 GitHub noreply@github.com 2023-11-02T14:10:33+02:00 cmake : disable LLAMA_NATIVE by default (#3906)
2756c4fbffab097736d5116007872d86456a544a 1efae9b7dca2a5cc5aa21c1997b538022964ea19 Georgi Gerganov ggerganov@gmail.com 2023-11-02T11:20:21+02:00 GitHub noreply@github.com 2023-11-02T11:20:21+02:00 gguf : remove special-case code for GGUFv1 (#3901)
1efae9b7dca2a5cc5aa21c1997b538022964ea19 b12fa0d1c13596869c512f49a526b979c94787cc Georgi Gerganov ggerganov@gmail.com 2023-11-02T09:54:18+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-02T09:54:44+02:00 llm : prevent from 1-D tensors being GPU split (#3697)
b12fa0d1c13596869c512f49a526b979c94787cc 4d719a6d4e74b9a98e75f826f865f3153717d54b cebtenzzre cebtenzzre@gmail.com 2023-11-02T02:50:16-04:00 GitHub noreply@github.com 2023-11-02T08:50:16+02:00 build : link against build info instead of compiling against it (#3879)
4d719a6d4e74b9a98e75f826f865f3153717d54b 183b3fac6c28e65d23ac0230c1dd6fb84bf0154d Georgi Gerganov ggerganov@gmail.com 2023-11-02T08:35:10+02:00 GitHub noreply@github.com 2023-11-02T08:35:10+02:00 cuda : check if this fixes Pascal card regression (#3882)
183b3fac6c28e65d23ac0230c1dd6fb84bf0154d 2fffa0d61fa10e4b466e78cabcc6a4e16717b580 Georgi Gerganov ggerganov@gmail.com 2023-11-02T08:33:37+02:00 GitHub noreply@github.com 2023-11-02T08:33:37+02:00 metal : fix build errors and kernel sig after #2268 (#3898)
2fffa0d61fa10e4b466e78cabcc6a4e16717b580 0eb332a10f3f14a3746c391bf80ff5e7bdf29d5d cebtenzzre cebtenzzre@gmail.com 2023-11-02T01:49:44-04:00 GitHub noreply@github.com 2023-11-02T07:49:44+02:00 cuda : fix RoPE after #2268 (#3897)
0eb332a10f3f14a3746c391bf80ff5e7bdf29d5d d02e98cde035d91ed8032ab943d1d504fe9da394 cebtenzzre cebtenzzre@gmail.com 2023-11-01T19:29:14-04:00 GitHub noreply@github.com 2023-11-01T19:29:14-04:00 llama : fix llama_context_default_params after #2268 (#3893)
d02e98cde035d91ed8032ab943d1d504fe9da394 898aeca90a9bb992f506234cf3b8b7f7fa28a1df slaren slarengh@gmail.com 2023-11-01T23:10:09+01:00 GitHub noreply@github.com 2023-11-01T23:10:09+01:00 ggml-cuda : compute ptrs for cublasGemmBatchedEx in a kernel (#3891)
898aeca90a9bb992f506234cf3b8b7f7fa28a1df c43c2da8afacaddfe51c09b21dbd9922cd0ea46b cebtenzzre cebtenzzre@gmail.com 2023-11-01T18:04:33-04:00 GitHub noreply@github.com 2023-11-01T18:04:33-04:00 llama : implement YaRN RoPE scaling (#2268)
c43c2da8afacaddfe51c09b21dbd9922cd0ea46b 523e49b11174368cd73460fa5eae7b39d856f300 Georgi Gerganov ggerganov@gmail.com 2023-11-01T23:08:30+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-01T23:08:30+02:00 llm : fix llm_build_kqv taking unused tensor (benign, #3837)
523e49b11174368cd73460fa5eae7b39d856f300 e16b9fa4baa8a09c6619b116159830e898050942 Georgi Gerganov ggerganov@gmail.com 2023-11-01T23:00:50+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-01T23:00:50+02:00 llm : fix falcon norm after refactoring (#3837)
e16b9fa4baa8a09c6619b116159830e898050942 ff8f9a88da0018972dfdf6fe64b5c8992caabd9c Georgi Gerganov ggerganov@gmail.com 2023-11-01T21:25:00+02:00 GitHub noreply@github.com 2023-11-01T21:25:00+02:00 metal : multi-simd softmax (#3710)
ff8f9a88da0018972dfdf6fe64b5c8992caabd9c 50337961a678fce4081554b24e56e86b67660163 Georgi Gerganov ggerganov@gmail.com 2023-11-01T21:15:55+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-01T21:15:55+02:00 common : minor (#3715)
50337961a678fce4081554b24e56e86b67660163 0e40806c1cb3bdf9955ed807ffbe212be85b4c67 Georgi Gerganov ggerganov@gmail.com 2023-11-01T20:11:02+02:00 GitHub noreply@github.com 2023-11-01T20:11:02+02:00 llm : add llm_build_context (#3881)
0e40806c1cb3bdf9955ed807ffbe212be85b4c67 a2758d08e44ce3624d233af4d23c6843e2e735b5 bandoti 141645996+bandoti@users.noreply.github.com 2023-11-01T14:42:01-03:00 GitHub noreply@github.com 2023-11-01T19:42:01+02:00 common : allow caller to handle help/argument exceptions (#3715)
a2758d08e44ce3624d233af4d23c6843e2e735b5 e75dfdd31b6a3dfa0627ba4ac3bb4b36e9db588e staviq staviq@gmail.com 2023-11-01T15:18:27+01:00 GitHub noreply@github.com 2023-11-01T16:18:27+02:00 log : make generating separate log files optional (#3787)
e75dfdd31b6a3dfa0627ba4ac3bb4b36e9db588e 9a3b4f6c86503c9cfc049d4d0fdeafef12806f5e l3utterfly gc.pthzfoldr@gmail.com 2023-11-01T21:40:43+08:00 GitHub noreply@github.com 2023-11-01T15:40:43+02:00 sampling : null grammar field after reset (#3885)
9a3b4f6c86503c9cfc049d4d0fdeafef12806f5e 73bdcb395ef9a997d9c02950c7cd4249546162cd Georgi Gerganov ggerganov@gmail.com 2023-11-01T13:50:45+02:00 Georgi Gerganov ggerganov@gmail.com 2023-11-01T13:50:45+02:00 ggml : fix UNUSED macro (#3762)
73bdcb395ef9a997d9c02950c7cd4249546162cd f0e209324a7f663225791897877bf610f1af152d Andrew Godfrey AndrewGodfrey@users.noreply.github.com 2023-11-01T04:49:04-07:00 GitHub noreply@github.com 2023-11-01T13:49:04+02:00 finetune : add -ngl parameter (#3762)
f0e209324a7f663225791897877bf610f1af152d ca190bca8e844d171020d6147687e71472d71734 Georgi Gerganov ggerganov@gmail.com 2023-11-01T11:29:07+02:00 GitHub noreply@github.com 2023-11-01T11:29:07+02:00 scripts : add server-llm.sh (#3868)
ca190bca8e844d171020d6147687e71472d71734 71e3718abdb2771b50c9606d3a7569623a0b0afe Adrian Hesketh a-h@users.noreply.github.com 2023-11-01T09:28:28Z GitHub noreply@github.com 2023-11-01T11:28:28+02:00 server : re-enable completion and embedded at the same time (#3876)
71e3718abdb2771b50c9606d3a7569623a0b0afe 238657db2364cfb728c694470a4a81702afea760 Georgi Gerganov ggerganov@gmail.com 2023-11-01T08:04:02+02:00 GitHub noreply@github.com 2023-11-01T08:04:02+02:00 llama : refactor graph build code (#3837)
238657db2364cfb728c694470a4a81702afea760 07178c98e1b61a5e2af39d347add12e7eb9e08e1 kalomaze 66376113+kalomaze@users.noreply.github.com 2023-10-31T14:44:49-05:00 GitHub noreply@github.com 2023-10-31T20:44:49+01:00 samplers : Min-P sampler implementation [alternative to Top P/Top K] (#3841)
07178c98e1b61a5e2af39d347add12e7eb9e08e1 207b51900e15cc7f89763a3bb1c565fe11cbb45d Tungsten842 886724vf@anonaddy.me 2023-10-31T18:24:03+01:00 GitHub noreply@github.com 2023-10-31T19:24:03+02:00 flake.nix: fix for rocm 5.7 (#3853)
207b51900e15cc7f89763a3bb1c565fe11cbb45d 6e08281e588bbba1a5d180290a94a43f167f3a1a Georgi Gerganov ggerganov@gmail.com 2023-10-30T19:19:15+02:00 GitHub noreply@github.com 2023-10-30T19:19:15+02:00 ggml : move FP16 <-> FP32 code to ggml-impl.h (#3861)
6e08281e588bbba1a5d180290a94a43f167f3a1a 2046eb4345e62c4575b3cdc0115a51db89f3fb70 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-29T11:31:40-06:00 GitHub noreply@github.com 2023-10-29T11:31:40-06:00 Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843)
2046eb4345e62c4575b3cdc0115a51db89f3fb70 71a09da301705b9c5ad4ca3cf3fbd966dd3f1ec5 cebtenzzre cebtenzzre@gmail.com 2023-10-29T12:33:47-04:00 GitHub noreply@github.com 2023-10-29T18:33:47+02:00 make : remove unnecessary dependency on build-info.h (#3842)
71a09da301705b9c5ad4ca3cf3fbd966dd3f1ec5 d69d777c02b9ac405a95f3cbfba219a990caefff Georgi Gerganov ggerganov@gmail.com 2023-10-29T18:32:51+02:00 GitHub noreply@github.com 2023-10-29T18:32:51+02:00 llama : fix kv shift bug (#3835)
d69d777c02b9ac405a95f3cbfba219a990caefff ff3bad83e29e3009010cbc923bebd769055eaa7f Georgi Gerganov ggerganov@gmail.com 2023-10-29T18:32:28+02:00 GitHub noreply@github.com 2023-10-29T18:32:28+02:00 ggml : quantization refactoring (#3833)
ff3bad83e29e3009010cbc923bebd769055eaa7f 82a6646e0221216c41edcdf99f5a44bb051391f5 Erik Scholz Green-Sky@users.noreply.github.com 2023-10-28T16:41:07+02:00 GitHub noreply@github.com 2023-10-28T16:41:07+02:00 flake : update flake.lock for newer transformers version + provide extra dev shell (#3797)
82a6646e0221216c41edcdf99f5a44bb051391f5 ba231e8a6dd8ad82acfe0e4d492ff7cef6b3f0a1 Aarni Koskela akx@iki.fi 2023-10-28T15:43:01+03:00 GitHub noreply@github.com 2023-10-28T15:43:01+03:00 metal : try cwd for ggml-metal.metal if bundle lookup fails (#3793)
ba231e8a6dd8ad82acfe0e4d492ff7cef6b3f0a1 8a2f2fea2914aaa3f4b2f82800c7de15f15bdb09 Georgi Gerganov ggerganov@gmail.com 2023-10-28T15:25:33+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-28T15:35:26+03:00 issues : change label from bug to bug-unconfirmed (#3748)
8a2f2fea2914aaa3f4b2f82800c7de15f15bdb09 bd6d9e205982b34e0ba2c3b22bbf31a1ef1a1bb5 Georgi Gerganov ggerganov@gmail.com 2023-10-28T15:25:15+03:00 GitHub noreply@github.com 2023-10-28T06:25:15-06:00 convert : ignore tokens if their IDs are within [0, vocab_size) (#3831)
bd6d9e205982b34e0ba2c3b22bbf31a1ef1a1bb5 ee1a0ec9cb367ba41d138134795cbbbe93d2bf1c Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-28T05:54:24-06:00 GitHub noreply@github.com 2023-10-28T14:54:24+03:00 llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747)
ee1a0ec9cb367ba41d138134795cbbbe93d2bf1c 177461104b454163473dced2a5038f4e016cdb7e Georgi Gerganov ggerganov@gmail.com 2023-10-28T14:23:11+03:00 GitHub noreply@github.com 2023-10-28T14:23:11+03:00 llama : add option for greedy sampling with probs (#3813)
177461104b454163473dced2a5038f4e016cdb7e fdee152e4eebb78c191df0b074857111d7f2aba7 Henk Poley HenkPoley@gmail.com 2023-10-28T12:16:33+02:00 GitHub noreply@github.com 2023-10-28T13:16:33+03:00 common : print that one line of the syntax help *also* to standard output (#3823)
fdee152e4eebb78c191df0b074857111d7f2aba7 41aee4df821854f37d90a45281f03b6db8d27de2 Georgi Gerganov ggerganov@gmail.com 2023-10-28T12:06:08+03:00 GitHub noreply@github.com 2023-10-28T12:06:08+03:00 starcoder : add GPU offloading (#3827)
41aee4df821854f37d90a45281f03b6db8d27de2 6d459cbfbe5a011dfca94f9550527a504b6f9aa1 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-27T15:40:07-06:00 GitHub noreply@github.com 2023-10-28T00:40:07+03:00 speculative : ensure draft and target model vocab matches (#3812)
6d459cbfbe5a011dfca94f9550527a504b6f9aa1 c8d6a1f34ab6f1b6bd468d256e535a61f98f114c cebtenzzre cebtenzzre@gmail.com 2023-10-27T17:33:53-04:00 GitHub noreply@github.com 2023-10-27T17:33:53-04:00 llama : correctly report GGUFv3 format (#3818)
c8d6a1f34ab6f1b6bd468d256e535a61f98f114c 2f9ec7e271220a78fe27c9e6ccbcc0dda31cda0f Thibault Terrasson thibault.terrasson@gmail.com 2023-10-27T16:37:41+02:00 GitHub noreply@github.com 2023-10-27T08:37:41-06:00 simple : fix batch handling (#3803)
2f9ec7e271220a78fe27c9e6ccbcc0dda31cda0f 34b2a5e1ee4fe6295fb4420eb91131d743694c65 Georgi Gerganov ggerganov@gmail.com 2023-10-27T17:01:23+03:00 GitHub noreply@github.com 2023-10-27T17:01:23+03:00 cuda : improve text-generation and batched decoding performance (#3776)
34b2a5e1ee4fe6295fb4420eb91131d743694c65 6961c4bd0b5176e10ab03b35394f1e9eab761792 Georgi Gerganov ggerganov@gmail.com 2023-10-26T22:53:37+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-26T22:54:17+03:00 server : do not release slot on image input (#3798)
6961c4bd0b5176e10ab03b35394f1e9eab761792 cc448774866e6479c750bd7c135cd8f92cedee67 Georgi Gerganov ggerganov@gmail.com 2023-10-25T10:26:27+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-25T10:26:27+03:00 batched-bench : print params at start
cc448774866e6479c750bd7c135cd8f92cedee67 ad939626577cd25b462e8026cc543efb71528472 Georgi Gerganov ggerganov@gmail.com 2023-10-25T10:09:16+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-25T10:09:16+03:00 log : disable pid in log filenames
ad939626577cd25b462e8026cc543efb71528472 1717521cdb976a2219888b0e5cba36e210eee9df cebtenzzre cebtenzzre@gmail.com 2023-10-24T16:10:43-04:00 GitHub noreply@github.com 2023-10-24T23:10:43+03:00 server : add parameter -tb N, --threads-batch N (#3584) (#3768)
1717521cdb976a2219888b0e5cba36e210eee9df b2f7e04bd312eaf97eee0523aa09d950d585626b Georgi Gerganov ggerganov@gmail.com 2023-10-24T23:08:20+03:00 GitHub noreply@github.com 2023-10-24T23:08:20+03:00 server : do not block system prompt update (#3767)
b2f7e04bd312eaf97eee0523aa09d950d585626b abd21fc99f1d35e2081e4c01dc09c71a86bf3c5a Georgi Gerganov ggerganov@gmail.com 2023-10-24T21:51:20+03:00 GitHub noreply@github.com 2023-10-24T21:51:20+03:00 sync : ggml (conv ops + cuda MSVC fixes) (#3765)
abd21fc99f1d35e2081e4c01dc09c71a86bf3c5a 2b4ea35e56792064598e922e46d081e02bc96b94 John Smith 67539080+kingsidelee@users.noreply.github.com 2023-10-25T01:48:45+08:00 GitHub noreply@github.com 2023-10-24T20:48:45+03:00 cmake : add missed dependencies (#3763)
2b4ea35e56792064598e922e46d081e02bc96b94 daab3d7f45832e10773c99f3484b0d5b14d86c0c Georgi Gerganov ggerganov@gmail.com 2023-10-24T16:48:37+03:00 GitHub noreply@github.com 2023-10-24T16:48:37+03:00 cuda : add batched cuBLAS GEMM for faster attention (#3749)
daab3d7f45832e10773c99f3484b0d5b14d86c0c 469c9addef75893e6be12edda852d12e840bf064 Galunid karolek1231456@gmail.com 2023-10-24T09:17:17+02:00 GitHub noreply@github.com 2023-10-24T09:17:17+02:00 Add more tokenizer tests (#3742)
469c9addef75893e6be12edda852d12e840bf064 e3932593d46c30145301a13097895f9376cba509 Georgi Gerganov ggerganov@gmail.com 2023-10-24T09:46:50+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-24T09:47:22+03:00 metal : handle ggml_scale for n%4 != 0 (close #3754)
e3932593d46c30145301a13097895f9376cba509 9d02956443e5c1ded29b7b5ed8a21bc01ba6f563 Georgi Gerganov ggerganov@gmail.com 2023-10-23T23:46:05+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-23T23:46:05+03:00 Revert "make : add optional CUDA_NATIVE_ARCH (#2482)"
9d02956443e5c1ded29b7b5ed8a21bc01ba6f563 69a6735087c3634963c642fd69f0851ac479cd78 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-10-23T22:57:16+03:00 GitHub noreply@github.com 2023-10-23T22:57:16+03:00 issues : separate bug and enhancement template + no default title (#3748)
69a6735087c3634963c642fd69f0851ac479cd78 5be6c803fa5378f62a1590f3ad8c6b64c7c0c2ce Galunid karolek1231456@gmail.com 2023-10-23T21:46:00+02:00 GitHub noreply@github.com 2023-10-23T21:46:00+02:00 Update special token handling in conversion scripts for gpt2 derived tokenizers (#3746)
5be6c803fa5378f62a1590f3ad8c6b64c7c0c2ce 6336701c9378c23c85d1c0e464b663ca2bbb8e60 Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2023-10-23T12:40:03-07:00 GitHub noreply@github.com 2023-10-23T22:40:03+03:00 llama : remove token functions with `context` args in favor of `model` (#3720)
6336701c9378c23c85d1c0e464b663ca2bbb8e60 96981f37b1e3f450d9e63e571514217bf60f0a7f Galunid karolek1231456@gmail.com 2023-10-23T17:47:03+02:00 GitHub noreply@github.com 2023-10-23T17:47:03+02:00 Fix baichuan convert script not detecing model (#3739)
96981f37b1e3f450d9e63e571514217bf60f0a7f 438c2ca83045a00ef244093d27e9ed41a8cb4ea9 Alex awhill19@icloud.com 2023-10-22T15:56:53-04:00 GitHub noreply@github.com 2023-10-22T22:56:53+03:00 make : add optional CUDA_NATIVE_ARCH (#2482)
438c2ca83045a00ef244093d27e9ed41a8cb4ea9 9e70cc03229df19ca2d28ce23cc817198f897278 Georgi Gerganov ggerganov@gmail.com 2023-10-22T22:53:08+03:00 GitHub noreply@github.com 2023-10-22T22:53:08+03:00 server : parallel decoding and multimodal (#3677)
9e70cc03229df19ca2d28ce23cc817198f897278 5a42a5f8e8a86da9ac88008d748cf232a83aa0e1 goerch jhr.walter@t-online.de 2023-10-22T21:21:42+02:00 GitHub noreply@github.com 2023-10-22T21:21:42+02:00 Add test for MPT tokenization (#3728)
5a42a5f8e8a86da9ac88008d748cf232a83aa0e1 a5e7dbd6141128bfa3c40a19c2945a181df625d3 Ian Scrivener github@zilogy.asia 2023-10-23T05:16:43+11:00 GitHub noreply@github.com 2023-10-22T21:16:43+03:00 readme : remove unsupported node.js library (#3703)
a5e7dbd6141128bfa3c40a19c2945a181df625d3 d3956aea53369455008159cc405ed4c496976692 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-22T12:14:56-06:00 GitHub noreply@github.com 2023-10-22T21:14:56+03:00 llama : validate special token ids are in range when loading GGUF model (#3635)
d3956aea53369455008159cc405ed4c496976692 22c69a27945e7acf9690dd3210d316f22182751c vvhg1 94630311+vvhg1@users.noreply.github.com 2023-10-22T20:09:51+02:00 GitHub noreply@github.com 2023-10-22T21:09:51+03:00 main : escape prompt for cfg_negative_prompt and consecutive inputs in main with interactive (#3623)
22c69a27945e7acf9690dd3210d316f22182751c 465219b9143ac01db0990bbcb0a081ef72ec2008 Georgi Gerganov ggerganov@gmail.com 2023-10-22T08:37:20+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-22T08:37:20+03:00 batched : add len CLI argument
465219b9143ac01db0990bbcb0a081ef72ec2008 d1031cf49c3b958b915fd558e23453471c29ac33 shibe2 shibe@tuta.io 2023-10-12T16:01:23+04:00 shibe2 shibe@tuta.io 2023-10-20T22:30:52+04:00 CLBlast: Add outer loops over src0 for broadcasting in mulmat
d1031cf49c3b958b915fd558e23453471c29ac33 8cf19d60dc93809db8e51fedc811595eed9134c5 Georgi Gerganov ggerganov@gmail.com 2023-10-20T21:07:23+03:00 GitHub noreply@github.com 2023-10-20T21:07:23+03:00 sampling : refactor init to use llama_sampling_params (#3696)
8cf19d60dc93809db8e51fedc811595eed9134c5 a0edf73bda31c7c4e649e6f07c6fd30a729929cd Qin Yue Chen 71813199+chenqiny@users.noreply.github.com 2023-10-20T06:19:40-05:00 GitHub noreply@github.com 2023-10-20T14:19:40+03:00 gguf : support big endian platform (#3552)
a0edf73bda31c7c4e649e6f07c6fd30a729929cd f439e506e8ae8b01df2ae2156380f8156d7553e3 Georgi Gerganov ggerganov@gmail.com 2023-10-20T13:06:10+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-20T13:06:10+03:00 server : fix uninitialized sampling context (close #3685)
f439e506e8ae8b01df2ae2156380f8156d7553e3 e78f3ef24af4ca74e77e725644b41ae8ca3b10a5 Herman Semenov GermanAizek@yandex.ru 2023-10-20T10:02:12Z GitHub noreply@github.com 2023-10-20T13:02:12+03:00 ggml : fix rope + llama minor optimizations (#3560)
e78f3ef24af4ca74e77e725644b41ae8ca3b10a5 f3b25e40438b3c8383caabf4e7b89863145a9f0e cebtenzzre cebtenzzre@gmail.com 2023-10-20T01:32:08-04:00 GitHub noreply@github.com 2023-10-20T08:32:08+03:00 convert : restore compat with old Falcon models (#3680)
f3b25e40438b3c8383caabf4e7b89863145a9f0e 60abea9798f47b918a9f38c66edfd88c526d20f6 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-10-19T19:40:41+03:00 GitHub noreply@github.com 2023-10-19T19:40:41+03:00 multimodal : add BakLLaVA conversion support (#3682)
60abea9798f47b918a9f38c66edfd88c526d20f6 004797f6ac135383f8c1d1f5bd415ddee2f79318 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-10-19T16:59:11+03:00 GitHub noreply@github.com 2023-10-19T16:59:11+03:00 llava : avoid segfault in case of non-existent mmproj file (#3674)
004797f6ac135383f8c1d1f5bd415ddee2f79318 4e82b2ea3fa6482915d147bc9f46e70b9ada7700 Georgi Gerganov ggerganov@gmail.com 2023-10-18T21:44:43+03:00 GitHub noreply@github.com 2023-10-18T21:44:43+03:00 readme : update hot topics
4e82b2ea3fa6482915d147bc9f46e70b9ada7700 0e89203b517c95ec6675eda75d200a60d1e8921d Georgi Gerganov ggerganov@gmail.com 2023-10-18T18:49:40+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-18T18:49:40+03:00 speculative : bug fixes
0e89203b517c95ec6675eda75d200a60d1e8921d c67fe68e417f766970fb1feaf2e66458aa24116a Georgi Gerganov ggerganov@gmail.com 2023-10-18T16:21:57+03:00 GitHub noreply@github.com 2023-10-18T16:21:57+03:00 speculative : add tree-based sampling example (#3624)
c67fe68e417f766970fb1feaf2e66458aa24116a 1117d06607d2d885640ac501f05f0aae5494e2c5 Jhen-Jie Hong iainst0409@gmail.com 2023-10-18T07:21:48-05:00 GitHub noreply@github.com 2023-10-18T15:21:48+03:00 metal : implement q5_0 and q5_1 kernels (#3648)
1117d06607d2d885640ac501f05f0aae5494e2c5 cb33f43a2a9f5a5a5f8d290dd97c625d9ba97a2f shibe2 shibe@tuta.io 2023-10-18T16:09:22+04:00 GitHub noreply@github.com 2023-10-18T15:09:22+03:00 opencl : fix element-wise multiplication (#3656)
cb33f43a2a9f5a5a5f8d290dd97c625d9ba97a2f e1675d133c31e1c8de2f06be7164e12c0ba6cf2c slaren slarengh@gmail.com 2023-10-17T22:24:50+02:00 GitHub noreply@github.com 2023-10-17T22:24:50+02:00 fix embeddings when using CUDA (#3657)
e1675d133c31e1c8de2f06be7164e12c0ba6cf2c 8402566a7c436bfbde8e7b0461faee50298106a0 Georgi Gerganov ggerganov@gmail.com 2023-10-17T22:34:26+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-17T22:34:26+03:00 llama : avoid fprintf in favor of LLAMA_LOG (#3538)
8402566a7c436bfbde8e7b0461faee50298106a0 40e5ce054f4c4fa555e4510ea5f760bb29185332 BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2023-10-17T14:13:21-04:00 GitHub noreply@github.com 2023-10-17T21:13:21+03:00 readme : update hot-topics & models, detail windows release in usage (#3615)
40e5ce054f4c4fa555e4510ea5f760bb29185332 a5e8c1d8c71f01d98ae2ec63a57c118664f9764d shibe2 shibe@tuta.io 2023-10-11T21:30:06+04:00 shibe2 shibe@tuta.io 2023-10-17T21:02:30+04:00 CLBlast: Fix temporary buffer size for f16 conversion (wsize)
a5e8c1d8c71f01d98ae2ec63a57c118664f9764d e74c705e15cd228ad696c4a3cdea6d6fb4ff434c slaren slarengh@gmail.com 2023-10-17T19:00:58+02:00 GitHub noreply@github.com 2023-10-17T20:00:58+03:00 train-text-from-scratch : fix assert failure in ggml-alloc (#3618)
e74c705e15cd228ad696c4a3cdea6d6fb4ff434c 3ad1e3f1a10c1f66b4f1cd7510e0977fadbc0dfd Georgi Gerganov ggerganov@gmail.com 2023-10-17T19:52:53+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-17T19:52:53+03:00 editorconfig : remove trailing spaces
3ad1e3f1a10c1f66b4f1cd7510e0977fadbc0dfd 1142013da40e98946a109b141dd858f0ed996051 coezbek c.oezbek@gmail.com 2023-10-17T18:51:02+02:00 GitHub noreply@github.com 2023-10-17T19:51:02+03:00 server : documentation of JSON return value of /completion endpoint (#3632)
1142013da40e98946a109b141dd858f0ed996051 5fe268a4d9ce09f3a6c77239af583d3a8e49d54c Georgi Gerganov ggerganov@gmail.com 2023-10-17T19:12:46+03:00 GitHub noreply@github.com 2023-10-17T19:12:46+03:00 save-load-state : fix example + add ci test (#3655)
5fe268a4d9ce09f3a6c77239af583d3a8e49d54c 1a159553f921a9209fed8c714494e57b3649f232 ldwang ftgreat@163.com 2023-10-17T23:52:33+08:00 GitHub noreply@github.com 2023-10-17T18:52:33+03:00 readme : add Aquila2 links (#3610)
1a159553f921a9209fed8c714494e57b3649f232 281ef73c258cc1eebec8a64264240432d5878c4b staviq staviq@gmail.com 2023-10-17T17:11:01+02:00 GitHub noreply@github.com 2023-10-17T18:11:01+03:00 tokenizer : special token handling (#3538)
281ef73c258cc1eebec8a64264240432d5878c4b 940efa95fec0b8a98c226a889d2ad839dfeeae0d Georgi Gerganov ggerganov@gmail.com 2023-10-17T09:19:28+03:00 GitHub noreply@github.com 2023-10-17T09:19:28+03:00 k-quants : fix quantization ranges (#3646)
940efa95fec0b8a98c226a889d2ad839dfeeae0d 11bff290458f12f020b588792707f76ec658a27a Georgi Gerganov ggerganov@gmail.com 2023-10-16T23:58:00+03:00 GitHub noreply@github.com 2023-10-16T23:58:00+03:00 llava : fix tokenization to not add bos between image embeddings and user prompt (#3645)
11bff290458f12f020b588792707f76ec658a27a 11dc1091f64b24ca6d643acc6d0051117ba60161 cebtenzzre cebtenzzre@gmail.com 2023-10-15T02:32:06-04:00 GitHub noreply@github.com 2023-10-15T09:32:06+03:00 MPT : support GQA for replit-code-v1.5 (#3627)
11dc1091f64b24ca6d643acc6d0051117ba60161 2a4bcbacead886996f175f33479d1d874a3e577f M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-10-14T13:52:44+03:00 GitHub noreply@github.com 2023-10-14T04:52:44-06:00 Honor -ngl option for Cuda offloading in llava (#3621)
2a4bcbacead886996f175f33479d1d874a3e577f 424b6381c4daeed62e6600e0402e72f39845b58d Daniel Bevenius daniel.bevenius@gmail.com 2023-10-13T12:33:16+02:00 GitHub noreply@github.com 2023-10-13T13:33:16+03:00 llama : remove n_threads from llama_decode_internal (#3614)
424b6381c4daeed62e6600e0402e72f39845b58d 1e0e873c373c33989beb6bc64d83cd572ab7fe2b slaren slarengh@gmail.com 2023-10-13T12:23:10+02:00 GitHub noreply@github.com 2023-10-13T12:23:10+02:00 ggml : add context enumeration functions (#3605)
1e0e873c373c33989beb6bc64d83cd572ab7fe2b 370359e5baf619f3a8d461023143d1494b1e8fde shibe2 shibe@tuta.io 2023-10-12T23:59:47+04:00 GitHub noreply@github.com 2023-10-12T21:59:47+02:00 CLBlast: Fix matrix-vector multiplication (#3544)
370359e5baf619f3a8d461023143d1494b1e8fde 9e24cc6e2e589d405bd1720c400f5b0b9d0ca3ee M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-10-12T18:23:18+03:00 GitHub noreply@github.com 2023-10-12T18:23:18+03:00 examples: support LLaVA v1.5 (multimodal model) (#3436)
9e24cc6e2e589d405bd1720c400f5b0b9d0ca3ee d28e572c0270eb72649dbcc3a347e36c05c2934a uint256_t maekawatoshiki1017@gmail.com 2023-10-12T22:36:16+09:00 GitHub noreply@github.com 2023-10-12T16:36:16+03:00 docs : fix typo GOMP_CPU_AFFINITY (#3597)
d28e572c0270eb72649dbcc3a347e36c05c2934a f3040beaab5228b1a9dfe5675a200379478f7204 Georgi Gerganov ggerganov@gmail.com 2023-10-12T14:31:05+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-12T14:31:05+03:00 cmake : fix add_compile_options on macOS
f3040beaab5228b1a9dfe5675a200379478f7204 1a8c8795d64b04df96c28f29faac2d6e256f53bc Ian Scrivener github@zilogy.asia 2023-10-12T22:10:50+11:00 GitHub noreply@github.com 2023-10-12T14:10:50+03:00 typo : it is `--n-gpu-layers` not `--gpu-layers` (#3592)
1a8c8795d64b04df96c28f29faac2d6e256f53bc b016596d903641f8825cd94bb6742e1de0c21017 Georgi Gerganov ggerganov@gmail.com 2023-10-12T13:44:56+03:00 GitHub noreply@github.com 2023-10-12T13:44:56+03:00 ci : check if there is enough VRAM (#3596)
b016596d903641f8825cd94bb6742e1de0c21017 6b3ae4da92485f979a0f45774fcf68597634db0b Aarni Koskela akx@iki.fi 2023-10-12T15:51:53+09:00 GitHub noreply@github.com 2023-10-12T09:51:53+03:00 server : add completion mode (no chat) (#3582)
6b3ae4da92485f979a0f45774fcf68597634db0b 57dd55e2c742bfc50e0f5c6fb95c14118cff44f6 Georgi Gerganov ggerganov@gmail.com 2023-10-12T09:35:19+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-12T09:35:30+03:00 prompts : add mnemonics.txt
57dd55e2c742bfc50e0f5c6fb95c14118cff44f6 b8fe4b5cc9cb237ca98e5bc51b5d189e3c446d13 Georgi Gerganov ggerganov@gmail.com 2023-10-12T09:29:04+03:00 GitHub noreply@github.com 2023-10-12T09:29:04+03:00 server : fix kv cache management (#3588)
b8fe4b5cc9cb237ca98e5bc51b5d189e3c446d13 a8bdd65525ae86dea905e9866ad369b53e30ac14 Georgi Gerganov ggerganov@gmail.com 2023-10-11T23:55:08+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-11T23:55:41+03:00 main : fix session loading bug (#3400)
a8bdd65525ae86dea905e9866ad369b53e30ac14 70c29da118cdb02bfcbd0376c32b5b2236e48e48 Michael Coppola m18coppola@gmail.com 2023-10-11T15:42:22-04:00 GitHub noreply@github.com 2023-10-11T22:42:22+03:00 server : add parameter -tb N, --threads-batch N (#3584)
70c29da118cdb02bfcbd0376c32b5b2236e48e48 8c70a5ff25964f0a81e20d142a2f5ac5baff22fc Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-11T13:35:46-06:00 GitHub noreply@github.com 2023-10-11T22:35:46+03:00 common : fix mirostat state when using multiple sequences (#3543)
8c70a5ff25964f0a81e20d142a2f5ac5baff22fc 24ba3d829e31a6eda3fa1723f692608c2fa3adda Georgi Gerganov ggerganov@gmail.com 2023-10-11T21:25:33+03:00 GitHub noreply@github.com 2023-10-11T21:25:33+03:00 batched : add bench tool (#3545)
24ba3d829e31a6eda3fa1723f692608c2fa3adda 9f6ede19f3cfa50d4a51a5babb056c3f8a450b80 Zane Shannon z@zcs.me 2023-10-11T04:14:05-07:00 GitHub noreply@github.com 2023-10-11T06:14:05-05:00 examples : add batched.swift + improve CI for swift (#3562)
9f6ede19f3cfa50d4a51a5babb056c3f8a450b80 233fc1c69f6f415f35363e18a755f9610e89161b Galunid karolek1231456@gmail.com 2023-10-11T01:02:49+02:00 GitHub noreply@github.com 2023-10-10T19:02:49-04:00 Add MPT model to supported models in README.md (#3574)
233fc1c69f6f415f35363e18a755f9610e89161b c5b49360d0d9e49f32e05a9116e90bd0b39a282d goerch jhr.walter@t-online.de 2023-10-10T18:59:52+02:00 GitHub noreply@github.com 2023-10-10T18:59:52+02:00 Minor improvements in GPT2 tokenizer (#3567)
c5b49360d0d9e49f32e05a9116e90bd0b39a282d 02d2875deff28599c6c2c6e1886fab002ffe43b1 Xingchen Song(宋星辰) xingchensong1996@163.com 2023-10-11T00:28:50+08:00 GitHub noreply@github.com 2023-10-10T19:28:50+03:00 readme : add bloom (#3570)
02d2875deff28599c6c2c6e1886fab002ffe43b1 0aa6595ae02f97f2e5ffd74bf57a8b21ac83b272 Xingchen Song(宋星辰) xingchensong1996@163.com 2023-10-10T22:48:21+08:00 GitHub noreply@github.com 2023-10-10T17:48:21+03:00 llm : add bloom models (#3553)
0aa6595ae02f97f2e5ffd74bf57a8b21ac83b272 f5f9121de140eff558f13b5c5e78a3a3b6b94377 Jhen-Jie Hong iainst0409@gmail.com 2023-10-10T06:31:13-05:00 GitHub noreply@github.com 2023-10-10T14:31:13+03:00 swift : improvements and fixes (#3564)
f5f9121de140eff558f13b5c5e78a3a3b6b94377 11ea5c7d96f2c28e1c99659e08ec0a44574056e2 Jan Ploski jpl@plosquare.com 2023-10-10T09:50:23+02:00 GitHub noreply@github.com 2023-10-10T10:50:23+03:00 llm : add MPT support (#3417)
11ea5c7d96f2c28e1c99659e08ec0a44574056e2 95bd60a0a69f57e9a2ff1269667ea484a1a9bb40 vvhg1 94630311+vvhg1@users.noreply.github.com 2023-10-10T09:31:21+02:00 GitHub noreply@github.com 2023-10-10T10:31:21+03:00 infill. : fix tokenization (#3508)
95bd60a0a69f57e9a2ff1269667ea484a1a9bb40 fcca0a700487999d52a525c96d6661e9f6a8703a slaren slarengh@gmail.com 2023-10-09T14:44:58+02:00 GitHub noreply@github.com 2023-10-09T15:44:58+03:00 ggml-alloc : fix assert in debug builds (#3555)
fcca0a700487999d52a525c96d6661e9f6a8703a dcc09d25961c5d0626bc148e558ee841141748f7 Georgi Gerganov ggerganov@gmail.com 2023-10-09T14:32:17+03:00 GitHub noreply@github.com 2023-10-09T14:32:17+03:00 refact : fix convert script + zero out KV cache to avoid nans (#3523)
dcc09d25961c5d0626bc148e558ee841141748f7 db3abcc114d5d1790ba814aa1a80ac673d4ccc3e Georgi Gerganov ggerganov@gmail.com 2023-10-09T14:28:27+03:00 GitHub noreply@github.com 2023-10-09T14:28:27+03:00 metal : do not use mul_mm kernels when ne00 < 64 (#3542)
db3abcc114d5d1790ba814aa1a80ac673d4ccc3e eee42c670e6fa6df9cf17e7ffc319f74cbd81354 Georgi Gerganov ggerganov@gmail.com 2023-10-08T20:19:14+03:00 GitHub noreply@github.com 2023-10-08T20:19:14+03:00 sync : ggml (ggml-backend) (#3548)
eee42c670e6fa6df9cf17e7ffc319f74cbd81354 8e6716a102e390e930594d51302730184dac83cc Matheus C. França matheus-catarino@hotmail.com 2023-10-08T10:59:20-03:00 GitHub noreply@github.com 2023-10-08T16:59:20+03:00 ci : add Zig CI/CD and fix build (#2996)
8e6716a102e390e930594d51302730184dac83cc 9c38d181d40b9d27f8f42152c18e7c70bfffcf37 Ryder Wishart ryderwishart@gmail.com 2023-10-08T03:55:58-07:00 GitHub noreply@github.com 2023-10-08T13:55:58+03:00 api_like_OAI.py : compat with Microsoft Guidance (#2746)
9c38d181d40b9d27f8f42152c18e7c70bfffcf37 a1202a31ed8c35705bd09fe91c3e7410c619bd70 arcrank arcrank@gmail.com 2023-10-08T06:52:57-04:00 GitHub noreply@github.com 2023-10-08T13:52:57+03:00 api_like_OAI.py : simplify function (#2796)
a1202a31ed8c35705bd09fe91c3e7410c619bd70 94e502dfb79430870b42b8e8ee132b4aaa93e4a8 Johannes Rudolph johannes.rudolph@gmail.com 2023-10-08T12:21:19+02:00 GitHub noreply@github.com 2023-10-08T13:21:19+03:00 k-quants : fix comments about block sizing (#3499)
94e502dfb79430870b42b8e8ee132b4aaa93e4a8 7d8b24932fe788a4cda76459a0c5df3e0073cb98 Georgi Gerganov ggerganov@gmail.com 2023-10-08T11:24:50+03:00 GitHub noreply@github.com 2023-10-08T11:24:50+03:00 ci : enable on obj-c changes + fix metal build (#3540)
7d8b24932fe788a4cda76459a0c5df3e0073cb98 b0ec5218c3d24755786b80ecce9cf4ffc07583f8 Luo Tian lt@basecity.com 2023-10-08T16:24:01+08:00 GitHub noreply@github.com 2023-10-08T11:24:01+03:00 zig : fix build by introducing train.cpp (#3539)
b0ec5218c3d24755786b80ecce9cf4ffc07583f8 63d3b06a4318329f92b078e8aa0be7ab6e9f871f Georgi Gerganov ggerganov@gmail.com 2023-10-08T10:01:53+03:00 GitHub noreply@github.com 2023-10-08T10:01:53+03:00 metal : support MTLGPUFamily < Apple7, formatting, style (#3524)
63d3b06a4318329f92b078e8aa0be7ab6e9f871f a16e89cec83b4bd5f6af8f1ce1400f94c12356f9 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-07T23:22:17-06:00 GitHub noreply@github.com 2023-10-08T08:22:17+03:00 llama : fix missing break in Persimmon arch case statements (#3535)
a16e89cec83b4bd5f6af8f1ce1400f94c12356f9 4d0383321184aadf91968d9e3c6a45286ed2473b Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-07T15:31:41-06:00 GitHub noreply@github.com 2023-10-07T15:31:41-06:00 Fix trying to strip newline from empty prompt and cfg prompt file content (#3534)
4d0383321184aadf91968d9e3c6a45286ed2473b c47066d833c6c112e0d23342aa62c3250dd33c81 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-10-07T22:14:10+03:00 GitHub noreply@github.com 2023-10-07T22:14:10+03:00 gguf.py : fix CI for publishing GGUF package (#3532)
c47066d833c6c112e0d23342aa62c3250dd33c81 f1782c68de13b64bb5283fc2038f584e47be9fd2 Tom C tom.corelis@gmail.com 2023-10-07T02:56:15-07:00 GitHub noreply@github.com 2023-10-07T12:56:15+03:00 py : change version of numpy requirement to 1.24.4 (#3515)
f1782c68de13b64bb5283fc2038f584e47be9fd2 c26765a0a148b47e5b541df32438c3ad2a0a8314 cebtenzzre cebtenzzre@gmail.com 2023-10-07T04:41:52-04:00 GitHub noreply@github.com 2023-10-07T11:41:52+03:00 quantize : fail fast on write errors (#3521)
c26765a0a148b47e5b541df32438c3ad2a0a8314 0e797c2fc571b866090f7d60ac7d39d8533593f2 Jhen-Jie Hong iainst0409@gmail.com 2023-10-07T03:40:27-05:00 GitHub noreply@github.com 2023-10-07T11:40:27+03:00 metal : support default.metallib load & reuse code for swift package (#3522)
0e797c2fc571b866090f7d60ac7d39d8533593f2 3a716b4dae545c3db307594fbc509a95d3e21b6e Phillip Kravtsov phillip@kravtsov.net 2023-10-07T00:12:43-07:00 GitHub noreply@github.com 2023-10-07T10:12:43+03:00 llm : support Adept Persimmon 8B (#3410)
3a716b4dae545c3db307594fbc509a95d3e21b6e 1faaae8c2bdc4a21302e367e0754c3fe74a8113e goerch jhr.walter@t-online.de 2023-10-07T06:57:01+02:00 GitHub noreply@github.com 2023-10-07T06:57:01+02:00 Fix for #3454 (#3455)
1faaae8c2bdc4a21302e367e0754c3fe74a8113e cb13d73a720c42d1958bff79b6869d77b26b8cea BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2023-10-06T15:13:36-04:00 GitHub noreply@github.com 2023-10-06T22:13:36+03:00 readme : update models, cuda + ppl instructions (#3510)
cb13d73a720c42d1958bff79b6869d77b26b8cea 9ca79d5cbbc8d43f2bff951404b6a40ff1ee3788 Mihai mihai.chirculescu@yahoo.com 2023-10-06T21:39:33+03:00 GitHub noreply@github.com 2023-10-06T21:39:33+03:00 server : docs fix default values and add n_probs (#3506)
9ca79d5cbbc8d43f2bff951404b6a40ff1ee3788 0c731ca4039ccff86ffab90eaae4ca98037c4496 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-06T10:10:13-06:00 GitHub noreply@github.com 2023-10-06T10:10:13-06:00 kv cache slot search improvements (#3493)
0c731ca4039ccff86ffab90eaae4ca98037c4496 a8777ad84e00cda0399e827cdf971e2c3fab1da2 Georgi Gerganov ggerganov@gmail.com 2023-10-06T16:35:55+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-06T16:36:32+03:00 prompts : fix editorconfig checks after #3416
a8777ad84e00cda0399e827cdf971e2c3fab1da2 97af49fa395df77e4c18af0e1655b2fee67c9686 pudepiedj pudepiedj@gmail.com 2023-10-06T14:16:38+01:00 GitHub noreply@github.com 2023-10-06T16:16:38+03:00 parallel : add option to load external prompt file (#3416)
97af49fa395df77e4c18af0e1655b2fee67c9686 16820a5a0d885113f21021ce934f0b0027b9d69a Jhen-Jie Hong iainst0409@gmail.com 2023-10-06T07:44:24-05:00 GitHub noreply@github.com 2023-10-06T15:44:24+03:00 server : reuse llama_sample_token common util (#3494)
16820a5a0d885113f21021ce934f0b0027b9d69a 04b2f4386eda0264287156104cbf9d1b87895422 l3utterfly gc.pthzfoldr@gmail.com 2023-10-06T18:47:59+08:00 GitHub noreply@github.com 2023-10-06T13:47:59+03:00 llama : correct hparams comparison (#3446)
04b2f4386eda0264287156104cbf9d1b87895422 48edda30ee545fdac2e7a33d505382888f748bbf Jhen-Jie Hong iainst0409@gmail.com 2023-10-06T05:36:43-05:00 GitHub noreply@github.com 2023-10-06T13:36:43+03:00 ci : fix xcodebuild destinations (#3491)
48edda30ee545fdac2e7a33d505382888f748bbf 45eba9369fbcbd7f677eba9a2d3e4ffcfdc81824 cebtenzzre cebtenzzre@gmail.com 2023-10-05T15:00:34-04:00 GitHub noreply@github.com 2023-10-05T15:00:34-04:00 convert : update Falcon script for new HF config (#3448)
45eba9369fbcbd7f677eba9a2d3e4ffcfdc81824 acec9eaaa93315711c11d15afa8d245d164b7cff Kenvix ⭐ kenvixzure@live.com 2023-10-06T01:16:39+08:00 GitHub noreply@github.com 2023-10-05T20:16:39+03:00 build : use std::make_tuple() for compatibility with older GCC versions (#3488)
acec9eaaa93315711c11d15afa8d245d164b7cff e2583cbc29cd7d6d1403f338842c07dfc0467e6c staviq staviq@gmail.com 2023-10-05T18:17:29+02:00 GitHub noreply@github.com 2023-10-05T19:17:29+03:00 common : process escape sequences in reverse prompts (#3461)
e2583cbc29cd7d6d1403f338842c07dfc0467e6c e8b8d32e8663ffc55a02c9721af3a5190382cbb0 shibe2 shibe@tuta.io 2023-10-05T15:57:03+04:00 shibe2 shibe@tuta.io 2023-10-05T18:25:23+04:00 CLBlast: Fix handling of on-device tensor data
e8b8d32e8663ffc55a02c9721af3a5190382cbb0 8f3a642ec1d878b2d0a0d15e3a4277f522790d4c Jhen-Jie Hong iainst0409@gmail.com 2023-10-05T09:02:55-05:00 GitHub noreply@github.com 2023-10-05T17:02:55+03:00 server : fix incorrect num_tokens_predicted (#3480)
8f3a642ec1d878b2d0a0d15e3a4277f522790d4c 0745384449fe8d89d6d99c93153569079e853247 Jhen-Jie Hong iainst0409@gmail.com 2023-10-05T09:00:07-05:00 GitHub noreply@github.com 2023-10-05T17:00:07+03:00 swift : disable ACCELERATE_NEW_LAPACK (#3481)
0745384449fe8d89d6d99c93153569079e853247 019ba1dcd0c7775a5ac0f7442634a330eb0173cc Jhen-Jie Hong iainst0409@gmail.com 2023-10-05T08:56:21-05:00 GitHub noreply@github.com 2023-10-05T16:56:21+03:00 ci : add swift build via xcodebuild (#3482)
019ba1dcd0c7775a5ac0f7442634a330eb0173cc beabc8cfb0145b48aad68fefc573d316fe9c3a8a Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-10-04T08:20:28-06:00 GitHub noreply@github.com 2023-10-04T17:20:28+03:00 convert : fix Baichuan2 models by using vocab size in config.json (#3299)
beabc8cfb0145b48aad68fefc573d316fe9c3a8a 0d152b37fecd5a4838330d47bb034cebf1681779 Georgi Gerganov ggerganov@gmail.com 2023-10-04T16:50:44+03:00 GitHub noreply@github.com 2023-10-04T16:50:44+03:00 readme : add project status link
0d152b37fecd5a4838330d47bb034cebf1681779 f8c90cdbaa729e64493164c1aba7ea80da7b716f Georgi Gerganov ggerganov@gmail.com 2023-10-04T16:25:41+03:00 Georgi Gerganov ggerganov@gmail.com 2023-10-04T16:25:41+03:00 ggml : fix build after #3329
f8c90cdbaa729e64493164c1aba7ea80da7b716f f93af02488179b9c52d0d391b08ae4c4d891b8d3 ds5t5 145942675+ds5t5@users.noreply.github.com 2023-10-04T06:23:39-07:00 GitHub noreply@github.com 2023-10-04T16:23:39+03:00 llm : add Refact model (#3329)
f93af02488179b9c52d0d391b08ae4c4d891b8d3 f72f8f22c9cb60465b2e79df2767e4ba9604e576 Georgi Gerganov ggerganov@gmail.com 2023-10-04T15:29:58+03:00 GitHub noreply@github.com 2023-10-04T15:29:58+03:00 sync : ggml (conv 1d + 2d updates, UB fixes) (#3468)
f72f8f22c9cb60465b2e79df2767e4ba9604e576 79f34abddb72ac5ddbf118f3d87520b611a10a7d Merrick Christensen merrick.christensen@gmail.com 2023-10-04T00:33:13-06:00 GitHub noreply@github.com 2023-10-04T09:33:13+03:00 finetune : readme fix typo (#3465)
79f34abddb72ac5ddbf118f3d87520b611a10a7d 8186242b6d67cf87ae179fb1a62f52fdf0e5c5eb Tameem 113388789+AhmadTameem@users.noreply.github.com 2023-10-03T23:38:19+05:00 GitHub noreply@github.com 2023-10-03T21:38:19+03:00 ggml : add RISC-V Vector Support for K-Quants and improved the existing intrinsics (#3453)
8186242b6d67cf87ae179fb1a62f52fdf0e5c5eb ac2219fef34eb5b713c286c34c6e4162c39c8f3b h-h-h-h 13482553+h-h-h-h@users.noreply.github.com 2023-10-03T20:16:15+02:00 GitHub noreply@github.com 2023-10-03T21:16:15+03:00 main : consistent prefix/suffix coloring (#3425)
ac2219fef34eb5b713c286c34c6e4162c39c8f3b 48be797ffbd80b062f55778e09e97180eb25d2ab Georgi Gerganov ggerganov@gmail.com 2023-10-03T21:04:01+03:00 GitHub noreply@github.com 2023-10-03T21:04:01+03:00 llama : fix session saving/loading (#3400)
48be797ffbd80b062f55778e09e97180eb25d2ab f56e1baec361b5381e32ee6b6e56e4f00e002dfe Alex Klinkhamer git@grencez.dev 2023-10-03T10:09:28-07:00 GitHub noreply@github.com 2023-10-03T20:09:28+03:00 llama : expose model's rope_freq_scale in the API (#3418)
f56e1baec361b5381e32ee6b6e56e4f00e002dfe 017efe899d8fa76118aef88e963210d48da01172 Jiahao Li liplus17@163.com 2023-10-04T00:55:21+08:00 GitHub noreply@github.com 2023-10-03T19:55:21+03:00 metal : alibi for arbitrary number of heads (#3426)
017efe899d8fa76118aef88e963210d48da01172 ff5a3f0c09dfa0a8e0bf76d1748df5c6dee0e8ff Eve 139727413+netrunnereve@users.noreply.github.com 2023-10-03T16:53:15Z GitHub noreply@github.com 2023-10-03T19:53:15+03:00 cmake : make LLAMA_NATIVE flag actually use the instructions supported by the processor (#3273)
ff5a3f0c09dfa0a8e0bf76d1748df5c6dee0e8ff 1c84003c08027f5d3a4cb876f51d6b6224a34d0e goerch jhr.walter@t-online.de 2023-10-03T09:16:26+02:00 GitHub noreply@github.com 2023-10-03T09:16:26+02:00 Work on the BPE tokenizer (#3252)
1c84003c08027f5d3a4cb876f51d6b6224a34d0e e78f0b0d0572168f328dd0e2ed3175a53fe52acc cebtenzzre cebtenzzre@gmail.com 2023-10-02T18:07:24-04:00 GitHub noreply@github.com 2023-10-02T18:07:24-04:00 convert : fix vocab size when not defined in hparams (#3421)
e78f0b0d0572168f328dd0e2ed3175a53fe52acc 665018c749101e81c816675198e731e47d6b1dbe cebtenzzre cebtenzzre@gmail.com 2023-10-02T15:38:43-04:00 GitHub noreply@github.com 2023-10-02T22:38:43+03:00 cmake : increase minimum version for add_link_options (#3444)
665018c749101e81c816675198e731e47d6b1dbe 29a404a951fb0b3f9c3b6ab8c4c9c76ac50d2bb3 shibe2 shibe@tuta.io 2023-10-02T23:26:15+04:00 GitHub noreply@github.com 2023-10-02T21:26:15+02:00 CLBlast: Add broadcast support for matrix multiplication (#3402)
29a404a951fb0b3f9c3b6ab8c4c9c76ac50d2bb3 0fe321031a5c670ab5fb5f49d69c4c91d783c93f cebtenzzre cebtenzzre@gmail.com 2023-10-02T15:20:28-04:00 GitHub noreply@github.com 2023-10-02T15:20:28-04:00 gguf : add BERT, MPT, and GPT-J arch info (#3408)
0fe321031a5c670ab5fb5f49d69c4c91d783c93f 9476b012260a2fb6c67976582d64484ce7406ed9 cebtenzzre cebtenzzre@gmail.com 2023-10-02T14:58:46-04:00 GitHub noreply@github.com 2023-10-02T14:58:46-04:00 gguf : general usability improvements (#3409)
9476b012260a2fb6c67976582d64484ce7406ed9 a03ce38455544121c5c00cf845def1443acd6ac8 cebtenzzre cebtenzzre@gmail.com 2023-10-02T09:16:50-04:00 GitHub noreply@github.com 2023-10-02T16:16:50+03:00 cmake : make CUDA flags more similar to the Makefile (#3420)
a03ce38455544121c5c00cf845def1443acd6ac8 a84767698495d72e44044f1f6db1c1cc721bfd15 xaedes xaedes@gmail.com 2023-10-02T15:15:45+02:00 GitHub noreply@github.com 2023-10-02T16:15:45+03:00 finetune : fix #3404 (#3437)
a84767698495d72e44044f1f6db1c1cc721bfd15 095231dfd32679e32300f8ffaf1770b693ea64b0 Adrian smith.adriane@gmail.com 2023-10-02T03:49:59-07:00 GitHub noreply@github.com 2023-10-02T13:49:59+03:00 metal : set log callback before initializing (#3427)
095231dfd32679e32300f8ffaf1770b693ea64b0 ea55295a745c084f588be20710f5a1a12abb1109 bandoti 141645996+bandoti@users.noreply.github.com 2023-10-02T06:51:49-03:00 GitHub noreply@github.com 2023-10-02T12:51:49+03:00 cmake : fix transient definitions in find pkg (#3411)
ea55295a745c084f588be20710f5a1a12abb1109 c97f01c362ac102c6994edb80008f8608539553a Kevin Ji 1146876+kevinji@users.noreply.github.com 2023-10-02T04:53:53-04:00 GitHub noreply@github.com 2023-10-02T11:53:53+03:00 docker : ignore Git files (#3314)
c97f01c362ac102c6994edb80008f8608539553a f5ef5cfb18148131fcf45bdd2331f0db5ab7c3d0 vvhg1 94630311+vvhg1@users.noreply.github.com 2023-10-02T09:42:02+02:00 GitHub noreply@github.com 2023-10-02T10:42:02+03:00 infill : add new example + extend server API (#3296)
f5ef5cfb18148131fcf45bdd2331f0db5ab7c3d0 40e07a60f9ce06e79f3ccd4c903eba300fb31b5e slaren slarengh@gmail.com 2023-09-30T18:12:57+02:00 GitHub noreply@github.com 2023-09-30T18:12:57+02:00 ggml-cuda : perform cublas mat mul of quantized types as f16 (#3412)
40e07a60f9ce06e79f3ccd4c903eba300fb31b5e bc34dd4f5b5a7c10ae3ed85a265ce6f2ed2fab79 slaren slarengh@gmail.com 2023-09-29T18:42:32+02:00 GitHub noreply@github.com 2023-09-29T18:42:32+02:00 llama.cpp : add documentation about rope_freq_base and scale values (#3401)
bc34dd4f5b5a7c10ae3ed85a265ce6f2ed2fab79 2777a84be429401a2b7d33c2b6a4ada1f0776f1b Georgi Gerganov ggerganov@gmail.com 2023-09-29T19:05:18+03:00 GitHub noreply@github.com 2023-09-29T19:05:18+03:00 train : fix KQ_pos allocation (#3392)
2777a84be429401a2b7d33c2b6a4ada1f0776f1b 0a4a4a098261ddd26480371eaccfe90d1bf6488a Cebtenzzre cebtenzzre@gmail.com 2023-09-29T09:48:45-04:00 GitHub noreply@github.com 2023-09-29T16:48:45+03:00 llama : quantize up to 31% faster on Linux and Windows with mmap (#3206)
0a4a4a098261ddd26480371eaccfe90d1bf6488a 569550df20c1ede59ff195a6b6e900957ad84d16 BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2023-09-29T08:50:35-04:00 GitHub noreply@github.com 2023-09-29T15:50:35+03:00 readme : update hot topics + model links (#3399)
569550df20c1ede59ff195a6b6e900957ad84d16 c71bf2c45c5140203184f50b259828107658e900 Andrew Duffy a10y@users.noreply.github.com 2023-09-29T07:15:57-04:00 GitHub noreply@github.com 2023-09-29T14:15:57+03:00 readme : add link to grammars app (#3388)
c71bf2c45c5140203184f50b259828107658e900 bc39553c901a91cfcb757863586250838c83eeab Jhen-Jie Hong iainst0409@gmail.com 2023-09-29T13:25:13+08:00 GitHub noreply@github.com 2023-09-29T08:25:13+03:00 swift : fix build on xcode 15 (#3387)
bc39553c901a91cfcb757863586250838c83eeab 0ccfc62a96a6b59a8faa14d1b350493f4cd51ae2 Cebtenzzre cebtenzzre@gmail.com 2023-09-28T17:41:44-04:00 GitHub noreply@github.com 2023-09-28T17:41:44-04:00 build : enable more non-default compiler warnings (#3200)
0ccfc62a96a6b59a8faa14d1b350493f4cd51ae2 7f1a0fe709ea1a861da2f3759f58a28bf8953c12 Hua Jiang allenhjiang@outlook.com 2023-09-28T13:06:18-07:00 GitHub noreply@github.com 2023-09-28T23:06:18+03:00 ggml_tensor: update the structure comments. (#3283)
7f1a0fe709ea1a861da2f3759f58a28bf8953c12 16bc66d9479edd5ee12ec734973554d4493c5dfa Qu Zongfu 43257352+yancaoweidaode@users.noreply.github.com 2023-09-29T03:51:52+08:00 GitHub noreply@github.com 2023-09-28T22:51:52+03:00 ggml : release the requested thread pool resource (#3292)
16bc66d9479edd5ee12ec734973554d4493c5dfa 0512d66670de3f650c579519833c085014b0f200 slaren slarengh@gmail.com 2023-09-28T21:42:38+02:00 GitHub noreply@github.com 2023-09-28T22:42:38+03:00 llama.cpp : split llama_context_params into model and context params (#3301)
0512d66670de3f650c579519833c085014b0f200 0e76a8992c8200237bbc6471a53fb8796b3872f7 Eve 139727413+netrunnereve@users.noreply.github.com 2023-09-28T19:31:04Z GitHub noreply@github.com 2023-09-28T22:31:04+03:00 ci : multithreaded builds (#3311)
0e76a8992c8200237bbc6471a53fb8796b3872f7 2db94d98eda56982d80238840b0652b4137a2a84 xaedes xaedes@gmail.com 2023-09-28T20:40:11+02:00 GitHub noreply@github.com 2023-09-28T21:40:11+03:00 train : finetune LORA (#2632)
2db94d98eda56982d80238840b0652b4137a2a84 ecf90b1a5114034bc0939b3968f549fe4d63cf6d Cebtenzzre cebtenzzre@gmail.com 2023-09-28T14:30:31-04:00 GitHub noreply@github.com 2023-09-28T14:30:31-04:00 gguf : basic type checking in gguf_get_* (#3346)
ecf90b1a5114034bc0939b3968f549fe4d63cf6d 2619109ad57d7a75388a9cce51e5da645410d92e Cebtenzzre cebtenzzre@gmail.com 2023-09-28T14:30:15-04:00 GitHub noreply@github.com 2023-09-28T14:30:15-04:00 gguf : make token scores and types optional (#3347)
2619109ad57d7a75388a9cce51e5da645410d92e ec893798b7a2a803466cc8f063051499ec3d96f7 Georgi Gerganov ggerganov@gmail.com 2023-09-28T19:36:36+03:00 GitHub noreply@github.com 2023-09-28T19:36:36+03:00 ci : disable freeBSD builds due to lack of VMs (#3381)
ec893798b7a2a803466cc8f063051499ec3d96f7 45855b3f1c7bdd0320aa632334d0b3e8965c26c4 Georgi Gerganov ggerganov@gmail.com 2023-09-28T19:04:36+03:00 GitHub noreply@github.com 2023-09-28T19:04:36+03:00 llama : custom attention mask + parallel decoding + no context swaps (#3228)
45855b3f1c7bdd0320aa632334d0b3e8965c26c4 4aea3b846ec151cc6d08f93a8889eae13b286b06 Kevin Ji 1146876+kevinji@users.noreply.github.com 2023-09-28T09:11:32-04:00 GitHub noreply@github.com 2023-09-28T09:11:32-04:00 docs : mark code as Bash (#3375)
4aea3b846ec151cc6d08f93a8889eae13b286b06 da0400344be12074e67dcabc565140289cf7efaa Pierre Alexandre SCHEMBRI pa.schembri@gmail.com 2023-09-28T14:13:37+02:00 GitHub noreply@github.com 2023-09-28T15:13:37+03:00 readme : add Mistral AI release 0.1 (#3362)
da0400344be12074e67dcabc565140289cf7efaa e519621010cac02c6fec0f8f3b16cda0591042c0 slaren slarengh@gmail.com 2023-09-28T12:08:28+02:00 GitHub noreply@github.com 2023-09-28T13:08:28+03:00 ggml-cuda : perform cublas fp16 matrix multiplication as fp16 (#3370)
e519621010cac02c6fec0f8f3b16cda0591042c0 ac43576124a75c2de6e333ac31a3444ff9eb9458 Zhang Peiyuan a1286225768@gmail.com 2023-09-28T02:45:20+08:00 GitHub noreply@github.com 2023-09-27T20:45:20+02:00 convert : remove bug in convert.py permute function (#3364)
ac43576124a75c2de6e333ac31a3444ff9eb9458 20c7e1e804690f3db58bd33eb56f8c6aa4735c63 Richard Roberson richardr1126@gmail.com 2023-09-27T10:25:12-06:00 GitHub noreply@github.com 2023-09-27T19:25:12+03:00 make-ggml.py : compatibility with more models and GGUF (#3290)
20c7e1e804690f3db58bd33eb56f8c6aa4735c63 dc6897404e141c74cbbf8030ecfebd74e1815411 Cebtenzzre cebtenzzre@gmail.com 2023-09-27T12:18:07-04:00 GitHub noreply@github.com 2023-09-27T12:18:07-04:00 gguf : fix a few general keys (#3341)
dc6897404e141c74cbbf8030ecfebd74e1815411 527e57cfd8a9a26bf622c0510c21c2508a24be26 Rickard Hallerbäck rickard.hallerback@gmail.com 2023-09-27T17:48:33+02:00 GitHub noreply@github.com 2023-09-27T18:48:33+03:00 metal : reusing llama.cpp logging (#3152)
527e57cfd8a9a26bf622c0510c21c2508a24be26 ffe88a36a913e5792aa383f0726bdbcf632e7191 Jag Chadha jagtesh@gmail.com 2023-09-27T11:34:32-04:00 GitHub noreply@github.com 2023-09-27T18:34:32+03:00 build : add ACCELERATE_NEW_LAPACK to fix warning on macOS Sonoma (#3342)
ffe88a36a913e5792aa383f0726bdbcf632e7191 99115f3fa654b593099c6719ad30e3f54ce231e1 BarfingLemurs 128182951+BarfingLemurs@users.noreply.github.com 2023-09-27T11:30:36-04:00 GitHub noreply@github.com 2023-09-27T18:30:36+03:00 readme : add some recent perplexity and bpw measurements to READMES, link for k-quants (#3340)
99115f3fa654b593099c6719ad30e3f54ce231e1 1726f9626f21f102d8e01df06c23a7f94add7990 DAN™ dranger003@gmail.com 2023-09-25T18:45:33-04:00 GitHub noreply@github.com 2023-09-25T18:45:33-04:00 cmake : fix build-info.h on MSVC (#3309)
1726f9626f21f102d8e01df06c23a7f94add7990 a98b1633d5a94d0aa84c7c16e1f8df5ac21fc850 2f38b454 dxf@protonmail.com 2023-09-26T02:24:52+08:00 GitHub noreply@github.com 2023-09-25T20:24:52+02:00 docs: Fix typo CLBlast_DIR var. (#3330)
a98b1633d5a94d0aa84c7c16e1f8df5ac21fc850 c091cdfb24621710c617ea85c92fcd347d0bf340 Erik Scholz Green-Sky@users.noreply.github.com 2023-09-25T13:48:30+02:00 GitHub noreply@github.com 2023-09-25T13:48:30+02:00 nix : add cuda, use a symlinked toolkit for cmake (#3202)
c091cdfb24621710c617ea85c92fcd347d0bf340 51a7cf5c6e490b2f51c82daa76c4ca4f8d845826 slaren slarengh@gmail.com 2023-09-23T21:48:24+02:00 GitHub noreply@github.com 2023-09-23T21:48:24+02:00 llama-bench : add README (#3317)
51a7cf5c6e490b2f51c82daa76c4ca4f8d845826 bedb92b603886768ad51e629f81eda15ff6b86f5 Cebtenzzre cebtenzzre@gmail.com 2023-09-23T05:28:50-04:00 GitHub noreply@github.com 2023-09-23T12:28:50+03:00 examples : fix RoPE defaults to match PR #3240 (#3315)
bedb92b603886768ad51e629f81eda15ff6b86f5 bc9d3e3971e5607a10ff4c24e39568ce1ac87271 Kevin Ji 1146876+kevinji@users.noreply.github.com 2023-09-22T23:52:23-04:00 GitHub noreply@github.com 2023-09-22T23:52:23-04:00 scripts : use `/usr/bin/env` in shebang (#3313)
bc9d3e3971e5607a10ff4c24e39568ce1ac87271 36b904e20003017f50108ae68359ef87a192dae2 Lee Drake b.lee.drake@gmail.com 2023-09-21T13:00:24-06:00 GitHub noreply@github.com 2023-09-21T21:00:24+02:00 Update README.md (#3289)
36b904e20003017f50108ae68359ef87a192dae2 324f3403d54ae4499a1d68623161015f7419fb76 shibe2 shibe@tuta.io 2023-09-21T22:10:26+04:00 GitHub noreply@github.com 2023-09-21T14:10:26-04:00 ggml-opencl.cpp: Make private functions static (#3300)
324f3403d54ae4499a1d68623161015f7419fb76 f56c418ab0a635c020bcb5bf44b8f00cb3c9e514 Edward Taylor edeetee@gmail.com 2023-09-21T21:08:20+12:00 GitHub noreply@github.com 2023-09-21T12:08:20+03:00 zig : fix for updated c lib (#3259)
f56c418ab0a635c020bcb5bf44b8f00cb3c9e514 8185710a80531e9ee0c0cb99d3a9c9af1019ab67 yuiseki yuiseki@gmail.com 2023-09-21T17:57:40+09:00 GitHub noreply@github.com 2023-09-21T11:57:40+03:00 embedding : update README.md (#3224)
8185710a80531e9ee0c0cb99d3a9c9af1019ab67 7eb41179edc56083ef4eb2df7967ac9ff38b34fb Johannes Gäßler johannesg@5d6.de 2023-09-21T10:43:53+02:00 GitHub noreply@github.com 2023-09-21T11:43:53+03:00 CUDA: use only 1 thread if fully offloaded (#2915)
7eb41179edc56083ef4eb2df7967ac9ff38b34fb a5661d7e71d15b8dfc81bc0510ba912ebe85dfa3 Georgi Gerganov ggerganov@gmail.com 2023-09-20T20:48:22+03:00 GitHub noreply@github.com 2023-09-20T20:48:22+03:00 readme : update hot topics
a5661d7e71d15b8dfc81bc0510ba912ebe85dfa3 65c2c1c5ab7c5089dbc6d10bc49b9c58f0164317 Cebtenzzre cebtenzzre@gmail.com 2023-09-20T12:12:47-04:00 GitHub noreply@github.com 2023-09-20T12:12:47-04:00 llama : allow gguf RoPE keys to be overridden with defaults (#3240)
65c2c1c5ab7c5089dbc6d10bc49b9c58f0164317 80834daecf4b9021770361a6d5e1b9c7a60e6854 Cebtenzzre cebtenzzre@gmail.com 2023-09-20T12:06:08-04:00 GitHub noreply@github.com 2023-09-20T12:06:08-04:00 benchmark-matmult : do not use integer abs() on a float (#3277)
80834daecf4b9021770361a6d5e1b9c7a60e6854 a40f2b656fab364ce0aff98dbefe9bd9c3721cc9 kang tpdns9032100@gmail.com 2023-09-20T22:48:22+09:00 GitHub noreply@github.com 2023-09-20T15:48:22+02:00 flake : Restore default package's buildInputs (#3262)
a40f2b656fab364ce0aff98dbefe9bd9c3721cc9 d119c04c159d015a93567df7e73e0e45a22d0f1d Alon alonfaraj@gmail.com 2023-09-20T15:06:36+03:00 GitHub noreply@github.com 2023-09-20T14:06:36+02:00 CI: FreeBSD fix (#3258)
d119c04c159d015a93567df7e73e0e45a22d0f1d 8781013ef654270cbead3e0011e33a6d690fb168 Georgi Gerganov ggerganov@gmail.com 2023-09-20T10:02:39+03:00 GitHub noreply@github.com 2023-09-20T10:02:39+03:00 examples : fix benchmark-matmult (#1554)
8781013ef654270cbead3e0011e33a6d690fb168 7ddf185537b712ea0ccbc5f222ee92bed654914e Cebtenzzre cebtenzzre@gmail.com 2023-09-18T10:03:53-04:00 GitHub noreply@github.com 2023-09-18T10:03:53-04:00 make : restore build-info.h dependency for several targets (#3205)
7ddf185537b712ea0ccbc5f222ee92bed654914e ee66942d7ef7c259528158f9a3bd1c314984d32f Erik Scholz Green-Sky@users.noreply.github.com 2023-09-18T02:21:47+02:00 GitHub noreply@github.com 2023-09-18T02:21:47+02:00 ci : switch cudatoolkit install on windows to networked (#3236)
ee66942d7ef7c259528158f9a3bd1c314984d32f 111163e2463171891680feed94371eb9becd9817 Johannes Gäßler johannesg@5d6.de 2023-09-17T23:35:20+02:00 GitHub noreply@github.com 2023-09-17T23:35:20+02:00 CUDA: fix peer access logic (#3231)
111163e2463171891680feed94371eb9becd9817 8b428c9bc84be6887d904600d1298b28baffd552 Johannes Gäßler johannesg@5d6.de 2023-09-17T16:37:53+02:00 GitHub noreply@github.com 2023-09-17T16:37:53+02:00 CUDA: enable peer access between devices (#2470)
8b428c9bc84be6887d904600d1298b28baffd552 578d8c8f5cb72f354bc115ba230ee5b2d803eee7 slaren slarengh@gmail.com 2023-09-17T14:33:28+02:00 GitHub noreply@github.com 2023-09-17T14:33:28+02:00 llama.cpp : show model size and BPW on load (#3223)
578d8c8f5cb72f354bc115ba230ee5b2d803eee7 b541b4f0b1d4d9871c831e47cd5ff661039d6101 Johannes Gäßler johannesg@5d6.de 2023-09-17T14:16:22+02:00 GitHub noreply@github.com 2023-09-17T14:16:22+02:00 CUDA: fix scratch malloced on non-main device (#3220)
b541b4f0b1d4d9871c831e47cd5ff661039d6101 5dbc2b3213126a31d3be4ade8ca042cb93019682 IsaacDynamo 61521674+IsaacDynamo@users.noreply.github.com 2023-09-16T19:35:25+02:00 GitHub noreply@github.com 2023-09-16T19:35:25+02:00 Enable BUILD_SHARED_LIBS=ON on all Windows builds (#3215)
5dbc2b3213126a31d3be4ade8ca042cb93019682 b08e75baea294e366628b898e85c0bd359b58115 Vlad spitfireage@gmail.com 2023-09-16T17:55:43+03:00 GitHub noreply@github.com 2023-09-16T16:55:43+02:00 Enable build with CUDA 11.0 (make) (#3132)
b08e75baea294e366628b898e85c0bd359b58115 e6616cf0db2b63189fc34d0076f654af9adecdf8 goerch jhr.walter@t-online.de 2023-09-16T13:41:33+02:00 GitHub noreply@github.com 2023-09-16T13:41:33+02:00 Fixing the last deviations from sentencepiece indicated by test-tokenizer-1 (#3170)
e6616cf0db2b63189fc34d0076f654af9adecdf8 3aefaab9e59335ebb07d5205dbc8633efd680e58 Cebtenzzre cebtenzzre@gmail.com 2023-09-15T16:59:49-04:00 GitHub noreply@github.com 2023-09-15T16:59:49-04:00 examples : add compiler version and target to build info (#2998)
3aefaab9e59335ebb07d5205dbc8633efd680e58 69eb67e28275cd2d57693405f768754a7b2245ad Cebtenzzre cebtenzzre@gmail.com 2023-09-15T15:38:27-04:00 GitHub noreply@github.com 2023-09-15T15:38:27-04:00 check C++ code with -Wmissing-declarations (#3184)
69eb67e28275cd2d57693405f768754a7b2245ad 4fe09dfe665c58a753dc9eb638dd4dca1cd35488 Cebtenzzre cebtenzzre@gmail.com 2023-09-15T15:18:15-04:00 GitHub noreply@github.com 2023-09-15T15:18:15-04:00 fix build numbers by setting fetch-depth=0 (#3197)
4fe09dfe665c58a753dc9eb638dd4dca1cd35488 80291a1d02a07f7f66666fb576c5b1e75aa48b46 Meng Zhang meng@tabbyml.com 2023-09-16T03:02:13+08:00 GitHub noreply@github.com 2023-09-15T22:02:13+03:00 llama : add support for StarCoder model architectures (#3187)
80291a1d02a07f7f66666fb576c5b1e75aa48b46 c6f1491da032238241e01021c8c58d7b540a043f Cebtenzzre cebtenzzre@gmail.com 2023-09-15T14:02:01-04:00 GitHub noreply@github.com 2023-09-15T21:02:01+03:00 common : do not use GNU zero-length __VA_ARGS__ extension (#3195)
c6f1491da032238241e01021c8c58d7b540a043f e3d87a6c36eadd84d58763143c6d56a0c771ca40 Georgi Gerganov ggerganov@gmail.com 2023-09-15T20:17:24+03:00 GitHub noreply@github.com 2023-09-15T20:17:24+03:00 metal : fix bug in soft_max kernels (out-of-bounds access) (#3194)
e3d87a6c36eadd84d58763143c6d56a0c771ca40 8c00b7a6ff38e27fa1e471452b8a480913772c2a Cebtenzzre cebtenzzre@gmail.com 2023-09-15T12:29:02-04:00 GitHub noreply@github.com 2023-09-15T12:29:02-04:00 convert : make ftype optional in simple scripts (#3185)
8c00b7a6ff38e27fa1e471452b8a480913772c2a 7e50d34be68aae2cc766203703dd188e910e033a Georgi Gerganov ggerganov@gmail.com 2023-09-15T19:06:03+03:00 GitHub noreply@github.com 2023-09-15T19:06:03+03:00 sync : ggml (Metal F32 support + reduce ggml-alloc size) (#3192)
7e50d34be68aae2cc766203703dd188e910e033a 235f7c193b02dacfb56319e41a28684b3a2c6db0 Engininja2 139037756+Engininja2@users.noreply.github.com 2023-09-15T06:24:30-06:00 GitHub noreply@github.com 2023-09-15T15:24:30+03:00 cmake : fix building shared libs for clang (rocm) on windows (#3176)
235f7c193b02dacfb56319e41a28684b3a2c6db0 a51b68765799e75e17c7622f376bfbeb66f1bd70 Evgeny Kurnevsky kurnevsky@gmail.com 2023-09-15T10:10:22+02:00 GitHub noreply@github.com 2023-09-15T11:10:22+03:00 flake : use pkg-config instead of pkgconfig (#3188)
a51b68765799e75e17c7622f376bfbeb66f1bd70 76164fe2e65c058e9ee2c3afd0ad6b182ca57e25 Georgi Gerganov ggerganov@gmail.com 2023-09-15T11:09:24+03:00 GitHub noreply@github.com 2023-09-15T11:09:24+03:00 metal : relax conditions on fast matrix multiplication kernel (#3168)
76164fe2e65c058e9ee2c3afd0ad6b182ca57e25 c2ab6fe661af9834ca6dd75f14e2439938cc22ac Andrei abetlen@gmail.com 2023-09-15T04:07:40-04:00 GitHub noreply@github.com 2023-09-15T11:07:40+03:00 cmake : fix llama.h location when built outside of root directory (#3179)
c2ab6fe661af9834ca6dd75f14e2439938cc22ac 2d770505a89a99ce78a5950cf14fc06d3176ffa4 Ali Tariq ali.tariq@10xengineers.ai 2023-09-15T13:06:56+05:00 GitHub noreply@github.com 2023-09-15T11:06:56+03:00 ci : Cloud-V for RISC-V builds (#3160)
2d770505a89a99ce78a5950cf14fc06d3176ffa4 98311c427739e3b06527c3ce6b5c021ab6692740 Roland 14355895+rbur0425@users.noreply.github.com 2023-09-15T03:28:45-04:00 GitHub noreply@github.com 2023-09-15T10:28:45+03:00 llama : remove mtest (#3177)
98311c427739e3b06527c3ce6b5c021ab6692740 feea179e9f9921e96e8fb1b8855d4a8f83682455 Cebtenzzre cebtenzzre@gmail.com 2023-09-14T21:09:53-04:00 GitHub noreply@github.com 2023-09-14T21:09:53-04:00 llama : make quantize example up to 2.7x faster (#3115)
feea179e9f9921e96e8fb1b8855d4a8f83682455 769266a543f68377a1d904ec2a8c27b38a4025ab jneem joeneeman@gmail.com 2023-09-14T13:54:47-05:00 GitHub noreply@github.com 2023-09-14T21:54:47+03:00 flake : allow $out/include to already exist (#3175)
769266a543f68377a1d904ec2a8c27b38a4025ab cf8238e7f43cb82a36426af392037e85cd2a3df6 Andrei abetlen@gmail.com 2023-09-14T13:38:16-04:00 GitHub noreply@github.com 2023-09-14T20:38:16+03:00 cmake : compile ggml-rocm with -fpic when building shared library (#3158)
cf8238e7f43cb82a36426af392037e85cd2a3df6 4b8560e72a936b5d536ebd1e7a5dd579984769f3 Asbjørn Olling asbjornolling@gmail.com 2023-09-14T19:25:00+02:00 GitHub noreply@github.com 2023-09-14T20:25:00+03:00 flake : include llama.h in nix output (#3159)
4b8560e72a936b5d536ebd1e7a5dd579984769f3 83a53b753a9499a2a3535c93975b430cb2c828a9 Cebtenzzre cebtenzzre@gmail.com 2023-09-14T13:22:47-04:00 GitHub noreply@github.com 2023-09-14T20:22:47+03:00 make : fix clang++ detection, move some definitions to CPPFLAGS (#3155)
83a53b753a9499a2a3535c93975b430cb2c828a9 5c872dbca2c7979b1f6dafc97db0774b8bbf9372 Alon alonfaraj@gmail.com 2023-09-14T20:21:25+03:00 GitHub noreply@github.com 2023-09-14T19:21:25+02:00 CI: add FreeBSD & simplify CUDA windows (#3053)
5c872dbca2c7979b1f6dafc97db0774b8bbf9372 990a5e226a1a0ac858abe3aa7e5f3b000d4fa665 akawrykow 142945436+akawrykow@users.noreply.github.com 2023-09-14T10:19:42-07:00 GitHub noreply@github.com 2023-09-14T20:19:42+03:00 falcon : use stated vocab size (#2914)
990a5e226a1a0ac858abe3aa7e5f3b000d4fa665 980ab41afba96106cd29cdf3aa6f948c251cb71f bandoti 141645996+bandoti@users.noreply.github.com 2023-09-14T14:04:40-03:00 GitHub noreply@github.com 2023-09-14T20:04:40+03:00 cmake : add relocatable Llama package (#2960)
980ab41afba96106cd29cdf3aa6f948c251cb71f e394084166baac09e8ee9a08a4686f907f7e5291 dylan canardleteer@users.noreply.github.com 2023-09-14T09:47:00-07:00 GitHub noreply@github.com 2023-09-14T19:47:00+03:00 docker : add gpu image CI builds (#3103)
e394084166baac09e8ee9a08a4686f907f7e5291 4c8643dd6ea1a163bc5979cb69c1e7ab0975bc93 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-09-14T10:32:26-06:00 GitHub noreply@github.com 2023-09-14T19:32:26+03:00 gguf-py : support identity operation in TensorNameMap (#3095)
4c8643dd6ea1a163bc5979cb69c1e7ab0975bc93 35f73049af6c676a106a5a990a819ae0bc3fcd7d jameswu2014 545426914@qq.com 2023-09-15T00:32:10+08:00 GitHub noreply@github.com 2023-09-14T12:32:10-04:00 feature : support Baichuan serial models (#3009)
35f73049af6c676a106a5a990a819ae0bc3fcd7d 71ca2fad7d6c0ef95ef9944fb3a1a843e481f314 Leng Yue lengyue@lengyue.me 2023-09-14T09:14:44-07:00 GitHub noreply@github.com 2023-09-14T19:14:44+03:00 speculative : add heuristic algorithm (#3006)
71ca2fad7d6c0ef95ef9944fb3a1a843e481f314 1b6c650d16048d6427dd502a9627e72837265844 goerch jhr.walter@t-online.de 2023-09-13T15:19:44+02:00 GitHub noreply@github.com 2023-09-13T16:19:44+03:00 whisper : tokenizer fix + re-enable tokenizer test for LLaMa (#3096)
1b6c650d16048d6427dd502a9627e72837265844 0a5eebb45d5697127b84418576dc479c400c4b3d Tristan Ross rosscomputerguy@protonmail.com 2023-09-13T06:08:52-07:00 GitHub noreply@github.com 2023-09-13T16:08:52+03:00 cmake : add a compiler flag check for FP16 format (#3086)
0a5eebb45d5697127b84418576dc479c400c4b3d 84e723653ca99d51a74b454984acf2c077468561 Johannes Gäßler johannesg@5d6.de 2023-09-13T11:20:24+02:00 GitHub noreply@github.com 2023-09-13T11:20:24+02:00 CUDA: mul_mat_q RDNA2 tunings (#2910)
84e723653ca99d51a74b454984acf2c077468561 b52b29ab9d601bb298050bcd2261169bc917ba2c FK sozforex@gmail.com 2023-09-13T08:50:46+02:00 GitHub noreply@github.com 2023-09-13T08:50:46+02:00 speculative: add --n-gpu-layers-draft option (#3063)
b52b29ab9d601bb298050bcd2261169bc917ba2c 4f7cd6ba9c88d3ca9a207b6e04f8b2b1efd707b8 Eric Sommerlade es0m@users.noreply.github.com 2023-09-13T02:54:20+01:00 GitHub noreply@github.com 2023-09-12T21:54:20-04:00 arm64 support for windows (#3007)
4f7cd6ba9c88d3ca9a207b6e04f8b2b1efd707b8 89e89599fd095172f8d67903b5e227467420f036 Johannes Gäßler johannesg@5d6.de 2023-09-13T00:15:33+02:00 GitHub noreply@github.com 2023-09-13T00:15:33+02:00 CUDA: fix LoRAs (#3130)
89e89599fd095172f8d67903b5e227467420f036 d54a4027a6ebda98ab0fef7fa0c2247d0bef132a Johannes Gäßler johannesg@5d6.de 2023-09-11T22:58:41+02:00 GitHub noreply@github.com 2023-09-11T22:58:41+02:00 CUDA: fix mul_mat_q not used for output tensor (#3127)
d54a4027a6ebda98ab0fef7fa0c2247d0bef132a 1b0d09259e37898c519edb6c52d58f4d096f10bd Johannes Gäßler johannesg@5d6.de 2023-09-11T19:55:51+02:00 GitHub noreply@github.com 2023-09-11T19:55:51+02:00 CUDA: lower GPU latency + fix Windows performance (#3110)
1b0d09259e37898c519edb6c52d58f4d096f10bd 8a4ca9af569853023ce87f047eb5165df13f2ff1 Jhen-Jie Hong iainst0409@gmail.com 2023-09-11T19:49:06+08:00 GitHub noreply@github.com 2023-09-11T19:49:06+08:00 cmake : support build for iOS/tvOS (#3116)
8a4ca9af569853023ce87f047eb5165df13f2ff1 f31b6f4e2d6def3c0bd7c75f75c0c1e8698e0589 Johannes Gäßler johannesg@5d6.de 2023-09-11T13:00:24+02:00 GitHub noreply@github.com 2023-09-11T13:00:24+02:00 CUDA: add device number to error messages (#3112)
f31b6f4e2d6def3c0bd7c75f75c0c1e8698e0589 6eeb4d90839bac1e6085e5544654ab5c319ad09a Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-09-11T09:30:11+02:00 GitHub noreply@github.com 2023-09-11T10:30:11+03:00 metal : PP speedup (#3084)
6eeb4d90839bac1e6085e5544654ab5c319ad09a 21ac3a1503001020122db5dce6adf34b761675f5 Erik Scholz Green-Sky@users.noreply.github.com 2023-09-10T17:06:53+02:00 GitHub noreply@github.com 2023-09-10T11:06:53-04:00 convert: remove most of the n_mult usage in convert.py (#3098)
21ac3a1503001020122db5dce6adf34b761675f5 4fd54779550e43e2a29f6840ebcf8f395a2f879e kchro3 62481661+kchro3@users.noreply.github.com 2023-09-09T02:12:10-07:00 GitHub noreply@github.com 2023-09-09T17:12:10+08:00 metal : support for Swift (#3078)
4fd54779550e43e2a29f6840ebcf8f395a2f879e ec2a24fedf1de8ebd5f170016953b09ff2806924 Jhen-Jie Hong iainst0409@gmail.com 2023-09-09T16:46:04+08:00 GitHub noreply@github.com 2023-09-09T11:46:04+03:00 metal : support build for iOS/tvOS (#3089)
ec2a24fedf1de8ebd5f170016953b09ff2806924 7d99aca759f2f8a1ff39f3bb02a840f69863428b takov751 40316768+takov751@users.noreply.github.com 2023-09-08T17:06:26+01:00 GitHub noreply@github.com 2023-09-08T19:06:26+03:00 flake : add train-text-from-scratch to flake.nix (#3042)
7d99aca759f2f8a1ff39f3bb02a840f69863428b ba7ffbb2517ff8cf4c689f94a9ad866f3ee71225 Ikko Eltociear Ashimine eltociear@gmail.com 2023-09-09T01:04:32+09:00 GitHub noreply@github.com 2023-09-08T19:04:32+03:00 readme : fix typo (#3043)
ba7ffbb2517ff8cf4c689f94a9ad866f3ee71225 e64f5b55783e910d8287363895d652b4bea6527a Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-09-08T18:01:04+02:00 GitHub noreply@github.com 2023-09-08T19:01:04+03:00 metal : Q3_K speedup (#2995)
e64f5b55783e910d8287363895d652b4bea6527a 94f10b91ed69980f299441e49c8dbdb448f0ccc6 Cebtenzzre cebtenzzre@gmail.com 2023-09-08T11:43:35-04:00 GitHub noreply@github.com 2023-09-08T11:43:35-04:00 examples : make n_ctx warning work again (#3066)
94f10b91ed69980f299441e49c8dbdb448f0ccc6 b3e9852e471d12cbbe5dad20c81c4766d969739a Georgi Gerganov ggerganov@gmail.com 2023-09-08T18:18:04+03:00 GitHub noreply@github.com 2023-09-08T18:18:04+03:00 readme : update hot tpoics
b3e9852e471d12cbbe5dad20c81c4766d969739a cb6c44c5e045709b6bb5cc9bb8c9be107c771a78 Georgi Gerganov ggerganov@gmail.com 2023-09-08T17:58:07+03:00 GitHub noreply@github.com 2023-09-08T17:58:07+03:00 sync : ggml (CUDA GLM RoPE + POSIX) (#3082)
cb6c44c5e045709b6bb5cc9bb8c9be107c771a78 a21baeb12202a9020b48c53beaaf4b355228e8ba Przemysław Pawełczyk przemoc@gmail.com 2023-09-08T14:09:21+02:00 GitHub noreply@github.com 2023-09-08T15:09:21+03:00 build : do not use _GNU_SOURCE gratuitously (#2035)
a21baeb12202a9020b48c53beaaf4b355228e8ba 6ff712a6d1a0c85d996e2f681df57a2554cfe5c1 hongbo.mo 352280764@qq.com 2023-09-08T18:57:55+08:00 GitHub noreply@github.com 2023-09-08T13:57:55+03:00 docker : add git to full-cuda.Dockerfile main-cuda.Dockerfile (#3044)
6ff712a6d1a0c85d996e2f681df57a2554cfe5c1 ebc96086af49fe70108cafcea6ab4bebd658a41a Yui dev@sleepyyui.com 2023-09-08T12:32:55+02:00 GitHub noreply@github.com 2023-09-08T12:32:55+02:00 Update deprecated GGML TheBloke links to GGUF (#3079)
ebc96086af49fe70108cafcea6ab4bebd658a41a 7f412dab9c8801f5d37904f7dce1faf4c2b43b42 slaren slarengh@gmail.com 2023-09-08T04:04:56+02:00 GitHub noreply@github.com 2023-09-08T04:04:56+02:00 ggml-alloc : correctly check mmap return value for errors (#3075)
7f412dab9c8801f5d37904f7dce1faf4c2b43b42 6336d834ec7bff3e93e24182c0f609d2f2bdce26 Kunshang Ji kunshang.ji@intel.com 2023-09-08T09:46:56+08:00 GitHub noreply@github.com 2023-09-08T03:46:56+02:00 enable CPU HBM (#2603)
6336d834ec7bff3e93e24182c0f609d2f2bdce26 00d62adb79bf914a95fb9a2e8f42f3029e76d62c Cebtenzzre cebtenzzre@gmail.com 2023-09-07T14:27:42-04:00 GitHub noreply@github.com 2023-09-07T14:27:42-04:00 convert : fix F32 ftype not being saved (#3048)
00d62adb79bf914a95fb9a2e8f42f3029e76d62c 4fa2cc1750b861880de42515cb19c13b2d776ee2 Cebtenzzre cebtenzzre@gmail.com 2023-09-07T13:22:29-04:00 GitHub noreply@github.com 2023-09-07T13:22:29-04:00 fix some warnings from gcc and clang-tidy (#3038)
4fa2cc1750b861880de42515cb19c13b2d776ee2 5ffab089a54bc06ae4a9ab533893b558756a1e80 Cebtenzzre cebtenzzre@gmail.com 2023-09-07T10:15:01-04:00 GitHub noreply@github.com 2023-09-07T10:15:01-04:00 make : improve test target (#3031)
5ffab089a54bc06ae4a9ab533893b558756a1e80 15b67a66c2f2d6032415b28a699b5131962318f1 Cebtenzzre cebtenzzre@gmail.com 2023-09-07T10:13:50-04:00 GitHub noreply@github.com 2023-09-07T10:13:50-04:00 make : fix CPPFLAGS (#3035)
15b67a66c2f2d6032415b28a699b5131962318f1 be8c9c245bd129ebabb80e0a7a8dd7daeb4d30af slaren slarengh@gmail.com 2023-09-07T15:52:34+02:00 GitHub noreply@github.com 2023-09-07T15:52:34+02:00 llama-bench : use two tokens in the warmup run for prompt evals (#3059)
be8c9c245bd129ebabb80e0a7a8dd7daeb4d30af be6beeb8d75294552c4918fce06d7b84eebf3d79 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-09-07T15:45:01+02:00 GitHub noreply@github.com 2023-09-07T16:45:01+03:00 metal : parallel RoPE on Metal (#3024)
be6beeb8d75294552c4918fce06d7b84eebf3d79 c4f496648c1e32efeb714200e7eae7fc7cfbb223 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-09-07T15:42:42+02:00 GitHub noreply@github.com 2023-09-07T16:42:42+03:00 metal : correct fix of kernel_norm (#3060)
c4f496648c1e32efeb714200e7eae7fc7cfbb223 fec2fb19e4229aac58c98171c46e77144b99f8a3 Georgi Gerganov ggerganov@gmail.com 2023-09-07T15:49:09+03:00 GitHub noreply@github.com 2023-09-07T15:49:09+03:00 metal : fix kernel_norm (fixes Falcon on Metal) (#3057)
fec2fb19e4229aac58c98171c46e77144b99f8a3 178b1850ebd21b349cebbee887950e435c5aa2d3 Przemysław Pawełczyk przemoc@gmail.com 2023-09-07T10:15:06+02:00 GitHub noreply@github.com 2023-09-07T11:15:06+03:00 ggml : posixify madvise and pagesize (#3037)
178b1850ebd21b349cebbee887950e435c5aa2d3 ea2c85d5d2a93d39d0172222917f3195f0e456ff Georgi Gerganov ggerganov@gmail.com 2023-09-06T12:40:57+03:00 Georgi Gerganov ggerganov@gmail.com 2023-09-06T12:40:57+03:00 k-quants : fix zero-weight guard in Q6_K (ref #3040)
ea2c85d5d2a93d39d0172222917f3195f0e456ff 9912b9efc8922321fe7202ab42ba913833cbe9cd Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-09-06T02:49:11-06:00 GitHub noreply@github.com 2023-09-06T02:49:11-06:00 convert-llama-ggml-to-gguf: Try to handle files older than GGJTv3 (#3023)
9912b9efc8922321fe7202ab42ba913833cbe9cd 9e2023156e5b5acabaf8632e66c6ae68d3703c31 Cebtenzzre cebtenzzre@gmail.com 2023-09-05T18:21:10-04:00 GitHub noreply@github.com 2023-09-05T18:21:10-04:00 build : add LLAMA_METAL_NDEBUG flag (#3033)
9e2023156e5b5acabaf8632e66c6ae68d3703c31 de2fe892af92a5c7b5ef1beb7efbc0524343fbab Cebtenzzre cebtenzzre@gmail.com 2023-09-05T15:12:00-04:00 GitHub noreply@github.com 2023-09-05T15:12:00-04:00 make : use new flag variables for recent changes (#3019)
de2fe892af92a5c7b5ef1beb7efbc0524343fbab c9c3220c485c7bea740a07cda7343677fb3beaae Cebtenzzre cebtenzzre@gmail.com 2023-09-05T15:10:27-04:00 GitHub noreply@github.com 2023-09-05T15:10:27-04:00 examples : replace fprintf to stdout with printf (#3017)
c9c3220c485c7bea740a07cda7343677fb3beaae d59bd97065cd7ded6c4ecab54b1d5e0b1b11e318 Erik Scholz Green-Sky@users.noreply.github.com 2023-09-05T19:41:00+02:00 GitHub noreply@github.com 2023-09-05T19:41:00+02:00 convert: fix convert.py not working with int filename_stem (#3028)
d59bd97065cd7ded6c4ecab54b1d5e0b1b11e318 35938ee3b0c16f1fbbf240dae21e0228864b938c Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-09-05T09:55:33+02:00 GitHub noreply@github.com 2023-09-05T09:55:33+02:00 Guard against all weights in a super-block being zero (#3010)
35938ee3b0c16f1fbbf240dae21e0228864b938c 921772104ba2219bfdc2b2980d05ebc0aa0c92a4 Georgi Gerganov ggerganov@gmail.com 2023-09-05T10:46:39+03:00 Georgi Gerganov ggerganov@gmail.com 2023-09-05T10:46:39+03:00 llama : update logic for number of threads when using BLAS
921772104ba2219bfdc2b2980d05ebc0aa0c92a4 2ba85c8609309a59d49c45ab43c31800b7ba141c Georgi Gerganov ggerganov@gmail.com 2023-09-05T08:46:17+03:00 GitHub noreply@github.com 2023-09-05T08:46:17+03:00 speculative : add grammar support (#2991)
2ba85c8609309a59d49c45ab43c31800b7ba141c e36ecdccc8754783f93ad3ac8a09e540101f2ca0 Georgi Gerganov ggerganov@gmail.com 2023-09-04T22:50:50+03:00 Georgi Gerganov ggerganov@gmail.com 2023-09-04T22:50:50+03:00 py : minor
e36ecdccc8754783f93ad3ac8a09e540101f2ca0 bd33e5ab92e7f214205792fc1cd9ca28e810f897 Georgi Gerganov ggerganov@gmail.com 2023-09-04T22:26:24+03:00 GitHub noreply@github.com 2023-09-04T22:26:24+03:00 build : on Mac OS enable Metal by default (#2901)
bd33e5ab92e7f214205792fc1cd9ca28e810f897 31035681445181fb414e0def7ec3f84462b3bd97 slaren slarengh@gmail.com 2023-09-04T14:59:52+02:00 GitHub noreply@github.com 2023-09-04T14:59:52+02:00 ggml-opencl : store GPU buffer in ggml_tensor::extra (#2994)
31035681445181fb414e0def7ec3f84462b3bd97 5b8530d88c489f9d0c0ef3d0886b369f655b792e Cebtenzzre cebtenzzre@gmail.com 2023-09-04T06:40:18-04:00 GitHub noreply@github.com 2023-09-04T13:40:18+03:00 llama-bench : make cpp file non-executable (#2999)
5b8530d88c489f9d0c0ef3d0886b369f655b792e e4386f417faf894f6706eec005e24d142b577fcb Leng Yue lengyue@lengyue.me 2023-09-04T03:39:57-07:00 GitHub noreply@github.com 2023-09-04T13:39:57+03:00 make : add speculative example (#3003)
e4386f417faf894f6706eec005e24d142b577fcb 35195689cd835464779c247b1c22ab9247418fd1 Aarni Koskela akx@iki.fi 2023-09-04T10:28:55+02:00 GitHub noreply@github.com 2023-09-04T16:28:55+08:00 server : add a subtle loading animation to the edit box (#2466)
35195689cd835464779c247b1c22ab9247418fd1 cf9b08485c4c2d4d945c6e74fe20f273a38b6104 Jiahao Li liplus17@163.com 2023-09-04T14:53:30+08:00 GitHub noreply@github.com 2023-09-04T08:53:30+02:00 2x faster (rms) norm cuda kernels (3.7% e2e improvement) (#2985)
cf9b08485c4c2d4d945c6e74fe20f273a38b6104 47068e517004d90f13c16352bb3b4cafd53a00cd slaren slarengh@gmail.com 2023-09-03T20:34:09+02:00 GitHub noreply@github.com 2023-09-03T20:34:09+02:00 ggml-alloc : use virtual memory for measurement (#2973)
47068e517004d90f13c16352bb3b4cafd53a00cd 8f429fa5111901f9646cf998643ac5310846d487 Georgi Gerganov ggerganov@gmail.com 2023-09-03T15:12:08+03:00 GitHub noreply@github.com 2023-09-03T15:12:08+03:00 speculative : PoC for speeding-up inference via speculative sampling (#2926)
8f429fa5111901f9646cf998643ac5310846d487 6519e9c99cffbad19b31bcba86df48c500628c09 Georgi Gerganov ggerganov@gmail.com 2023-09-03T13:42:56+03:00 Georgi Gerganov ggerganov@gmail.com 2023-09-03T13:43:17+03:00 perplexity : fix ETA by warming up the model with an empty run
6519e9c99cffbad19b31bcba86df48c500628c09 b7f2aa9e512c3be2e863d877cbb1056d7c4a03f8 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-09-03T04:38:43-06:00 GitHub noreply@github.com 2023-09-03T04:38:43-06:00 gguf(python): Fix special vocab handling when id < 0 (#2984)
b7f2aa9e512c3be2e863d877cbb1056d7c4a03f8 73a12a6344d5da4d8e2eba5d12221b8bc6895931 Georgi Gerganov ggerganov@gmail.com 2023-09-03T13:23:33+03:00 GitHub noreply@github.com 2023-09-03T13:23:33+03:00 metal : restore 363f0bf and fix reduce in F16_F32 kernels (#2986)
73a12a6344d5da4d8e2eba5d12221b8bc6895931 37301347767d555d0a66c043ce4ef6ead8e61c55 Alon alonfaraj@gmail.com 2023-09-03T13:19:01+03:00 GitHub noreply@github.com 2023-09-03T13:19:01+03:00 cov : disable comment in PRs (#2989)
37301347767d555d0a66c043ce4ef6ead8e61c55 d9151e6f570eb20bfd54427bd8a337d9b1a08018 opparco parco.opaai@gmail.com 2023-09-03T19:18:09+09:00 GitHub noreply@github.com 2023-09-03T13:18:09+03:00 llama : fix bpe tokenize from byte (#2889)
d9151e6f570eb20bfd54427bd8a337d9b1a08018 afc43d5f82588d2ed71ea104e8262f5e5da13980 Georgi Gerganov ggerganov@gmail.com 2023-09-03T12:40:56+03:00 Georgi Gerganov ggerganov@gmail.com 2023-09-03T12:40:56+03:00 metal : revert 6af0bab until we fix it
afc43d5f82588d2ed71ea104e8262f5e5da13980 6460f758dbd472653296044d36bed8c4554988f5 Alon alonfaraj@gmail.com 2023-09-03T11:48:49+03:00 GitHub noreply@github.com 2023-09-03T11:48:49+03:00 cov : add Code Coverage and codecov.io integration (#2928)
6460f758dbd472653296044d36bed8c4554988f5 ca82cf7bac0c91d03e3d320b3a865dd006f854ac Wentai Zhang rchardx@gmail.com 2023-09-03T16:46:44+08:00 GitHub noreply@github.com 2023-09-03T11:46:44+03:00 opencl : fix a bug in ggml_cl_pool_malloc() for ggml_cl_mul_mat_f32() (#2955)
ca82cf7bac0c91d03e3d320b3a865dd006f854ac 6a31a3bd9806c85ed08266f6ab65181da0f30d03 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-09-03T11:06:22+03:00 GitHub noreply@github.com 2023-09-03T11:06:22+03:00 metal : more optimizations (#2959)
6a31a3bd9806c85ed08266f6ab65181da0f30d03 cff7b0bf07cb46e1ad4fd199f6bdeb538925c8c4 kchro3 62481661+kchro3@users.noreply.github.com 2023-09-02T23:21:05-07:00 GitHub noreply@github.com 2023-09-03T09:21:05+03:00 swift : add support for k-quants (#2983)
cff7b0bf07cb46e1ad4fd199f6bdeb538925c8c4 340af42f09a80e32f4998857b4f0543e41124525 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-09-02T23:52:13-06:00 GitHub noreply@github.com 2023-09-03T08:52:13+03:00 convert.py : BPE fixes (#2938)
340af42f09a80e32f4998857b4f0543e41124525 c42f0ec6b344e14bd81c8612ab1445b3ff77358b Ido S ido.pluto@gmail.com 2023-09-03T08:50:51+03:00 GitHub noreply@github.com 2023-09-03T08:50:51+03:00 docs : add `catai` to `README.md` (#2967)
c42f0ec6b344e14bd81c8612ab1445b3ff77358b 2753415afdaf22a18c49608bd9d93cfffc05d435 momonga 115213907+mmnga@users.noreply.github.com 2023-09-03T14:36:28+09:00 GitHub noreply@github.com 2023-09-03T08:36:28+03:00 examples : fix gpt-neox (#2943)
2753415afdaf22a18c49608bd9d93cfffc05d435 bc054af97ac68a4b726e972cb283eb9565253ed5 kchro3 62481661+kchro3@users.noreply.github.com 2023-09-02T22:27:25-07:00 GitHub noreply@github.com 2023-09-03T08:27:25+03:00 swift : add missing c file to Package.swift (#2978)
bc054af97ac68a4b726e972cb283eb9565253ed5 3358c381f6251bf6e65855e1c93bfaa9ec82ddb3 Cebtenzzre cebtenzzre@gmail.com 2023-09-03T01:26:59-04:00 GitHub noreply@github.com 2023-09-03T08:26:59+03:00 make : support overriding CFLAGS/CXXFLAGS/CPPFLAGS/LDFLAGS (#2886)
3358c381f6251bf6e65855e1c93bfaa9ec82ddb3 52315a421674ff64305dbf082f69e4ec77f0a3f3 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-09-02T11:53:55-06:00 GitHub noreply@github.com 2023-09-02T11:53:55-06:00 logging: Fix creating empty file even when disabled (#2966)
52315a421674ff64305dbf082f69e4ec77f0a3f3 8b56b4f2c396eae1f4417e5a859557fed989e0ee bandoti 141645996+bandoti@users.noreply.github.com 2023-09-02T09:53:18-03:00 GitHub noreply@github.com 2023-09-02T15:53:18+03:00 readme : update clblast instructions (#2903)
8b56b4f2c396eae1f4417e5a859557fed989e0ee 21f3d1be867b4d7be07c26f5da6e4bc69bcf4d27 Karsten Weiss knweiss@gmail.com 2023-09-02T14:29:09+02:00 GitHub noreply@github.com 2023-09-02T15:29:09+03:00 metal : show all Metal device instances in the system (#2952)
21f3d1be867b4d7be07c26f5da6e4bc69bcf4d27 571083f508266c4eb5cb5457d836df5dd3c173ce Jhen-Jie Hong iainst0409@gmail.com 2023-09-02T20:23:45+08:00 GitHub noreply@github.com 2023-09-02T15:23:45+03:00 k-quants : fix build on armv7 (android only) (#2920)
571083f508266c4eb5cb5457d836df5dd3c173ce f04d0028444bc9b3d4225fba47e19d4c3aeb3741 Jhen-Jie Hong iainst0409@gmail.com 2023-09-02T08:31:46+08:00 GitHub noreply@github.com 2023-09-02T08:31:46+08:00 server : avoid aniprompt in probabilities of final response (#2849)
f04d0028444bc9b3d4225fba47e19d4c3aeb3741 69fdbb9abc8907dd2a9ffdd840cba92d678a660a Engininja2 139037756+Engininja2@users.noreply.github.com 2023-09-01T15:33:19-06:00 GitHub noreply@github.com 2023-09-01T23:33:19+02:00 cuda : vsubss4 for older versions of ROCm/clang (#2942)
69fdbb9abc8907dd2a9ffdd840cba92d678a660a 5d6f19f16b2173afe2d5c6aee2f5c9fc31038eba ZHAOKAI WANG sanxianwei@163.com 2023-09-01T22:06:44+08:00 GitHub noreply@github.com 2023-09-01T17:06:44+03:00 readme : quick start command fix (#2908)
5d6f19f16b2173afe2d5c6aee2f5c9fc31038eba 0d5893668625456c94bbadfddc53fc69cd51c223 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-09-01T08:02:48-06:00 GitHub noreply@github.com 2023-09-01T08:02:48-06:00 Allow quantize to only copy tensors, some other improvements (#2931)
0d5893668625456c94bbadfddc53fc69cd51c223 6c9c23429bf4e4fcaaddbebadc4638558430a7f2 Georgi Gerganov ggerganov@gmail.com 2023-09-01T17:00:40+03:00 Georgi Gerganov ggerganov@gmail.com 2023-09-01T17:01:11+03:00 llama2c : rename function
6c9c23429bf4e4fcaaddbebadc4638558430a7f2 ee8654bcd0146708988a703e54406d5b553712ea Cebtenzzre cebtenzzre@gmail.com 2023-09-01T09:53:14-04:00 GitHub noreply@github.com 2023-09-01T16:53:14+03:00 make : use unaligned vector moves on MinGW (#2945)
ee8654bcd0146708988a703e54406d5b553712ea 49bb9cbe0f598bc43be539b0df8eafb2130cfad3 m3ndax adrian.goessl@outlook.com 2023-09-01T15:47:27+02:00 GitHub noreply@github.com 2023-09-01T16:47:27+03:00 minor : add const qualifiers (#2853)
49bb9cbe0f598bc43be539b0df8eafb2130cfad3 ef156499721c67748cde01a5436cb6f0648bb4b4 Konstantin Herud konstantin.herud@denkbares.com 2023-09-01T15:36:14+02:00 GitHub noreply@github.com 2023-09-01T16:36:14+03:00 docs : add java-llama.cpp to README.md (#2935)
ef156499721c67748cde01a5436cb6f0648bb4b4 d8d6977f48f1fa402ade38ad32c5b5fb1358d059 Cebtenzzre cebtenzzre@gmail.com 2023-09-01T09:34:50-04:00 GitHub noreply@github.com 2023-09-01T16:34:50+03:00 build : fix most gcc and clang warnings (#2861)
d8d6977f48f1fa402ade38ad32c5b5fb1358d059 5aec2cfaac386eb09aebb75b805860828f00de91 Ben Siraphob bensiraphob@gmail.com 2023-09-01T09:32:14-04:00 GitHub noreply@github.com 2023-09-01T16:32:14+03:00 examples : add C grammar (#2357)
5aec2cfaac386eb09aebb75b805860828f00de91 13268c533177a4dc76bce0b465645d74f0d51d55 Tameem 113388789+AhmadTameem@users.noreply.github.com 2023-09-01T18:27:40+05:00 GitHub noreply@github.com 2023-09-01T16:27:40+03:00 ggml : add RISC-V vector intrinsics support (#2929)
13268c533177a4dc76bce0b465645d74f0d51d55 4dcd47d71df8ca4edcc31302744bd93f0c31298e Georgi Gerganov ggerganov@gmail.com 2023-09-01T13:42:41+03:00 GitHub noreply@github.com 2023-09-01T13:42:41+03:00 metal : slight speed-up for add and mul kernels (#2917)
4dcd47d71df8ca4edcc31302744bd93f0c31298e 18705a30ef3d6a89e1d7c6cb8cfe8633f760cb53 staviq staviq@gmail.com 2023-09-01T11:07:06+02:00 GitHub noreply@github.com 2023-09-01T12:07:06+03:00 logs : fix mingw-like builds (fixes #2898) (#2911)
18705a30ef3d6a89e1d7c6cb8cfe8633f760cb53 e8d91589258f9204397a7ac5f9b3c857835c98f8 Cebtenzzre cebtenzzre@gmail.com 2023-09-01T05:03:49-04:00 GitHub noreply@github.com 2023-09-01T12:03:49+03:00 llama2c : fix segfault and alloc-dealloc-mismatch (#2913)
e8d91589258f9204397a7ac5f9b3c857835c98f8 bce1fef328941499dc0acb76cc7fd7ac90449c2f Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-09-01T11:15:57+03:00 GitHub noreply@github.com 2023-09-01T11:15:57+03:00 metal: somewhat faster f16 x f32 matrix multiply kernel (#2951)
bce1fef328941499dc0acb76cc7fd7ac90449c2f 528134dd0267838d9c0250cf1d9621631dff09b2 Cebtenzzre cebtenzzre@gmail.com 2023-08-31T22:13:51-04:00 GitHub noreply@github.com 2023-08-31T22:13:51-04:00 convert : fix another python 3.8 issue (#2949)
528134dd0267838d9c0250cf1d9621631dff09b2 aeefac4ff760acea5afe66fbfe8d7eca1937b79c slaren slarengh@gmail.com 2023-09-01T01:32:09+02:00 GitHub noreply@github.com 2023-09-01T01:32:09+02:00 remove convert-llama-7b-pth-to-gguf.py and convert-llama-hf-to-gguf.py (#2906)
aeefac4ff760acea5afe66fbfe8d7eca1937b79c e8422de39e4aa2f7e50574124b060a80607e654a Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-08-31T16:49:24-06:00 GitHub noreply@github.com 2023-08-31T16:49:24-06:00 scripts: Use local gguf package when running from repo (#2927)
e8422de39e4aa2f7e50574124b060a80607e654a 92d0b751a77a089e650983e9f1564ef4d31b32b9 DannyDaemonic DannyDaemonic@gmail.com 2023-08-31T04:21:45-07:00 GitHub noreply@github.com 2023-08-31T04:21:45-07:00 @vxiiduu's fix for PrefetchVirtualMemory (#2930)
92d0b751a77a089e650983e9f1564ef4d31b32b9 8afe2280009ecbfc9de2c93b8f41283dc810609a Cebtenzzre cebtenzzre@gmail.com 2023-08-31T01:02:23-04:00 GitHub noreply@github.com 2023-08-31T08:02:23+03:00 convert : fix python 3.8 support, modernize type annotations (#2916)
8afe2280009ecbfc9de2c93b8f41283dc810609a 71d6975559acfd6c8407a4ef8275a9979c737765 Johannes Gäßler johannesg@5d6.de 2023-08-30T21:46:19+02:00 GitHub noreply@github.com 2023-08-30T21:46:19+02:00 CUDA: mul_mat_q=true llama_context_params default (#2912)
71d6975559acfd6c8407a4ef8275a9979c737765 b532a69b2fd08067f34f32f37a2fd9b37678a34a Henri Vasserman henv@hot.ee 2023-08-30T19:14:53+03:00 GitHub noreply@github.com 2023-08-30T19:14:53+03:00 [Docker] fix tools.sh argument passing. (#2884)
b532a69b2fd08067f34f32f37a2fd9b37678a34a c90d135eb433cf0d40fb95e46a48d1391d2352b5 Georgi Gerganov ggerganov@gmail.com 2023-08-30T13:29:40+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-30T13:29:40+03:00 convert.py : use dir name to name the llama
c90d135eb433cf0d40fb95e46a48d1391d2352b5 0d1c706181cd31e7f368dd14eeb16c1a2569e4df Georgi Gerganov ggerganov@gmail.com 2023-08-30T12:52:46+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-30T12:53:24+03:00 examples : fix underscore in beam-search + .gitignore (close #2900)
0d1c706181cd31e7f368dd14eeb16c1a2569e4df 950929442070874d45561d2a4b68b010457767de M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-08-30T12:47:40+03:00 GitHub noreply@github.com 2023-08-30T12:47:40+03:00 gguf : add workflow for Pypi publishing (#2896)
950929442070874d45561d2a4b68b010457767de 35092fb54712d032860f3976a6fc1ae1f84a4a28 alonfaraj alonfaraj@gmail.com 2023-08-30T12:42:51+03:00 GitHub noreply@github.com 2023-08-30T12:42:51+03:00 make : add test and update CI (#2897)
35092fb54712d032860f3976a6fc1ae1f84a4a28 dc07dc492ef9640bbb82904d7c7679f7bdcf6d76 Gilad S giladgd@users.noreply.github.com 2023-08-30T11:40:12+03:00 GitHub noreply@github.com 2023-08-30T11:40:12+03:00 docs : add `node-llama-cpp` to `README.md` (#2885)
dc07dc492ef9640bbb82904d7c7679f7bdcf6d76 ad9ddcff6ef322db5cf13785bd7c856b610d242e Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-08-30T02:25:50-06:00 GitHub noreply@github.com 2023-08-30T11:25:50+03:00 convert : various script cleanups/fixes + merges and special token handling (#2842)
ad9ddcff6ef322db5cf13785bd7c856b610d242e 8341a25957b319a03d4a811176cd5ad7f2b0fbd4 chaihahaha chai836275709@gmail.com 2023-08-30T14:50:55+08:00 GitHub noreply@github.com 2023-08-30T09:50:55+03:00 llm.vim : stop generation at multiple linebreaks, bind to <F2> (#2879)
8341a25957b319a03d4a811176cd5ad7f2b0fbd4 849408957c687cde4ab32c147107f643fc55130b staviq staviq@gmail.com 2023-08-30T08:29:32+02:00 GitHub noreply@github.com 2023-08-30T09:29:32+03:00 main : log file (#2748)
849408957c687cde4ab32c147107f643fc55130b 06abf8eebabe086ca4003dee2754ab45032cd3fd Cebtenzzre cebtenzzre@gmail.com 2023-08-30T02:20:26-04:00 GitHub noreply@github.com 2023-08-30T09:20:26+03:00 tests : add a C compliance test (#2848)
06abf8eebabe086ca4003dee2754ab45032cd3fd c03a243abf9f30889f31fefdfa94fe9d7034820c slaren slarengh@gmail.com 2023-08-29T23:24:42+02:00 GitHub noreply@github.com 2023-08-29T23:24:42+02:00 ggml : add view_src and view_offs to ggml_tensor for views (#2874)
c03a243abf9f30889f31fefdfa94fe9d7034820c fa3582f509a2715e80a473e79f88dcd1ebff44c2 slaren slarengh@gmail.com 2023-08-29T23:17:34+02:00 GitHub noreply@github.com 2023-08-29T23:17:34+02:00 remove outdated references to -eps and -gqa from README (#2881)
fa3582f509a2715e80a473e79f88dcd1ebff44c2 e37e69dcc3d52f21222a63cafed2a71b3f6b53c6 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-29T23:55:45+03:00 GitHub noreply@github.com 2023-08-29T23:55:45+03:00 Tell users attmepting to run perplexity with too few tokens to use more (#2882)
e37e69dcc3d52f21222a63cafed2a71b3f6b53c6 53885d7256909ec3e2176cdc2477f3986c15ec69 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-29T23:55:03+03:00 GitHub noreply@github.com 2023-08-29T23:55:03+03:00 10X faster BPE tokenizer (#2876)
53885d7256909ec3e2176cdc2477f3986c15ec69 bcce96ba4dd95482824700c4ce2455fe8c49055a maddes8cht 55592906+maddes8cht@users.noreply.github.com 2023-08-29T15:51:02+02:00 GitHub noreply@github.com 2023-08-29T16:51:02+03:00 py : fix "usage" messages (#2873)
bcce96ba4dd95482824700c4ce2455fe8c49055a 74e0caeb82fc9db77fa2cc93070bb919a9a935dd jameswu2014 545426914@qq.com 2023-08-29T17:48:41+08:00 GitHub noreply@github.com 2023-08-29T12:48:41+03:00 convert.py : fix baichuan7B support (#2870)
74e0caeb82fc9db77fa2cc93070bb919a9a935dd d4b5e16c32ba9c5fa6bbd035e80a99c113050cde Jhen-Jie Hong iainst0409@gmail.com 2023-08-29T17:30:10+08:00 GitHub noreply@github.com 2023-08-29T12:30:10+03:00 readme : add react-native binding (#2869)
d4b5e16c32ba9c5fa6bbd035e80a99c113050cde 3a007648f230ea37d6cca5e63850f04ebb12d2cf Cebtenzzre cebtenzzre@gmail.com 2023-08-29T04:42:41-04:00 GitHub noreply@github.com 2023-08-29T11:42:41+03:00 make : fix clang tests build, add missing examples (#2859)
3a007648f230ea37d6cca5e63850f04ebb12d2cf 611363ac791435497e66278dfe31ac8a4e11fa4f Georgi Gerganov ggerganov@gmail.com 2023-08-29T11:33:46+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-29T11:33:46+03:00 metal : add option to disable debug logs (close #2764)
611363ac791435497e66278dfe31ac8a4e11fa4f 95b6e5212f5e4e1419de1d833d7f8d788f9f2227 Georgi Gerganov ggerganov@gmail.com 2023-08-29T10:50:30+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-29T10:50:30+03:00 scripts : add pipefail
95b6e5212f5e4e1419de1d833d7f8d788f9f2227 44c117f41ee01c5ac8fb86bba041f08d8b87b46d Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2023-08-28T23:33:27-07:00 GitHub noreply@github.com 2023-08-29T09:33:27+03:00 added `struct` to llama_dump_timing_info_yaml's `llama_context` (#2857)
44c117f41ee01c5ac8fb86bba041f08d8b87b46d 43033b7bb4858da4f591715b3babdf906c9b7cbc xaedes xaedes@gmail.com 2023-08-28T21:51:47+02:00 GitHub noreply@github.com 2023-08-28T22:51:47+03:00 train : mem usage and other improvements (#2439)
43033b7bb4858da4f591715b3babdf906c9b7cbc 6b73ef120114beb5664ea94aab48d07ed248ee52 slaren slarengh@gmail.com 2023-08-28T19:19:18+02:00 GitHub noreply@github.com 2023-08-28T19:19:18+02:00 llama-bench : set locale to utf8 (#2832)
6b73ef120114beb5664ea94aab48d07ed248ee52 75fafcbcccc280a5b3883bc76d0a2dabf474d094 Johannes Gäßler johannesg@5d6.de 2023-08-28T17:59:39+02:00 GitHub noreply@github.com 2023-08-28T17:59:39+02:00 YAML result logging + preset script (#2657)
75fafcbcccc280a5b3883bc76d0a2dabf474d094 be475f60af1a54e8de81466ccc907d080cf6df1a alonfaraj alonfaraj@gmail.com 2023-08-28T18:38:35+03:00 GitHub noreply@github.com 2023-08-28T18:38:35+03:00 make : fix tests build (#2855)
be475f60af1a54e8de81466ccc907d080cf6df1a 3af6b86301ddfb11bb68e91dfc030b611b0d8426 grahameth 96447521+grahameth@users.noreply.github.com 2023-08-28T17:38:12+02:00 GitHub noreply@github.com 2023-08-28T18:38:12+03:00 llama.cpp : fix wrong vsnprintf call in MS compiler (#2856)
3af6b86301ddfb11bb68e91dfc030b611b0d8426 35feac6560387cf0484371af3d9b12bff678e0b9 Ronny Brendel ronnybrendel@gmail.com 2023-08-28T14:51:08+02:00 GitHub noreply@github.com 2023-08-28T15:51:08+03:00 ggml : tiny ggml_vec_dot_q4_K_q8_K AVX2 improvement (#2819)
35feac6560387cf0484371af3d9b12bff678e0b9 92b1bbd2ec43c82ec0530ba3c8758846c5790c75 Georgi Gerganov ggerganov@gmail.com 2023-08-28T14:24:53+03:00 GitHub noreply@github.com 2023-08-28T14:24:53+03:00 ggml : sync (mem align to header + conv_transpose_2d fixes + ggml_alloc) (#2852)
92b1bbd2ec43c82ec0530ba3c8758846c5790c75 dd0dc366dab10e8df28d3924e7f313b5c695e908 Johannes Gäßler johannesg@5d6.de 2023-08-28T13:23:55+02:00 GitHub noreply@github.com 2023-08-28T14:23:55+03:00 CUDA: fix RoPE asserts, block sizes (#2833)
dd0dc366dab10e8df28d3924e7f313b5c695e908 f55538c3ccba9b926846ef862fa830cea08c433e igarnier igarnier@protonmail.com 2023-08-28T10:19:59+02:00 GitHub noreply@github.com 2023-08-28T11:19:59+03:00 llama.h : add missing struct keyword for C compat in callback type (#2847)
f55538c3ccba9b926846ef862fa830cea08c433e ebcee207b6058b7f695bb5c203ad87b1066a9790 Georgi Gerganov ggerganov@gmail.com 2023-08-28T10:59:08+03:00 GitHub noreply@github.com 2023-08-28T10:59:08+03:00 metal : fix memory leak (#2762)
ebcee207b6058b7f695bb5c203ad87b1066a9790 3e8ff47af620a31e0810c58a41e4b089145982ef Cebtenzzre cebtenzzre@gmail.com 2023-08-28T02:32:25-04:00 GitHub noreply@github.com 2023-08-28T09:32:25+03:00 quantize : make output filename optional again (#2823)
3e8ff47af620a31e0810c58a41e4b089145982ef 103cfafc774f6feb3172b5d4d39681c965b17eba JohnnyB jboero@users.noreply.github.com 2023-08-28T07:31:24+01:00 GitHub noreply@github.com 2023-08-28T09:31:24+03:00 devops : added systemd units and set versioning to use date. (#2835)
103cfafc774f6feb3172b5d4d39681c965b17eba c10704d01e21e3dbe4d6ca1026ebff85349dd239 Georgi Gerganov ggerganov@gmail.com 2023-08-27T21:50:22+03:00 GitHub noreply@github.com 2023-08-27T21:50:22+03:00 gguf : fix strings to not be null-terminated (#2839)
c10704d01e21e3dbe4d6ca1026ebff85349dd239 230d46c723edf5999752e4cb67fd94edb19ef9c7 Georgi Gerganov ggerganov@gmail.com 2023-08-27T18:55:41+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-27T18:55:41+03:00 llama : fix MPI threads (close #2827)
230d46c723edf5999752e4cb67fd94edb19ef9c7 463173a6c0ff353055eb90665794884c888c790f Olivier Chafik ochafik@users.noreply.github.com 2023-08-27T15:13:31+01:00 GitHub noreply@github.com 2023-08-27T17:13:31+03:00 examples : update llama2.c converter to read vocab and write models in GGUF format (#2751)
463173a6c0ff353055eb90665794884c888c790f eaa13a48ff4136f01c1cdb79cacd61b67ec53095 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-27T16:50:33+03:00 GitHub noreply@github.com 2023-08-27T16:50:33+03:00 llama : speedup tokenization (#2831)
eaa13a48ff4136f01c1cdb79cacd61b67ec53095 da7455d0467b5f5cc2e45d0dcffaf098df13db63 Georgi Gerganov ggerganov@gmail.com 2023-08-27T16:40:48+03:00 GitHub noreply@github.com 2023-08-27T16:40:48+03:00 falcon : fix CUDA inference by making K and Q contiguous (#2830)
da7455d0467b5f5cc2e45d0dcffaf098df13db63 25423e9185b7c2a1881ed8f85cc752a12370be9d Georgi Gerganov ggerganov@gmail.com 2023-08-27T15:52:34+03:00 GitHub noreply@github.com 2023-08-27T15:52:34+03:00 readme : fix headings
25423e9185b7c2a1881ed8f85cc752a12370be9d a6d1189fdd4c1ab4ba23f9d777f8950901dcffb2 Georgi Gerganov ggerganov@gmail.com 2023-08-27T15:24:40+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-27T15:24:58+03:00 scripts : helper convert script
a6d1189fdd4c1ab4ba23f9d777f8950901dcffb2 c48c5bb0b06385f6c708339188d2aaf2bc278477 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-27T15:19:59+03:00 GitHub noreply@github.com 2023-08-27T15:19:59+03:00 k_quants tuning for Falcon-7b (#2816)
c48c5bb0b06385f6c708339188d2aaf2bc278477 d0cee0d36d5be95a0d9088b674dbb27354107221 Georgi Gerganov ggerganov@gmail.com 2023-08-27T14:44:35+03:00 GitHub noreply@github.com 2023-08-27T14:44:35+03:00 readme : update hot topics
d0cee0d36d5be95a0d9088b674dbb27354107221 edd4c1481708fcd788b0e423268304fd26e2b125 Georgi Gerganov ggerganov@gmail.com 2023-08-27T14:19:54+03:00 GitHub noreply@github.com 2023-08-27T14:19:54+03:00 gguf : add 64-bit support (GGUF v2) (#2821)
edd4c1481708fcd788b0e423268304fd26e2b125 1591e2e590762011b43b10a9b6e04f13f98f2aa5 Georgi Gerganov ggerganov@gmail.com 2023-08-27T14:19:19+03:00 GitHub noreply@github.com 2023-08-27T14:19:19+03:00 llama : more tokenizer fixes (#2810)
1591e2e590762011b43b10a9b6e04f13f98f2aa5 789c8c945a2814e1487e18e68823d9926e3b1454 Przemysław Pawełczyk przemoc@gmail.com 2023-08-27T10:10:25+02:00 GitHub noreply@github.com 2023-08-27T11:10:25+03:00 ggml : detect SSSE3 (#2825)
789c8c945a2814e1487e18e68823d9926e3b1454 c1ac54b77aaba10d029084d152be786102010eb2 slaren slarengh@gmail.com 2023-08-27T09:03:27+02:00 GitHub noreply@github.com 2023-08-27T10:03:27+03:00 ci : add LoRA test to CI (#2650)
c1ac54b77aaba10d029084d152be786102010eb2 730d9c681e339b76407659344e5a2cd50af7d7d5 Bruce MacDonald brucewmacdonald@gmail.com 2023-08-26T16:11:45-07:00 GitHub noreply@github.com 2023-08-27T07:11:45+08:00 server : add `/detokenize` endpoint (#2802)
730d9c681e339b76407659344e5a2cd50af7d7d5 c7d92e6dfec3f54849f3a0ba373054d29f321ea2 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-08-26T14:13:36-06:00 GitHub noreply@github.com 2023-08-26T23:13:36+03:00 convert.py : advanced option (#2753)
c7d92e6dfec3f54849f3a0ba373054d29f321ea2 61d1a2895eeca55e0c8b7018492f6ab9c90cff78 Tim Miller drasticactions@users.noreply.github.com 2023-08-27T03:27:07+09:00 GitHub noreply@github.com 2023-08-26T21:27:07+03:00 llama : use Unicode Escape Sequence to replace encoded characters (#2814)
61d1a2895eeca55e0c8b7018492f6ab9c90cff78 741ca7dd1cec0a0349494742b9083d6ef4cd73c5 Tungsten842 quantmint@protonmail.com 2023-08-26T20:19:44+02:00 GitHub noreply@github.com 2023-08-26T21:19:44+03:00 flake.nix : add rocm support and cleanup (#2808)
741ca7dd1cec0a0349494742b9083d6ef4cd73c5 72f895c923ba98b8f2af294440206f35915c0501 Cebtenzzre cebtenzzre@gmail.com 2023-08-26T14:17:51-04:00 GitHub noreply@github.com 2023-08-26T21:17:51+03:00 llama : move #includes out of _GNU_SOURCE conditional (#2817)
72f895c923ba98b8f2af294440206f35915c0501 50526f37eba0b28336700890242ff282b949cd83 Dr. Tom Murphy VII Ph.D 499244+tom7@users.noreply.github.com 2023-08-26T14:12:56-04:00 GitHub noreply@github.com 2023-08-26T21:12:56+03:00 main : fix bug (penalize_nl=false doesn't work) + suppress warning on mingw (#1528)
50526f37eba0b28336700890242ff282b949cd83 04f4b1eb10f3e25750ca3e530265ce2841730e6b Cebtenzzre cebtenzzre@gmail.com 2023-08-26T12:53:52-04:00 GitHub noreply@github.com 2023-08-26T19:53:52+03:00 llama : use std::abs in llama_sample_tail_free (#2800)
04f4b1eb10f3e25750ca3e530265ce2841730e6b 7592375403a0bd0456d5ec2cdf8350e591f04fb0 Georgi Gerganov ggerganov@gmail.com 2023-08-26T17:37:35+03:00 GitHub noreply@github.com 2023-08-26T17:37:35+03:00 k-quants : remove unnecessary tensor shape restrictions (#2811)
7592375403a0bd0456d5ec2cdf8350e591f04fb0 771551a793c9976ed9cdfe7b8c69536af32af9f9 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-26T17:27:49+03:00 GitHub noreply@github.com 2023-08-26T17:27:49+03:00 Better perplexity for 2- and 3-bit quantization for LLaMA-v2-70B (#2807)
771551a793c9976ed9cdfe7b8c69536af32af9f9 f305bad11e10ad09e396faed2e37f4f845f5d566 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-26T16:48:53+03:00 GitHub noreply@github.com 2023-08-26T16:48:53+03:00 Fix HellaSwag (#2805)
f305bad11e10ad09e396faed2e37f4f845f5d566 a2ca4e9de9da45ed0bb1c34935d5ec80cebc22d5 Volodymyr Vitvitskyi 72226+signalpillar@users.noreply.github.com 2023-08-26T14:25:39+01:00 GitHub noreply@github.com 2023-08-26T16:25:39+03:00 flake : build llama.cpp on Intel with nix (#2795)
a2ca4e9de9da45ed0bb1c34935d5ec80cebc22d5 2ba83c8685177faea3399db9564f9c52df75c366 Nigel Bosch pnigelb@gmail.com 2023-08-26T07:11:17-05:00 GitHub noreply@github.com 2023-08-26T14:11:17+02:00 Handle null rope scaling value (#2793)
2ba83c8685177faea3399db9564f9c52df75c366 bae5c5f679e043371bc2b4dffff8d4964d6cb953 klosax 131523366+klosax@users.noreply.github.com 2023-08-26T13:45:53+02:00 GitHub noreply@github.com 2023-08-26T13:45:53+02:00 Fix spm whitespaces (#2806)
bae5c5f679e043371bc2b4dffff8d4964d6cb953 232caf3c1581a6cb023571780ff41dc2d66d1ca0 lon 114724657+longregen@users.noreply.github.com 2023-08-26T10:07:43+02:00 GitHub noreply@github.com 2023-08-26T16:07:43+08:00 examples : skip unnecessary external lib in server README.md how-to (#2804)
232caf3c1581a6cb023571780ff41dc2d66d1ca0 d046dcee081118c9071bbc63dacdb359a58c467a Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2023-08-25T09:17:15-07:00 GitHub noreply@github.com 2023-08-25T19:17:15+03:00 llama : fix struct decl (#2790)
d046dcee081118c9071bbc63dacdb359a58c467a c82742ac9cd96fd34aa961978805c1d8a361d589 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-25T19:05:02+03:00 GitHub noreply@github.com 2023-08-25T19:05:02+03:00 Faster perplexity computation (#2786)
c82742ac9cd96fd34aa961978805c1d8a361d589 28b2c996ca0ab90a5669946084f13443ec98e241 Matt Pulver matt.pulver@heavy.ai 2023-08-25T11:18:48-04:00 GitHub noreply@github.com 2023-08-25T18:18:48+03:00 llama : add llama_beam_search() (#2267)
28b2c996ca0ab90a5669946084f13443ec98e241 154725c5436808e5c519685d0279e850596dbe62 Nigel Bosch pnigelb@gmail.com 2023-08-25T09:41:52-05:00 GitHub noreply@github.com 2023-08-25T16:41:52+02:00 convert.py : Get rope scale from HuggingFace models (#2772)
154725c5436808e5c519685d0279e850596dbe62 12e2e33a977af73e75885eeee91c5575a77f4e5f slaren slarengh@gmail.com 2023-08-25T15:16:19+02:00 GitHub noreply@github.com 2023-08-25T15:16:19+02:00 llama-bench : add model sizes (#2771)
12e2e33a977af73e75885eeee91c5575a77f4e5f 29674ab4e847fcaba60cc6558f0d46d5f74ae279 slaren slarengh@gmail.com 2023-08-25T14:08:53+02:00 GitHub noreply@github.com 2023-08-25T14:08:53+02:00 convert.py : export rope freq_base when converting CodeLlama from an HF model (#2773)
29674ab4e847fcaba60cc6558f0d46d5f74ae279 5439a0ab57c16b556ffa91a0953df5e46b1e7fb4 Jhen-Jie Hong iainst0409@gmail.com 2023-08-25T18:32:45+08:00 GitHub noreply@github.com 2023-08-25T18:32:45+08:00 server : display token probabilities in the UI (#2489)
5439a0ab57c16b556ffa91a0953df5e46b1e7fb4 8194cd8772c58b8a43aa07a2fc468f5366d7e320 Georgi Gerganov ggerganov@gmail.com 2023-08-25T13:03:25+03:00 GitHub noreply@github.com 2023-08-25T13:03:25+03:00 ci : pip install gguf in editable mode (#2782)
8194cd8772c58b8a43aa07a2fc468f5366d7e320 6bbc598a632560cb45dd2c51ad403bda8723b629 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-08-25T12:43:41+03:00 GitHub noreply@github.com 2023-08-25T12:43:41+03:00 gguf : export objects to user code (#2780)
6bbc598a632560cb45dd2c51ad403bda8723b629 3f460a2b723c8b936ac29ecfd02f244b3adeba55 Henri Vasserman henv@hot.ee 2023-08-25T12:09:42+03:00 GitHub noreply@github.com 2023-08-25T12:09:42+03:00 ROCm Port (#1087)
3f460a2b723c8b936ac29ecfd02f244b3adeba55 87e3733f24a85d894cc16e1cbdfa1ea1e81a76f3 Georgi Gerganov ggerganov@gmail.com 2023-08-25T11:55:59+03:00 GitHub noreply@github.com 2023-08-25T11:55:59+03:00 cuda : add RoPE kernel for mode == 2 (NeoX) (#2760)
87e3733f24a85d894cc16e1cbdfa1ea1e81a76f3 b91ad7f46134d0d051dc516eb59a76f402de55c2 M. Yusuf Sarıgöz yusufsarigoz@gmail.com 2023-08-25T09:26:05+03:00 GitHub noreply@github.com 2023-08-25T09:26:05+03:00 gguf : make gguf pip-installable
b91ad7f46134d0d051dc516eb59a76f402de55c2 2e5f70a25fc4576e9ed78603fe493eb7702c37a3 Shouzheng Liu lshzh.hi@gmail.com 2023-08-25T01:58:00-04:00 GitHub noreply@github.com 2023-08-25T08:58:00+03:00 ggml-alloc : enlarge size of parse_seq (#2776)
2e5f70a25fc4576e9ed78603fe493eb7702c37a3 d0f77b1353fc820d1ff1e6b87bc6bedde315938d Marcus Dunn 51931484+MarcusDunn@users.noreply.github.com 2023-08-24T14:49:30-07:00 GitHub noreply@github.com 2023-08-24T23:49:30+02:00 Added `enum` to `llama_token_get_type` return type (#2774)
d0f77b1353fc820d1ff1e6b87bc6bedde315938d 0d3094f0c742ce61f84feb6e4f0b59beee6194d7 slaren slarengh@gmail.com 2023-08-24T21:10:39+02:00 GitHub noreply@github.com 2023-08-24T21:10:39+02:00 convert.py : try to determine n_ctx automatically for CodeLlama (#2770)
0d3094f0c742ce61f84feb6e4f0b59beee6194d7 01f2224682b08185af609b28b1268b95c8b4cfa2 slaren slarengh@gmail.com 2023-08-24T20:04:05+02:00 GitHub noreply@github.com 2023-08-24T21:04:05+03:00 gguf : add rope_freq_base parameter for CodeLlama (#2769)
01f2224682b08185af609b28b1268b95c8b4cfa2 38b16dfca6e5032e6cfb90c1653bf1ba4cf647b4 Georgi Gerganov ggerganov@gmail.com 2023-08-24T19:58:30+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-24T19:58:30+03:00 falcon : write file type
38b16dfca6e5032e6cfb90c1653bf1ba4cf647b4 8f8c28e89cb9531211783da697d6e7c445e2af1d Shouzheng Liu lshzh.hi@gmail.com 2023-08-24T12:27:25-04:00 GitHub noreply@github.com 2023-08-24T19:27:25+03:00 metal : bug-fix when enable ggml-alloc (#2757)
8f8c28e89cb9531211783da697d6e7c445e2af1d 7694adda8d1111b3cf758ad6c91d754a0a4cacff Georgi Gerganov ggerganov@gmail.com 2023-08-24T19:26:19+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-24T19:26:47+03:00 convert : auto-determine model name based on dir + scripts update
7694adda8d1111b3cf758ad6c91d754a0a4cacff fea95c682d0028fdd25853bea58035794a0c964d Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-08-24T10:11:13-06:00 GitHub noreply@github.com 2023-08-24T10:11:13-06:00 Fix for main example getting stuck when -n -2 and --interactive (#2767)
fea95c682d0028fdd25853bea58035794a0c964d ef955fbd230c571cc1cda0d19baaeec347523175 slaren slarengh@gmail.com 2023-08-24T17:44:11+02:00 GitHub noreply@github.com 2023-08-24T17:44:11+02:00 fix convert.py for codellama, add llama 34B to the list of recognized models (#2768)
ef955fbd230c571cc1cda0d19baaeec347523175 d67777c202c03bcb74372690599ef3c03affb3ba DannyDaemonic DannyDaemonic@gmail.com 2023-08-24T06:58:02-07:00 GitHub noreply@github.com 2023-08-24T15:58:02+02:00 Tag release with build number (#2732)
d67777c202c03bcb74372690599ef3c03affb3ba c3e53b421a9910548be0345f85712c535f467a98 Georgi Gerganov ggerganov@gmail.com 2023-08-24T16:19:57+03:00 GitHub noreply@github.com 2023-08-24T16:19:57+03:00 metal : add Q8_0 support (#2763)
c3e53b421a9910548be0345f85712c535f467a98 6e91a1b0706c2e0e52b9d9be7ee82d3c1e7a33c1 Georgi Gerganov ggerganov@gmail.com 2023-08-24T12:26:01+03:00 GitHub noreply@github.com 2023-08-24T12:26:01+03:00 llama : escape all U+2581 in a string (#2750)
6e91a1b0706c2e0e52b9d9be7ee82d3c1e7a33c1 44d5462b5cddc1c5cbcd7647646f7b55b175b01f Evan Jones evan.q.jones@gmail.com 2023-08-24T00:07:13-04:00 GitHub noreply@github.com 2023-08-24T07:07:13+03:00 llama : fix grammar sometimes generating null char (#2756)
44d5462b5cddc1c5cbcd7647646f7b55b175b01f c7868b075377c8c3fa916ea7c1aca600f44bed55 Georgi Gerganov ggerganov@gmail.com 2023-08-23T23:44:19+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-23T23:44:19+03:00 readme : fix link
c7868b075377c8c3fa916ea7c1aca600f44bed55 79da24b58c1ea72340e64f799a4717d372207676 Georgi Gerganov ggerganov@gmail.com 2023-08-23T23:43:00+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-23T23:43:00+03:00 minor : fix trailing whitespace
79da24b58c1ea72340e64f799a4717d372207676 cf658adc832badaaa2ca119fe86070e5a830f8f6 Georgi Gerganov ggerganov@gmail.com 2023-08-23T23:41:16+03:00 GitHub noreply@github.com 2023-08-23T23:41:16+03:00 readme : update hot topics
cf658adc832badaaa2ca119fe86070e5a830f8f6 a192860cfec89a38d59a943623bf595b1fe4495b Georgi Gerganov ggerganov@gmail.com 2023-08-23T23:08:04+03:00 GitHub noreply@github.com 2023-08-23T23:08:04+03:00 llm : add Falcon support (#2717)
a192860cfec89a38d59a943623bf595b1fe4495b 95385241a91a616788a3bb76d12c9b7b2379ca2d Georgi Gerganov ggerganov@gmail.com 2023-08-23T22:37:39+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-23T22:37:39+03:00 minor : fix trailing whitespace
95385241a91a616788a3bb76d12c9b7b2379ca2d 335acd2ffd7b04501c6d8773ab9fcee6e7bf8639 Olivier Chafik ochafik@users.noreply.github.com 2023-08-23T20:33:05+01:00 GitHub noreply@github.com 2023-08-23T22:33:05+03:00 examples : restore the functionality to import llama2.c models (#2685)
335acd2ffd7b04501c6d8773ab9fcee6e7bf8639 5290c38e6e9b66ee2b543e560e301c1a1a90929c slaren slarengh@gmail.com 2023-08-23T16:46:54+02:00 GitHub noreply@github.com 2023-08-23T16:46:54+02:00 fix convert-lora-to-ggml.py (#2738)
5290c38e6e9b66ee2b543e560e301c1a1a90929c cc34dbda9681418a2b18382446b90cdcec398d82 klosax 131523366+klosax@users.noreply.github.com 2023-08-23T16:46:03+02:00 GitHub noreply@github.com 2023-08-23T16:46:03+02:00 main : insert bos if no tokens (#2727)
cc34dbda9681418a2b18382446b90cdcec398d82 7c2227a1972a4add4b5c118e4914c086513d0382 akawrykow 142945436+akawrykow@users.noreply.github.com 2023-08-23T07:31:34-07:00 GitHub noreply@github.com 2023-08-23T17:31:34+03:00 gitignore : fix for windows (#2729)
7c2227a1972a4add4b5c118e4914c086513d0382 f19dca04ea5fbf9a0b2753091d93464585d5c73b Cebtenzzre cebtenzzre@gmail.com 2023-08-23T10:29:09-04:00 GitHub noreply@github.com 2023-08-23T17:29:09+03:00 chmod : make scripts executable (#2675)
f19dca04ea5fbf9a0b2753091d93464585d5c73b 8207214b6a37a46526cee9e72d4c9092b9d1872f JohnnyB jboero@users.noreply.github.com 2023-08-23T15:28:22+01:00 GitHub noreply@github.com 2023-08-23T17:28:22+03:00 devops : RPM Specs (#2723)
8207214b6a37a46526cee9e72d4c9092b9d1872f 62959e740e8759d246ac8d09036950efde09981c Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-23T12:57:12+03:00 GitHub noreply@github.com 2023-08-23T12:57:12+03:00 Fix values shown in the quantize tool help (#2735)
62959e740e8759d246ac8d09036950efde09981c 7f7ddd5002040804e33fcdbde44aa22f8635f57d Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-23T12:56:42+03:00 GitHub noreply@github.com 2023-08-23T12:56:42+03:00 Strided perplexity (#2714)
7f7ddd5002040804e33fcdbde44aa22f8635f57d b8ad1b66b23f9b2e6e4531e9a62753323036a556 IgnacioFDM ignaciofdm@gmail.com 2023-08-23T06:31:09-03:00 GitHub noreply@github.com 2023-08-23T03:31:09-06:00 Fix ggml to gguf conversion on Windows (#2733)
b8ad1b66b23f9b2e6e4531e9a62753323036a556 f5fe98d11bdf9e7797bcfb05c0c3601ffc4b9d26 Xiao-Yong Jin jinxiaoyong@gmail.com 2023-08-23T02:12:12-05:00 GitHub noreply@github.com 2023-08-23T15:12:12+08:00 server : allow json array in prompt or content for direct token input (#2306)
f5fe98d11bdf9e7797bcfb05c0c3601ffc4b9d26 777f42ba18b29f25c71ff8de3ecf97b8017304c0 Evan Jones evan.q.jones@gmail.com 2023-08-22T21:01:57-04:00 GitHub noreply@github.com 2023-08-22T21:01:57-04:00 docs : add grammar docs (#2701)
777f42ba18b29f25c71ff8de3ecf97b8017304c0 46ef5b5fcf4c366e1fb27726b6394adbbf8fd0ea Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-08-22T17:39:39-06:00 GitHub noreply@github.com 2023-08-22T17:39:39-06:00 Improve handling of special tokens in GGML to GGUF converter (#2725)
46ef5b5fcf4c366e1fb27726b6394adbbf8fd0ea c63bb1d16a70c03440671b76954bb767513cead8 goerch jhr.walter@t-online.de 2023-08-22T23:10:42+02:00 GitHub noreply@github.com 2023-08-23T00:10:42+03:00 llama : fix whitespace escaping in tokenizer (#2724)
c63bb1d16a70c03440671b76954bb767513cead8 3b6cfe7c927df178ca3c11643c3ec93e143471c9 Johannes Gäßler johannesg@5d6.de 2023-08-22T22:47:05+02:00 GitHub noreply@github.com 2023-08-22T22:47:05+02:00 CUDA: use mul_mat_q kernels by default (#2683)
3b6cfe7c927df178ca3c11643c3ec93e143471c9 800c9635b4a9390126f397870f3a825fc7455bd1 Alex Petenchea alex.petenchea@gmail.com 2023-08-22T21:58:16+03:00 GitHub noreply@github.com 2023-08-22T21:58:16+03:00 convert.py : clarifying error message (#2718)
800c9635b4a9390126f397870f3a825fc7455bd1 deb7dfca4b9725cd295d1426db75fe8e0a6d5312 Jiahao Li liplus17@163.com 2023-08-23T02:27:06+08:00 GitHub noreply@github.com 2023-08-22T20:27:06+02:00 Fix CUDA softmax by subtracting max value before exp (#2665)
deb7dfca4b9725cd295d1426db75fe8e0a6d5312 bac66994cf356cf488078c056831396eb4ce31d5 Georgi Gerganov ggerganov@gmail.com 2023-08-22T20:05:59+03:00 GitHub noreply@github.com 2023-08-22T20:05:59+03:00 gguf : add ftype meta info to the model (#2710)
bac66994cf356cf488078c056831396eb4ce31d5 519c981f8b65ee6c87c2965539685ced0a17223b Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-22T19:14:09+03:00 GitHub noreply@github.com 2023-08-22T19:14:09+03:00 Quantization imrovements for k_quants (#2707)
519c981f8b65ee6c87c2965539685ced0a17223b 1123f7fbdfb8012e46f05e903e6f675922916378 slaren slarengh@gmail.com 2023-08-22T16:03:12+02:00 GitHub noreply@github.com 2023-08-22T16:03:12+02:00 embedding : evaluate prompt in batches (#2713)
1123f7fbdfb8012e46f05e903e6f675922916378 ef3f333d3775600d1646a9fa249aca532d15fb89 slaren slarengh@gmail.com 2023-08-22T15:25:19+02:00 GitHub noreply@github.com 2023-08-22T15:25:19+02:00 ggml-cuda : use graph allocator (#2684)
ef3f333d3775600d1646a9fa249aca532d15fb89 8e4364f2af9cd5d57240f23e83c0e29bc068bc02 Georgi Gerganov ggerganov@gmail.com 2023-08-22T14:22:08+03:00 GitHub noreply@github.com 2023-08-22T14:22:08+03:00 ggml : sync latest (SAM + SD operators, CUDA alibi) (#2709)
8e4364f2af9cd5d57240f23e83c0e29bc068bc02 1e3bc523d8053a77df3ac7126a84d0297ee97ef6 slaren slarengh@gmail.com 2023-08-22T09:56:03+02:00 GitHub noreply@github.com 2023-08-22T10:56:03+03:00 llama-bench : minor fixes (#2695)
1e3bc523d8053a77df3ac7126a84d0297ee97ef6 14b1d7e6f720dee41ce5a826376df738096d9033 Kylin 56434533+KyL0N@users.noreply.github.com 2023-08-22T15:14:23+08:00 GitHub noreply@github.com 2023-08-22T10:14:23+03:00 ggml : support CUDA's half type for aarch64(#1455) (#2670)
14b1d7e6f720dee41ce5a826376df738096d9033 226255b44ef2c2794bfac48d101d35a9c2dbb965 Shouzheng Liu lshzh.hi@gmail.com 2023-08-22T02:18:40-04:00 GitHub noreply@github.com 2023-08-22T09:18:40+03:00 metal : add missing barriers for mul-mat (#2699)
226255b44ef2c2794bfac48d101d35a9c2dbb965 930523c8e1cbbee5449c055daa894917fac6805e Jhen-Jie Hong iainst0409@gmail.com 2023-08-22T08:32:00+08:00 GitHub noreply@github.com 2023-08-22T08:32:00+08:00 server : fallback to default if client param is null (#2688)
930523c8e1cbbee5449c055daa894917fac6805e c8dba409e6d6a754090f08e6a862c5ffdd52e421 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-08-21T18:01:34-06:00 GitHub noreply@github.com 2023-08-21T18:01:34-06:00 Fix convert-llama-ggmlv3-to-gguf.py vocab conversion (#2698)
c8dba409e6d6a754090f08e6a862c5ffdd52e421 6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9 Georgi Gerganov ggerganov@gmail.com 2023-08-21T23:40:22+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-21T23:40:22+03:00 py : remove obsolete script
6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9 dadbed99e65252d79f81101a392d0d6497b86caa Georgi Gerganov ggerganov@gmail.com 2023-08-21T23:07:43+03:00 GitHub noreply@github.com 2023-08-21T23:07:43+03:00 gguf : new file format with flexible meta data (beta) (#2398)
dadbed99e65252d79f81101a392d0d6497b86caa cb1c0727bd59803b439b6a3af121c99e6393ff3d Shouzheng Liu lshzh.hi@gmail.com 2023-08-21T06:59:29-04:00 GitHub noreply@github.com 2023-08-21T13:59:29+03:00 metal : fix synchronization in new matrix multiplication kernel (#2686)
cb1c0727bd59803b439b6a3af121c99e6393ff3d 9e232f0234073358e7031c1b8d7aa45020469a3b Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-21T11:11:31+03:00 GitHub noreply@github.com 2023-08-21T11:11:31+03:00 HellaSwag: split token evaluation into batches if needed (#2681)
9e232f0234073358e7031c1b8d7aa45020469a3b 5e9ff54a675d163d9f42aad1b5b3e734f17b2701 slaren slarengh@gmail.com 2023-08-20T22:17:53+02:00 GitHub noreply@github.com 2023-08-20T22:17:53+02:00 ggml : move all type info to ggml_type_traits (#2663)
5e9ff54a675d163d9f42aad1b5b3e734f17b2701 1f0bccb27929e261744c979bc75114955da49e98 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-08-20T16:44:46+03:00 GitHub noreply@github.com 2023-08-20T16:44:46+03:00 More efficient Hellaswag implementation (#2677)
1f0bccb27929e261744c979bc75114955da49e98 f63564adfaa157ca387071d6b9a06cfaef0ef576 Georgi Gerganov ggerganov@gmail.com 2023-08-19T00:45:36+03:00 GitHub noreply@github.com 2023-08-19T05:45:36+08:00 server : better default prompt (#2646)
f63564adfaa157ca387071d6b9a06cfaef0ef576 2d8b76a110d76ff6b5728ff0af8477531e4db60e Jhen-Jie Hong iainst0409@gmail.com 2023-08-19T05:41:32+08:00 GitHub noreply@github.com 2023-08-19T05:41:32+08:00 server : update xxd usage for older versions compatibility (#2649)
2d8b76a110d76ff6b5728ff0af8477531e4db60e 7af633aec339367e36c867ae709088d6a801aa75 Adrian smith.adriane@gmail.com 2023-08-18T12:39:22-07:00 GitHub noreply@github.com 2023-08-18T21:39:22+02:00 Add link to clojure bindings to Readme. (#2659)
7af633aec339367e36c867ae709088d6a801aa75 097e121e2f17ed3541cf02c55ff7e9febc091b19 Georgi Gerganov ggerganov@gmail.com 2023-08-18T17:48:31+03:00 GitHub noreply@github.com 2023-08-18T17:48:31+03:00 readme : incoming BREAKING CHANGE
097e121e2f17ed3541cf02c55ff7e9febc091b19 eaf98c2649d7da705de255712f0038ac7e47c610 slaren slarengh@gmail.com 2023-08-18T12:44:58+02:00 GitHub noreply@github.com 2023-08-18T12:44:58+02:00 llama : add benchmark example (#2626)
eaf98c2649d7da705de255712f0038ac7e47c610 e9b12c332ec6e215fbac4b2ef165353acbcd8319 mdrokz mohammadmunshi@gmail.com 2023-08-18T15:47:58+05:30 GitHub noreply@github.com 2023-08-18T13:17:58+03:00 readme : add link to Rust bindings (#2656)
e9b12c332ec6e215fbac4b2ef165353acbcd8319 604b8bdfa6320bbcb018eebcc1252dfede603c6b Georgi Gerganov ggerganov@gmail.com 2023-08-18T12:48:55+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-18T12:48:55+03:00 perplexity : more meaningful ETA number - 2 decimal points
604b8bdfa6320bbcb018eebcc1252dfede603c6b 10151bee2e38b5711335c4a38f6ca93b50223acf Evan Jones evan.q.jones@gmail.com 2023-08-17T19:54:44-04:00 GitHub noreply@github.com 2023-08-17T19:54:44-04:00 Fix unicode in grammars (fixes #2501) (#2553)
10151bee2e38b5711335c4a38f6ca93b50223acf 0992a7b8b18a89e29a205efb48ceb559c9a08203 staviq staviq@gmail.com 2023-08-17T23:34:01Z GitHub noreply@github.com 2023-08-18T07:34:01+08:00 server : support for saving templates in browser LocalStorage (#2486)
0992a7b8b18a89e29a205efb48ceb559c9a08203 6ddeefad9b634c5c79e6bcf046523493ff1fdf7d Johannes Gäßler johannesg@5d6.de 2023-08-17T23:57:59+02:00 GitHub noreply@github.com 2023-08-17T23:57:59+02:00 README: fix LLAMA_CUDA_MMV_Y documentation (#2647)
6ddeefad9b634c5c79e6bcf046523493ff1fdf7d 8dae7ce68437faf1fa96ec0e7687b8700956ef20 Henri Vasserman henv@hot.ee 2023-08-17T23:11:18+03:00 GitHub noreply@github.com 2023-08-17T23:11:18+03:00 [Zig] Fixing Zig build and improvements (#2554)
8dae7ce68437faf1fa96ec0e7687b8700956ef20 a73ccf1aa34de49f61bfeb7f8a679c3bfdb3abe3 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-08-17T07:29:44-06:00 GitHub noreply@github.com 2023-08-17T07:29:44-06:00 Add --cfg-negative-prompt-file option for examples (#2591)
a73ccf1aa34de49f61bfeb7f8a679c3bfdb3abe3 7cf54e1f746941279d81d485796777c01f88049c Georgi Gerganov ggerganov@gmail.com 2023-08-17T10:47:09+03:00 GitHub noreply@github.com 2023-08-17T10:47:09+03:00 llama : replace (permute + reshape + view_1d) with (view_3d) (#2538)
7cf54e1f746941279d81d485796777c01f88049c a872a2b28eaefc8d464eaa535c94deeb501666f9 drbh david.richard.holtz@gmail.com 2023-08-17T03:41:01-04:00 GitHub noreply@github.com 2023-08-17T10:41:01+03:00 tests : adds simple llama grammar tests (#2618)
a872a2b28eaefc8d464eaa535c94deeb501666f9 0919a0f73d95cfb93a1646a1d1741a0615fe2c5e Shouzheng Liu lshzh.hi@gmail.com 2023-08-17T03:35:53-04:00 GitHub noreply@github.com 2023-08-17T10:35:53+03:00 ggml-alloc : fix discrepency between measure&eval (#2639)
0919a0f73d95cfb93a1646a1d1741a0615fe2c5e ed53db86c3b0e0815331a96d7a379edb5e62472c Kolen Cheung ickc@users.noreply.github.com 2023-08-16T21:09:49+01:00 GitHub noreply@github.com 2023-08-16T23:09:49+03:00 cmake : install ggml-meta.metal if LLAMA_METAL (#2449)
ed53db86c3b0e0815331a96d7a379edb5e62472c fc8ef549e50087762a0b4f901cd74b2defcc6ae3 Jhen-Jie Hong iainst0409@gmail.com 2023-08-17T04:09:03+08:00 GitHub noreply@github.com 2023-08-16T23:09:03+03:00 metal : print error of load pipeline state (#2564)
fc8ef549e50087762a0b4f901cd74b2defcc6ae3 bf83bff6742c0f1795b4c18695a13a34ac7adf62 Shouzheng Liu lshzh.hi@gmail.com 2023-08-16T16:08:28-04:00 GitHub noreply@github.com 2023-08-16T23:08:28+03:00 metal : enable ggml-alloc (#2627)
bf83bff6742c0f1795b4c18695a13a34ac7adf62 b5ffb2849d23afe73647f68eec7b68187af09be6 Shouzheng Liu lshzh.hi@gmail.com 2023-08-16T16:07:04-04:00 GitHub noreply@github.com 2023-08-16T23:07:04+03:00 metal : matrix-matrix multiplication kernel (#2615)
b5ffb2849d23afe73647f68eec7b68187af09be6 3ebb00935f3f0522b75df49c2769ab1774b91380 Georgi Gerganov ggerganov@gmail.com 2023-08-15T10:04:58+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-15T10:05:25+03:00 scripts : add helper script to get wikitext
3ebb00935f3f0522b75df49c2769ab1774b91380 d783f7982e0e823a2626a9956359c0d36c1a7e21 Jhen-Jie Hong iainst0409@gmail.com 2023-08-15T06:14:14+08:00 GitHub noreply@github.com 2023-08-15T06:14:14+08:00 server : add missing /json-schema-to-grammar.mjs (#2616)
d783f7982e0e823a2626a9956359c0d36c1a7e21 d75561df207d22790609ee0ad924302f66ac2599 Jhen-Jie Hong iainst0409@gmail.com 2023-08-14T21:37:39+08:00 GitHub noreply@github.com 2023-08-14T16:37:39+03:00 metal : return null instead of exit(1) (#2573)
d75561df207d22790609ee0ad924302f66ac2599 348acf188c9fbe66396990f2dc83229df367969b Cheng Shao terrorjack@type.dance 2023-08-14T15:36:42+02:00 GitHub noreply@github.com 2023-08-14T16:36:42+03:00 server : add --numa support (#2524)
348acf188c9fbe66396990f2dc83229df367969b 1cd06fa25eb859b14b3427a1d815a48f25fc3c34 Kamil Tomšík info@tomsik.cz 2023-08-14T15:35:16+02:00 GitHub noreply@github.com 2023-08-14T16:35:16+03:00 llama : add missing enum keyword in function signatures (#2610)
1cd06fa25eb859b14b3427a1d815a48f25fc3c34 2feb8934eb75ca63f3c42724229cce1df9579c8e Johannes Gäßler johannesg@5d6.de 2023-08-14T10:41:22+02:00 GitHub noreply@github.com 2023-08-14T10:41:22+02:00 CUDA: launch_bounds, small q4_K, q5_K mmq refactor (#2596)
2feb8934eb75ca63f3c42724229cce1df9579c8e 5517d6e69214cdead000a76983b9fe175c3f8329 Jhen-Jie Hong iainst0409@gmail.com 2023-08-14T16:20:17+08:00 GitHub noreply@github.com 2023-08-14T16:20:17+08:00 server : fix default grammar by use empty string in the UI (#2604)
5517d6e69214cdead000a76983b9fe175c3f8329 f31b5397143009d682db90fd2a6cde83f1ef00eb Jhen-Jie Hong iainst0409@gmail.com 2023-08-14T15:16:54+08:00 GitHub noreply@github.com 2023-08-14T15:16:54+08:00 server : implement json-schema-to-grammar.mjs & add grammar param in the UI (#2588)
f31b5397143009d682db90fd2a6cde83f1ef00eb ee77efea2a1e3f7d153976b0934522b6bbaa62e6 vxiiduu 73044267+vxiiduu@users.noreply.github.com 2023-08-14T13:59:16+10:00 GitHub noreply@github.com 2023-08-13T20:59:16-07:00 Enhance Windows 7 and below compatibility. (#2592)
ee77efea2a1e3f7d153976b0934522b6bbaa62e6 f64d44a9b9581cd58f7ec40f4fa1c3ca5ca18e1e drbh david.richard.holtz@gmail.com 2023-08-13T10:00:48-04:00 GitHub noreply@github.com 2023-08-13T17:00:48+03:00 test : add simple grammar parsing tests (#2594)
f64d44a9b9581cd58f7ec40f4fa1c3ca5ca18e1e b19edd54d51cef5e3616c18b1d0d8626895b2cba Johannes Gäßler johannesg@5d6.de 2023-08-13T00:24:45+02:00 GitHub noreply@github.com 2023-08-13T00:24:45+02:00 CUDA: Fixed OpenLLaMA 3b mmq, reduced compile time (#2590)
b19edd54d51cef5e3616c18b1d0d8626895b2cba 53dc399472d5bd35ee739b865e843b1996bd3814 byte-6174 88070277+byte-6174@users.noreply.github.com 2023-08-11T19:17:25-04:00 GitHub noreply@github.com 2023-08-12T01:17:25+02:00 Adding support for llama2.c models (#2559)
53dc399472d5bd35ee739b865e843b1996bd3814 9ca4abed893685692f90413e4d43153af12342d9 Equim sayaka@ekyu.moe 2023-08-12T06:35:14+08:00 GitHub noreply@github.com 2023-08-12T00:35:14+02:00 server: fixed wrong variable name in timing json (#2579)
9ca4abed893685692f90413e4d43153af12342d9 e59fcb2bc129881f4a269fee748fb38bce0a64de DannyDaemonic DannyDaemonic@gmail.com 2023-08-10T13:11:36-07:00 GitHub noreply@github.com 2023-08-10T13:11:36-07:00 Handle `ENABLE_VIRTUAL_TERMINAL_PROCESSING` more gracefully on earlier versions of Windows.
e59fcb2bc129881f4a269fee748fb38bce0a64de 1638757767072a4957f52b9e3594f0b67610631b Christian Demsar crasm@git.vczf.us 2023-08-10T10:28:27-04:00 GitHub noreply@github.com 2023-08-10T16:28:27+02:00 Add --n-predict -2 for stopping generation on full context (#2565)
1638757767072a4957f52b9e3594f0b67610631b 916a9acdd0a411426690400ebe2bb7ce840a6bba Martin Krasser krasserm@googlemail.com 2023-08-10T12:16:38+02:00 GitHub noreply@github.com 2023-08-10T13:16:38+03:00 Fix grammar-based sampling issue in server (#2566)
916a9acdd0a411426690400ebe2bb7ce840a6bba ea04a4ca1940d92becc0ee26523aa2c4a18cf938 Sam Spilsbury smspillaz@gmail.com 2023-08-09T23:47:42+03:00 GitHub noreply@github.com 2023-08-09T22:47:42+02:00 ggml-alloc: Don't try to re-use buffers of external tensors (#2562)
ea04a4ca1940d92becc0ee26523aa2c4a18cf938 25d43e0eb578b6e73046d9d6644a3a14d460600d grahameth 96447521+grahameth@users.noreply.github.com 2023-08-09T22:46:40+02:00 GitHub noreply@github.com 2023-08-09T22:46:40+02:00 add log_callback to llama_context_params for custom logging. (#2234)
25d43e0eb578b6e73046d9d6644a3a14d460600d f5bfea0580e417f99850d5456ca541d871a3e48c Johannes Gäßler johannesg@5d6.de 2023-08-09T09:42:34+02:00 GitHub noreply@github.com 2023-08-09T09:42:34+02:00 CUDA: tuned mul_mat_q kernels (#2546)
f5bfea0580e417f99850d5456ca541d871a3e48c acfc5478ff3446ca3b54553967a3dea09b7c771a Martin Krasser krasserm@googlemail.com 2023-08-08T15:29:19+02:00 GitHub noreply@github.com 2023-08-08T16:29:19+03:00 Allow passing grammar to completion endpoint (#2532)
acfc5478ff3446ca3b54553967a3dea09b7c771a 7ed8d1fe7f8cbe6a6763e6b46759795ac8d21e12 Johannes Gäßler johannesg@5d6.de 2023-08-08T14:38:16+02:00 GitHub noreply@github.com 2023-08-08T14:38:16+02:00 CUDA: tighter VRAM scratch size for 65b/70b (#2551)
7ed8d1fe7f8cbe6a6763e6b46759795ac8d21e12 e7f94d6fdc83b41ba449b4b8c80821673dd12ffc chaihahaha chai836275709@gmail.com 2023-08-08T20:07:02+08:00 GitHub noreply@github.com 2023-08-08T15:07:02+03:00 llm.vim : multiline autocompletion, get rid of "^@" (#2543)
e7f94d6fdc83b41ba449b4b8c80821673dd12ffc 2d7baaf50f3277e65cf71071f61ea34823d14c30 Georgi Gerganov ggerganov@gmail.com 2023-08-08T15:05:30+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-08T15:06:18+03:00 vim : bring back simple llm.vim example
2d7baaf50f3277e65cf71071f61ea34823d14c30 f3c3b4b1672d860800639c87d3b5d17564692469 AustinMroz austinmroz@utexas.edu 2023-08-08T06:44:48-05:00 GitHub noreply@github.com 2023-08-08T14:44:48+03:00 vim : streaming and more (#2495)
f3c3b4b1672d860800639c87d3b5d17564692469 93356bdb7a324a8f6570f99d02af392cd4c45796 klosax 131523366+klosax@users.noreply.github.com 2023-08-07T19:07:19+02:00 GitHub noreply@github.com 2023-08-07T19:07:19+02:00 Add --rope-scale parameter (#2544)
93356bdb7a324a8f6570f99d02af392cd4c45796 60baff7c8584ec369e53469cad5f92e102b1efe4 Georgi Gerganov ggerganov@gmail.com 2023-08-07T14:25:58+03:00 GitHub noreply@github.com 2023-08-07T14:25:58+03:00 ggml : mul mat tweaks (#2372)
60baff7c8584ec369e53469cad5f92e102b1efe4 9082b5dfbfae01243a0b822dcd2812877e63bf1b Georgi Gerganov ggerganov@gmail.com 2023-08-07T14:24:42+03:00 Georgi Gerganov ggerganov@gmail.com 2023-08-07T14:24:42+03:00 ggml : pad result of ggml_nbytes()
9082b5dfbfae01243a0b822dcd2812877e63bf1b 99d29c0094476c4962023036ecd61a3309d0e16b Georgi Gerganov ggerganov@gmail.com 2023-08-07T13:55:18+03:00 GitHub noreply@github.com 2023-08-07T13:55:18+03:00 ggml : change params pointer (style change) (#2539)
99d29c0094476c4962023036ecd61a3309d0e16b 3d9a55181603e85a26378a850a14068034e5002d Georgi Gerganov ggerganov@gmail.com 2023-08-07T13:20:09+03:00 GitHub noreply@github.com 2023-08-07T13:20:09+03:00 ggml : sync (custom ops) (#2537)
3d9a55181603e85a26378a850a14068034e5002d f6f9896ac3d2ff207e18f87dab85d126ceef5236 Johannes Gäßler johannesg@5d6.de 2023-08-07T10:09:40+02:00 GitHub noreply@github.com 2023-08-07T10:09:40+02:00 Fixed mmap prefetch for GPU offloading (#2529)
f6f9896ac3d2ff207e18f87dab85d126ceef5236 34a14b28ff7f3c98730339bacee035091b2a812a Georgi Gerganov ggerganov@gmail.com 2023-08-07T10:52:57+03:00 GitHub noreply@github.com 2023-08-07T10:52:57+03:00 metal : fix out-of-bounds access + inc concurrency nodes (#2416)
34a14b28ff7f3c98730339bacee035091b2a812a 7297128db8159c7b12db4c28a4532b993025c2e5 GiviMAD GiviMAD@users.noreply.github.com 2023-08-06T23:21:46-07:00 GitHub noreply@github.com 2023-08-07T09:21:46+03:00 [Makefile] Move ARM CFLAGS before compilation (#2536)
7297128db8159c7b12db4c28a4532b993025c2e5 86c32198954a2bc482025703d6875e11f1c2a574 Henri Vasserman henv@hot.ee 2023-08-07T08:35:53+03:00 GitHub noreply@github.com 2023-08-07T08:35:53+03:00 [Zig] Rewrite build for Zig 0.11 (#2514)
86c32198954a2bc482025703d6875e11f1c2a574 2e8265ae1764d6288aab0e2df641909072e2d58e DannyDaemonic DannyDaemonic@gmail.com 2023-08-05T23:49:34-07:00 GitHub noreply@github.com 2023-08-06T09:49:34+03:00 console : fix issue related to Windows 11 PowerShell console mode persistence (#2521)
2e8265ae1764d6288aab0e2df641909072e2d58e f514d1b306e1114c2884fcb25dd9bd48ae64ba32 Keiichi Tabata keiichi.tabata@outlook.com 2023-08-06T15:34:05+09:00 GitHub noreply@github.com 2023-08-06T09:34:05+03:00 convert.py : add missing abstract methods for quantized data (#2491)
f514d1b306e1114c2884fcb25dd9bd48ae64ba32 332311234a0aa2974b2450710e22e09d90dd6b0b Johannes Gäßler johannesg@5d6.de 2023-08-05T18:20:44+02:00 GitHub noreply@github.com 2023-08-05T18:20:44+02:00 CUDA: faster k-quant mul_mat_q kernels (#2525)
332311234a0aa2974b2450710e22e09d90dd6b0b 182af739c4ce237f7579facfe8f94dc53a1f573f Jonas Wunderlich 32615971+jonas-w@users.noreply.github.com 2023-08-04T20:16:11Z GitHub noreply@github.com 2023-08-04T22:16:11+02:00 fix firefox autoscroll (#2519)
182af739c4ce237f7579facfe8f94dc53a1f573f 4329d1acb01c353803a54733b8eef9d93d0b84b2 Cebtenzzre cebtenzzre@gmail.com 2023-08-04T15:00:57-04:00 GitHub noreply@github.com 2023-08-04T21:00:57+02:00 server: regenerate completion.js.hpp (#2515)
4329d1acb01c353803a54733b8eef9d93d0b84b2 02f9d96a866268700b8d8e7acbbcb4392c5ff345 Cebtenzzre cebtenzzre@gmail.com 2023-08-04T11:35:22-04:00 GitHub noreply@github.com 2023-08-04T17:35:22+02:00 CUDA: use min compute capability of GPUs actually used (#2506)
02f9d96a866268700b8d8e7acbbcb4392c5ff345 3498588e0fb4daf040c4e3c698595cb0bfd345c0 Cebtenzzre cebtenzzre@gmail.com 2023-08-04T11:34:32-04:00 GitHub noreply@github.com 2023-08-04T17:34:32+02:00 CUDA: check if event is NULL before cudaStreamWaitEvent (#2505)
3498588e0fb4daf040c4e3c698595cb0bfd345c0 5f631c26794b6371fcf2660e8d0c53494a5575f7 DannyDaemonic DannyDaemonic@gmail.com 2023-08-04T08:20:12-07:00 GitHub noreply@github.com 2023-08-04T08:20:12-07:00 Add --simple-io option for subprocesses and break out console.h and cpp (#1558)
5f631c26794b6371fcf2660e8d0c53494a5575f7 415e99fec27be5a2e4283f1937afd17eb33fbd66 Stephen Nichols snichols@users.noreply.github.com 2023-08-04T06:37:24-05:00 GitHub noreply@github.com 2023-08-04T13:37:24+02:00 Fixing race condition in server and partial stream handling in frontend. (#2391)
415e99fec27be5a2e4283f1937afd17eb33fbd66 ff966e7ca6af127c9405523cdb07ef8fa01bf6d6 l3utterfly gc.pthzfoldr@gmail.com 2023-08-04T19:29:52+08:00 GitHub noreply@github.com 2023-08-04T13:29:52+02:00 Stream save llama context data to file instead of allocating entire buffer upfront (#2488)
ff966e7ca6af127c9405523cdb07ef8fa01bf6d6 8183159cf3def112f6d1fe94815fce70e1bffa12 Borislav Stanimirov b.stanimirov@abv.bg 2023-08-04T13:07:21+03:00 GitHub noreply@github.com 2023-08-04T13:07:21+03:00 build : fix several cast and printf warnings (#2499)
8183159cf3def112f6d1fe94815fce70e1bffa12 468ea24fb4633a0d681f7ac84089566c1c6190cb Evan Jones evan.q.jones@gmail.com 2023-08-02T22:05:44-04:00 GitHub noreply@github.com 2023-08-02T22:05:44-04:00 examples : generate JSON according to schema (#1887)
468ea24fb4633a0d681f7ac84089566c1c6190cb 4f6b60c77652cdfc9d5545fe247ae5d764815598 Johannes Gäßler johannesg@5d6.de 2023-08-02T18:04:04+02:00 GitHub noreply@github.com 2023-08-02T18:04:04+02:00 CUDA: faster non k-quant mul_mat_q kernels (#2483)
4f6b60c77652cdfc9d5545fe247ae5d764815598 220d9318647a8ce127dbf7c9de5400455f41e7d8 Johannes Gäßler johannesg@5d6.de 2023-08-02T16:48:10+02:00 GitHub noreply@github.com 2023-08-02T16:48:10+02:00 CUDA: Fix models with output size != 32000 (#2480)
220d9318647a8ce127dbf7c9de5400455f41e7d8 81844fbcfd93a162b7aeaea9e4f2ab1358f7f97e ldwang ftgreat@163.com 2023-08-02T16:21:11+08:00 GitHub noreply@github.com 2023-08-02T11:21:11+03:00 readme : add Aquila-7B model series to supported models (#2487)
81844fbcfd93a162b7aeaea9e4f2ab1358f7f97e a312193e184b919047f33a5e844d846c5538dbe6 Eve 139727413+netrunnereve@users.noreply.github.com 2023-08-02T04:06:19-04:00 GitHub noreply@github.com 2023-08-02T11:06:19+03:00 tests : Fix compilation warnings (Linux/GCC) (#2451)
a312193e184b919047f33a5e844d846c5538dbe6 c574bddb368424b5996cbee2ec45ec050967d404 Yiming Cui conandiy@vip.qq.com 2023-08-02T14:18:31+08:00 GitHub noreply@github.com 2023-08-02T09:18:31+03:00 readme : Add Chinese LLaMA-2 / Alpaca-2 to supported models (#2475)
c574bddb368424b5996cbee2ec45ec050967d404 86aeb27734481751592abd85590020b40d9224c8 Bono Lv lvscar@users.noreply.github.com 2023-08-01T20:54:28+08:00 GitHub noreply@github.com 2023-08-01T14:54:28+02:00 fix a typo in examples/server/README.md (#2478)
86aeb27734481751592abd85590020b40d9224c8 1873ff586bd8499a18f763632711bf15d253585e ebraminio ebraminio@gmail.com 2023-08-01T01:56:23-07:00 GitHub noreply@github.com 2023-08-01T10:56:23+02:00 server : Support dark mode (#2414)
1873ff586bd8499a18f763632711bf15d253585e 49e7cb5bb1f75c91dd5db7d2d88cbc11bd9ee0c5 Matteo Boschini 12133566+mbosc@users.noreply.github.com 2023-08-01T09:43:12+02:00 GitHub noreply@github.com 2023-08-01T10:43:12+03:00 metal : add gqa8 kernel to allow llama-2-70B on metal (#2459)
49e7cb5bb1f75c91dd5db7d2d88cbc11bd9ee0c5 b772bba42e3bbca3cdab224456f8ff2ce427fd0b Johannes Gäßler johannesg@5d6.de 2023-07-31T21:02:19+02:00 GitHub noreply@github.com 2023-07-31T21:02:19+02:00 CUDA: fixed LLAMA_FAST compilation option (#2473)
b772bba42e3bbca3cdab224456f8ff2ce427fd0b 0728c5a8b9569183ffca0399caac099afef87595 Johannes Gäßler johannesg@5d6.de 2023-07-31T19:52:22+02:00 GitHub noreply@github.com 2023-07-31T19:52:22+02:00 CUDA: fixed cmake F16 option (#2471)
0728c5a8b9569183ffca0399caac099afef87595 1215ed7d5ccf854a55eccb52661427bb985e7f2c Johannes Gäßler johannesg@5d6.de 2023-07-31T15:44:35+02:00 GitHub noreply@github.com 2023-07-31T15:44:35+02:00 CUDA: mmq CLI option, fixed mmq build issues (#2453)
1215ed7d5ccf854a55eccb52661427bb985e7f2c 2dbf518911926ef5a30f43aa83a0b1b1cdeaaa11 Johannes Gäßler johannesg@5d6.de 2023-07-31T14:32:30+02:00 GitHub noreply@github.com 2023-07-31T14:32:30+02:00 CUDA: Implemented row flattening for non-glm RoPE (#2468)
2dbf518911926ef5a30f43aa83a0b1b1cdeaaa11 9d2382b3e45b5815fc6a054045a2f2c2b18c22a2 Johannes Gäßler johannesg@5d6.de 2023-07-31T13:18:51+02:00 GitHub noreply@github.com 2023-07-31T13:18:51+02:00 CUDA: fewer memory bank conflicts for mul_mat_q (#2458)
9d2382b3e45b5815fc6a054045a2f2c2b18c22a2 a113689571420fb4d6540f1a324d12965781356a slaren slarengh@gmail.com 2023-07-31T11:02:53+02:00 GitHub noreply@github.com 2023-07-31T11:02:53+02:00 Fix Metal backend broken from the allocator changes (#2455)
a113689571420fb4d6540f1a324d12965781356a 11f3ca06b8c66b0427aab0a472479da22553b472 slaren slarengh@gmail.com 2023-07-30T15:58:01+02:00 GitHub noreply@github.com 2023-07-30T15:58:01+02:00 ggml : add graph tensor allocator (#2411)
11f3ca06b8c66b0427aab0a472479da22553b472 9baf9ef304f330009d5a93b7390280a0fd27c9a1 Johannes Gäßler johannesg@5d6.de 2023-07-29T23:04:44+02:00 GitHub noreply@github.com 2023-07-29T23:04:44+02:00 CUDA: Quantized matrix matrix multiplication (#2160)
9baf9ef304f330009d5a93b7390280a0fd27c9a1 8a88e5855c3b93024be0f93290b01a4206b65b38 Johannes Gäßler johannesg@5d6.de 2023-07-29T23:04:10+02:00 GitHub noreply@github.com 2023-07-29T23:04:10+02:00 CUDA: faster multi GPU synchronization (#2448)
8a88e5855c3b93024be0f93290b01a4206b65b38 a9559bf77b903d94eb21614ceae5ae8950f0f1fc klosax 131523366+klosax@users.noreply.github.com 2023-07-28T20:25:36+02:00 GitHub noreply@github.com 2023-07-28T21:25:36+03:00 perplexity : add Hellaswag calculation (#2389)
a9559bf77b903d94eb21614ceae5ae8950f0f1fc ee1b497c985f61d6ec519c39fcfed78a3c6f1d06 Lee 44310445+lx200916@users.noreply.github.com 2023-07-29T02:17:45+08:00 GitHub noreply@github.com 2023-07-28T21:17:45+03:00 ggml : workaround for missing _mm256_setr_m128i in GCC < 8 in k_quants.c (#2405)
ee1b497c985f61d6ec519c39fcfed78a3c6f1d06 d73b8d48b45d6e2c0ae9bb8c39623c4024adc275 eric8607242 e0928021388@gmail.com 2023-07-29T02:10:05+08:00 GitHub noreply@github.com 2023-07-28T21:10:05+03:00 llama : support more diverse tokenizers? (#2420)
d73b8d48b45d6e2c0ae9bb8c39623c4024adc275 34ae1caf7fdea4c4c09087002e15e6230102e6a9 Georgi Gerganov ggerganov@gmail.com 2023-07-28T21:05:08+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-28T21:05:08+03:00 examples : fix whitespace
34ae1caf7fdea4c4c09087002e15e6230102e6a9 d91f3f0c55663719ea03b76311e8c36ed55eb0e2 nhamanasu 45545786+nhamanasu@users.noreply.github.com 2023-07-29T03:02:10+09:00 GitHub noreply@github.com 2023-07-28T21:02:10+03:00 examples : server chat mode with llama2 (#2400)
d91f3f0c55663719ea03b76311e8c36ed55eb0e2 65cdf34bdc469fa86248e667a5880992684ef114 Weird Constructor weirdconstructor@gmail.com 2023-07-28T10:44:43+02:00 GitHub noreply@github.com 2023-07-28T11:44:43+03:00 readme : fix the description of the Tail free sampling (TFS) method (#2431)
65cdf34bdc469fa86248e667a5880992684ef114 edcc7ae7d26007bbf83136e9d33f863fcad9b871 Rand Xie randxiexyy29@gmail.com 2023-07-28T01:42:53-07:00 GitHub noreply@github.com 2023-07-28T11:42:53+03:00 llama : use n_embd_gqa instead of n_embd to handle llama-2 70B (#2433)
edcc7ae7d26007bbf83136e9d33f863fcad9b871 7c529cede6e84054e77a3eceab31c53de7b2f55b niansa/tuxifan tuxifan@posteo.de 2023-07-28T03:14:11+02:00 GitHub noreply@github.com 2023-07-28T03:14:11+02:00 Obtaining LLaMA 2 instructions (#2308)
7c529cede6e84054e77a3eceab31c53de7b2f55b 1a941869cbef8e9cc351a6c6987e4ae3b0f021f7 mj-shifu 77107165+mj-shifu@users.noreply.github.com 2023-07-27T22:39:17+02:00 GitHub noreply@github.com 2023-07-27T14:39:17-06:00 convert.py : Update to support 70B HF format model files (#2427)
1a941869cbef8e9cc351a6c6987e4ae3b0f021f7 b5472ea0ada081a6e1c06998ebbc9a24aa2cd4a4 Georgi Gerganov ggerganov@gmail.com 2023-07-27T11:00:54+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-27T11:00:54+03:00 metal : disable graph concurrency optimization due to bug (#2413)
b5472ea0ada081a6e1c06998ebbc9a24aa2cd4a4 6df1f5940f889adde32fe47dc8881f010dcf9aba slaren slarengh@gmail.com 2023-07-26T23:57:23+02:00 GitHub noreply@github.com 2023-07-26T23:57:23+02:00 ggml : fix assert in ggml_set_unary_op (#2410)
6df1f5940f889adde32fe47dc8881f010dcf9aba 5488fb789ea5692268309baa76f67598155060be Cebtenzzre cebtenzzre@gmail.com 2023-07-26T14:00:04-04:00 GitHub noreply@github.com 2023-07-26T21:00:04+03:00 make : build with -Wmissing-prototypes (#2394)
5488fb789ea5692268309baa76f67598155060be eb542d39324574a6778fad9ba9e34ba7a14a82a3 slaren slarengh@gmail.com 2023-07-26T15:56:53+02:00 GitHub noreply@github.com 2023-07-26T15:56:53+02:00 ggml : allocate graphs in a context (#2392)
eb542d39324574a6778fad9ba9e34ba7a14a82a3 07aaa0f63fccaeab099b3a732abda20b921bc5a5 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-25T18:35:53+03:00 GitHub noreply@github.com 2023-07-25T18:35:53+03:00 Add LLAMA_DEFAULT_RMS_EPS so we can change the default (#2384)
07aaa0f63fccaeab099b3a732abda20b921bc5a5 fce48caf9a6b9930eee9e2a5971428cdff403ba8 slaren slarengh@gmail.com 2023-07-25T16:20:12+02:00 GitHub noreply@github.com 2023-07-25T16:20:12+02:00 ggml : fix ggml_flash_attn to use op_params (#2387)
fce48caf9a6b9930eee9e2a5971428cdff403ba8 875086bdb95ce1f3294439811536533199e3b579 ldwang ftgreat@163.com 2023-07-25T21:22:09+08:00 GitHub noreply@github.com 2023-07-25T16:22:09+03:00 convert.py : support bpe tokenizer (#2228)
875086bdb95ce1f3294439811536533199e3b579 da1889834a036a63ead2b0ca5c9ed8967712568c Jiahao Li liplus17@163.com 2023-07-25T20:58:32+08:00 GitHub noreply@github.com 2023-07-25T15:58:32+03:00 ggml : relax contiguous constraints in activation function (#2371)
da1889834a036a63ead2b0ca5c9ed8967712568c 82552b7f5403ca13957ac9a2cdc1732470057b62 slaren slarengh@gmail.com 2023-07-25T14:32:20+02:00 GitHub noreply@github.com 2023-07-25T15:32:20+03:00 ggml : improve graph build time via hash table lookup (#2329)
82552b7f5403ca13957ac9a2cdc1732470057b62 0c06204fb39aa5560e883e0ae74be9518c57d88e Hesen Peng hesen.peng@gmail.com 2023-07-25T05:24:09-07:00 GitHub noreply@github.com 2023-07-25T15:24:09+03:00 build : fix line breaking error in build-info.sh (#2349)
0c06204fb39aa5560e883e0ae74be9518c57d88e 1fed755b1fb9babb6dbc1b4023e492950cd5a5be Xiao-Yong Jin jinxiaoyong@gmail.com 2023-07-25T07:19:11-05:00 GitHub noreply@github.com 2023-07-25T15:19:11+03:00 main : add `--in-prefix-bos` to prefix BOS to user inputs; keep EOS (#2304)
1fed755b1fb9babb6dbc1b4023e492950cd5a5be be2301bcdaf6c9f0921141bd071de7788e2a351a Eve 139727413+netrunnereve@users.noreply.github.com 2023-07-25T08:16:13-04:00 GitHub noreply@github.com 2023-07-25T15:16:13+03:00 ci : add non-AVX scalar build/test (#2356)
be2301bcdaf6c9f0921141bd071de7788e2a351a 1aa18ef994a6a2b531434eb13251ef48e56d345b katsu560 118887472+katsu560@users.noreply.github.com 2023-07-25T21:13:41+09:00 GitHub noreply@github.com 2023-07-25T15:13:41+03:00 k_quants : add AVX support to dot functions with QK_K as 64 (#2339)
1aa18ef994a6a2b531434eb13251ef48e56d345b 9a08eaf3c4010962d0126e9e5bfbe9af64b2ac90 Shouzheng Liu lshzh.hi@gmail.com 2023-07-25T08:00:19-04:00 GitHub noreply@github.com 2023-07-25T15:00:19+03:00 metal : concurrently dispatch commands (#2358)
9a08eaf3c4010962d0126e9e5bfbe9af64b2ac90 129d844c87d90e74aafc23dcc84c980fd408def4 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-25T13:48:29+03:00 GitHub noreply@github.com 2023-07-25T13:48:29+03:00 Another speed gain for Q4_0 and Q4_1 on Metal (#2375)
129d844c87d90e74aafc23dcc84c980fd408def4 d5512b782b27ff698007dcd175da18959d5f163f Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-25T13:48:04+03:00 GitHub noreply@github.com 2023-07-25T13:48:04+03:00 Fix Q4_K and Q5_K for QK_K = 64 on CUDA (#2359)
d5512b782b27ff698007dcd175da18959d5f163f c798308e3a425eae050a1f249a576fa8c6433327 slaren slarengh@gmail.com 2023-07-25T11:36:17+02:00 GitHub noreply@github.com 2023-07-25T12:36:17+03:00 server: add rms_norm_eps parameter (#2380)
c798308e3a425eae050a1f249a576fa8c6433327 41c674161fb2459bdf7806d1eebead15bc5d046e Henri Vasserman henv@hot.ee 2023-07-25T10:27:34+03:00 GitHub noreply@github.com 2023-07-25T10:27:34+03:00 [Server] Escape HTML in webchat (#2368)
41c674161fb2459bdf7806d1eebead15bc5d046e b3f138d05849ccbce67303ac17b50ebbc268128a slaren slarengh@gmail.com 2023-07-24T17:57:12+02:00 GitHub noreply@github.com 2023-07-24T17:57:12+02:00 make rms_norm_eps a parameter (#2374)
b3f138d05849ccbce67303ac17b50ebbc268128a 5b2b2dc6ae8086bff7c9b3c17fb435cf319b7185 Aarni Koskela akx@iki.fi 2023-07-24T17:54:22+03:00 GitHub noreply@github.com 2023-07-24T17:54:22+03:00 Chat UI extras (#2366)
5b2b2dc6ae8086bff7c9b3c17fb435cf319b7185 42f70cb2f6a8089e0a0560a459e4ba317bac4d49 Georgi Gerganov ggerganov@gmail.com 2023-07-24T14:46:21+03:00 GitHub noreply@github.com 2023-07-24T14:46:21+03:00 ggml : sync (unary ops refactor, static-correctness) (#2370)
42f70cb2f6a8089e0a0560a459e4ba317bac4d49 84e09a7d8bc4ab6d658b5cd81295ac0add60be78 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-24T12:55:02+03:00 GitHub noreply@github.com 2023-07-24T12:55:02+03:00 Fix scalar version of Q5_K when QK_K = 64 (#2362)
84e09a7d8bc4ab6d658b5cd81295ac0add60be78 2f9cf974a066ac0e03fbb235d834b01b0164d743 Evan Jones evan.q.jones@gmail.com 2023-07-23T23:58:10-04:00 GitHub noreply@github.com 2023-07-23T23:58:10-04:00 llama : add grammar-based sampling (#1773)
2f9cf974a066ac0e03fbb235d834b01b0164d743 4f06592cc6b83979e4b442e8cb97b3948c857188 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-24T00:19:47+03:00 GitHub noreply@github.com 2023-07-24T00:19:47+03:00 Some more Q4_K and Q5_K speedup on CUDA (#2346)
4f06592cc6b83979e4b442e8cb97b3948c857188 70d26ac3883009946e737525506fa88f52727132 IgnacioFDM ignaciofdm@gmail.com 2023-07-23T17:31:17-03:00 GitHub noreply@github.com 2023-07-23T23:31:17+03:00 Add gqa parameter support to the server (#2351)
70d26ac3883009946e737525506fa88f52727132 57921ca6db53e08eb90010fba99add85be28b5a1 Johannes Gäßler johannesg@5d6.de 2023-07-23T17:49:06+02:00 GitHub noreply@github.com 2023-07-23T17:49:06+02:00 Fix __dp4a documentation (#2348)
57921ca6db53e08eb90010fba99add85be28b5a1 3602ac4255fd4cd589821346290b464a64b955d1 wzy 32936898+Freed-Wu@users.noreply.github.com 2023-07-23T21:33:02+08:00 GitHub noreply@github.com 2023-07-23T16:33:02+03:00 common : n_threads == -1 uses std::thread::hardware_concurrency() (#2347)
3602ac4255fd4cd589821346290b464a64b955d1 95a6c595e7ca8dbe47ccf8824e04213e10357f9a slaren slarengh@gmail.com 2023-07-23T15:19:39+02:00 GitHub noreply@github.com 2023-07-23T15:19:39+02:00 fix n_tasks (#2342)
95a6c595e7ca8dbe47ccf8824e04213e10357f9a e76d630df17e235e6b9ef416c45996765d2e36fb slaren slarengh@gmail.com 2023-07-23T14:36:02+02:00 GitHub noreply@github.com 2023-07-23T14:36:02+02:00 ggml: move op parameters from tensors to ggml_tensor::op_params (#2333)
e76d630df17e235e6b9ef416c45996765d2e36fb 1d0824b2476e7fda09751a0235c9e571b76d6f2c Georgi Gerganov ggerganov@gmail.com 2023-07-23T15:09:47+03:00 GitHub noreply@github.com 2023-07-23T15:09:47+03:00 llama : grouped-query attention + LLaMAv2 70B support (#2276)
1d0824b2476e7fda09751a0235c9e571b76d6f2c bc3ec2cdc9ea20b0faba2e1b4576fab3a911e4d1 maddes8cht 55592906+maddes8cht@users.noreply.github.com 2023-07-23T13:59:48+02:00 GitHub noreply@github.com 2023-07-23T14:59:48+03:00 llama : print help to stdout (#2338)
bc3ec2cdc9ea20b0faba2e1b4576fab3a911e4d1 a940458e4814e87bd0d3fbdb3f3d2733b4a3ccb1 wzy 32936898+Freed-Wu@users.noreply.github.com 2023-07-23T19:57:02+08:00 GitHub noreply@github.com 2023-07-23T14:57:02+03:00 flake : support `nix build '.#opencl'` (#2337)
a940458e4814e87bd0d3fbdb3f3d2733b4a3ccb1 91171b8072f6f0c8ae3a61e23451acb538bb9ece Christian Demsar christian@github.email.demsar.us 2023-07-23T07:56:34-04:00 GitHub noreply@github.com 2023-07-23T14:56:34+03:00 llama : print max tensor size to stderr (#2336)
91171b8072f6f0c8ae3a61e23451acb538bb9ece 355c80f49e32b0b15c0a457f3bad380e57f5b9ac Jose Maldonado 63384398+yukiteruamano@users.noreply.github.com 2023-07-23T07:52:08-04:00 GitHub noreply@github.com 2023-07-23T14:52:08+03:00 make : fix CLBLAST compile support in FreeBSD (#2331)
355c80f49e32b0b15c0a457f3bad380e57f5b9ac 83a00ce69bef9124c0702424a012ea799128b77d AustinMroz austinmroz@utexas.edu 2023-07-23T06:16:48-05:00 GitHub noreply@github.com 2023-07-23T14:16:48+03:00 examples : simplify vim plugin (#2327)
83a00ce69bef9124c0702424a012ea799128b77d d2a43664f93ba30a84e42713bb69f936cbdacf2a Jiahao Li liplus17@163.com 2023-07-23T19:00:37+08:00 GitHub noreply@github.com 2023-07-23T14:00:37+03:00 metal : support bcast add & dup & cont op (#2323)
d2a43664f93ba30a84e42713bb69f936cbdacf2a b9b7d94fc10a8039befd1bc3af4f4b09c620c351 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-23T08:49:20+03:00 GitHub noreply@github.com 2023-07-23T08:49:20+03:00 Speed up Q4_K (#2322)
b9b7d94fc10a8039befd1bc3af4f4b09c620c351 b47b8a9cfeb439d271bf997fb985fd6d82b3af5e Johannes Gäßler johannesg@5d6.de 2023-07-22T21:27:34+02:00 GitHub noreply@github.com 2023-07-22T21:27:34+02:00 CUDA: Fixed 7b q3_K_S with mul_mat_vec_q (#2313)
b47b8a9cfeb439d271bf997fb985fd6d82b3af5e b5fe67f8c69113bd9354bc1adcfe2df6be323740 Georgi Gerganov ggerganov@gmail.com 2023-07-22T21:17:57+03:00 GitHub noreply@github.com 2023-07-22T21:17:57+03:00 llama : optimize memory buffers (#2325)
b5fe67f8c69113bd9354bc1adcfe2df6be323740 24baa54ac1ff3d4156a2360deb1473af04a9b1a2 klosax 131523366+klosax@users.noreply.github.com 2023-07-22T14:21:24+02:00 GitHub noreply@github.com 2023-07-22T14:21:24+02:00 Perplexity: Compute scores correlated to HellaSwag (#2312)
24baa54ac1ff3d4156a2360deb1473af04a9b1a2 dd6c67d3cbb7b360747f776412bf01976aa32f4b whoreson 139810751+whoreson@users.noreply.github.com 2023-07-22T12:34:51+02:00 GitHub noreply@github.com 2023-07-22T13:34:51+03:00 examples : basic VIM plugin
dd6c67d3cbb7b360747f776412bf01976aa32f4b 5d500e8ccf5eee3de3ae66685cc3be75e43e08b9 Georgi Gerganov ggerganov@gmail.com 2023-07-22T12:00:56+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-22T12:00:56+03:00 ci : fix args
5d500e8ccf5eee3de3ae66685cc3be75e43e08b9 7d5f18468ceabd7a38f414f9f21b26b0c137f994 Georgi Gerganov ggerganov@gmail.com 2023-07-22T11:48:22+03:00 GitHub noreply@github.com 2023-07-22T11:48:22+03:00 ci : add 7B CUDA tests (#2319)
7d5f18468ceabd7a38f414f9f21b26b0c137f994 d924522a46c5ef097af4a88087d91673e8e87e4d Richard Roberson richardr1126@gmail.com 2023-07-21T13:01:10-06:00 GitHub noreply@github.com 2023-07-21T22:01:10+03:00 examples : add easy python script to create quantized (k-bit support) GGML models from local HF Transformer models (#2311)
d924522a46c5ef097af4a88087d91673e8e87e4d 4d76a5f49b9b5382dba5d13d92edb9159536c225 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-21T17:27:51+03:00 GitHub noreply@github.com 2023-07-21T17:27:51+03:00 Custom RoPE + bettter memory management for CUDA (#2295)
4d76a5f49b9b5382dba5d13d92edb9159536c225 0db14fef06836caaa13cc123c0a24dc598bdb9f0 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-21T17:05:30+03:00 GitHub noreply@github.com 2023-07-21T17:05:30+03:00 Faster Q3_K implementation on Metal (#2307)
0db14fef06836caaa13cc123c0a24dc598bdb9f0 03e566977b277937c5f706180171c5d12b597b0b Georgi Gerganov ggerganov@gmail.com 2023-07-21T15:16:55+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-21T15:16:55+03:00 ggml : fix the rope fix (513f8619535a64fa9ace808cdcbcf66211535f5c)
03e566977b277937c5f706180171c5d12b597b0b 513f8619535a64fa9ace808cdcbcf66211535f5c Ikko Eltociear Ashimine eltociear@gmail.com 2023-07-21T20:53:07+09:00 GitHub noreply@github.com 2023-07-21T14:53:07+03:00 examples : fix typo in minigpt4.py (#2298)
513f8619535a64fa9ace808cdcbcf66211535f5c 3973b25a64a37a47eac156a3fd28f83c16f14bf2 Georgi Gerganov ggerganov@gmail.com 2023-07-21T14:51:34+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-21T14:51:34+03:00 ggml : fix rope args order + assert (#2054)
3973b25a64a37a47eac156a3fd28f83c16f14bf2 ab0e26bdfb7b3adb1e3145c61a0fa92d1abd21d0 Georgi Gerganov ggerganov@gmail.com 2023-07-21T14:42:41+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-21T14:42:41+03:00 gitignore : fix final newline
ab0e26bdfb7b3adb1e3145c61a0fa92d1abd21d0 73643f5fb1136dc2b65ae910bdc5a431520d70a2 Guillaume "Vermeille" Sanchez Guillaume.V.Sanchez@gmail.com 2023-07-21T12:58:36+02:00 GitHub noreply@github.com 2023-07-21T13:58:36+03:00 llama : remove cfg smooth factor as it is only a reparameterization of the guidance scale (#2280)
73643f5fb1136dc2b65ae910bdc5a431520d70a2 a814d04f81121e0429b39a61fe4afd946cd42046 Jose Maldonado 63384398+yukiteruamano@users.noreply.github.com 2023-07-21T06:53:27-04:00 GitHub noreply@github.com 2023-07-21T13:53:27+03:00 gitignore : changes for Poetry users + chat examples (#2284)
a814d04f81121e0429b39a61fe4afd946cd42046 4c013bb7385a0e52ce721480c40c45bec5ef103f Georgi Gerganov ggerganov@gmail.com 2023-07-21T13:50:55+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-21T13:50:55+03:00 make : fix indentation
4c013bb7385a0e52ce721480c40c45bec5ef103f 42c7c2e2e9cae79330f57456fbc0eae1eaff17fa Georgi Gerganov ggerganov@gmail.com 2023-07-21T13:48:18+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-21T13:49:18+03:00 ci : fix MNT realpath usage (#2250)
42c7c2e2e9cae79330f57456fbc0eae1eaff17fa 78a3d13424b01c3f8ea94ea7e59650ab0501e902 Sky Yan skyan83@gmail.com 2023-07-21T18:38:57+08:00 GitHub noreply@github.com 2023-07-21T13:38:57+03:00 make : support customized LLAMA_CUDA_NVCC and LLAMA_CUDA_CCBIN (#2275)
78a3d13424b01c3f8ea94ea7e59650ab0501e902 ae178ab46bfd6ecb2422da5dad441a4e2fef8b7e wzy 32936898+Freed-Wu@users.noreply.github.com 2023-07-21T18:26:34+08:00 GitHub noreply@github.com 2023-07-21T13:26:34+03:00 flake : remove intel mkl from flake.nix due to missing files (#2277)
ae178ab46bfd6ecb2422da5dad441a4e2fef8b7e 54e3bc76fed914f8d4a30a7a50c19867cccb1338 Georgi Gerganov ggerganov@gmail.com 2023-07-21T13:10:51+03:00 GitHub noreply@github.com 2023-07-21T13:10:51+03:00 llama : make tensor_split ptr instead of array (#2272)
54e3bc76fed914f8d4a30a7a50c19867cccb1338 019fe257bbf699f400231683a8b816ad90281275 Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2023-07-21T12:09:16+02:00 GitHub noreply@github.com 2023-07-21T13:09:16+03:00 make : add new target for test binaries (#2244)
019fe257bbf699f400231683a8b816ad90281275 e68c96f7fee8fc22814a4a1209ffc97bbf35f7bd Hatsune Miku 129688334+at8u@users.noreply.github.com 2023-07-21T08:13:18Z GitHub noreply@github.com 2023-07-21T11:13:18+03:00 MIKU MAYHEM: Upgrading the Default Model for Maximum Fun 🎉 (#2287)
e68c96f7fee8fc22814a4a1209ffc97bbf35f7bd 9cf022a1889e50113fd348dc96b4557fc75a6296 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-21T10:44:40+03:00 GitHub noreply@github.com 2023-07-21T10:44:40+03:00 Faster Q2_K on Metal (#2297)
9cf022a1889e50113fd348dc96b4557fc75a6296 e782c9e735f93ab4767ffc37462c523b73a17ddc Przemysław Pawełczyk przemoc@gmail.com 2023-07-21T09:42:21+02:00 GitHub noreply@github.com 2023-07-21T10:42:21+03:00 make : fix embdinput library and server examples building on MSYS2 (#2235)
e782c9e735f93ab4767ffc37462c523b73a17ddc 785829dfe8baf0213f2ff66963d28c62f92d7930 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-20T18:19:45+03:00 GitHub noreply@github.com 2023-07-20T18:19:45+03:00 Faster Q5_K and Q6_K on Metal (#2294)
785829dfe8baf0213f2ff66963d28c62f92d7930 fff0e0eafe817eef429ecb64f892ab7bdae31846 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-20T15:18:43+03:00 GitHub noreply@github.com 2023-07-20T15:18:43+03:00 Faster Q4_K on Metal (#2290)
fff0e0eafe817eef429ecb64f892ab7bdae31846 417a85a0010519224cf154eb85d383ffeafeeead Georgi Gerganov ggerganov@gmail.com 2023-07-20T13:47:26+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-20T13:47:26+03:00 llama : fix regression from #2000 - could not load no-mmap models
417a85a0010519224cf154eb85d383ffeafeeead 294f424554c1599784ac9962462fc39ace92d8a5 Shouzheng Liu lshzh.hi@gmail.com 2023-07-20T06:32:22-04:00 GitHub noreply@github.com 2023-07-20T13:32:22+03:00 metal: minor q4 optimization and reduce code size (#2248)
294f424554c1599784ac9962462fc39ace92d8a5 45a1b07e9b20c33d71d8c849ff27d693a75a0269 Rinne AsakusaRinne@gmail.com 2023-07-19T15:06:40+08:00 GitHub noreply@github.com 2023-07-19T10:06:40+03:00 llama : extend API to get max devices at runtime (#2253)
45a1b07e9b20c33d71d8c849ff27d693a75a0269 b1f429095328a34556c0e9a7a2fefced3db3368c wzy 32936898+Freed-Wu@users.noreply.github.com 2023-07-19T15:01:55+08:00 GitHub noreply@github.com 2023-07-19T10:01:55+03:00 flake : update flake.nix (#2270)
b1f429095328a34556c0e9a7a2fefced3db3368c d01bccde9f759b24449fdaa16306b406a50eb367 wzy 32936898+Freed-Wu@users.noreply.github.com 2023-07-19T15:01:11+08:00 GitHub noreply@github.com 2023-07-19T10:01:11+03:00 cmake : install targets (#2256)
d01bccde9f759b24449fdaa16306b406a50eb367 6cbf9dfb32f0e23ed3afd02d30ab066ed53e2c4d Georgi Gerganov ggerganov@gmail.com 2023-07-18T14:24:43+03:00 GitHub noreply@github.com 2023-07-18T14:24:43+03:00 ci : integrate with ggml-org/ci (#2250)
6cbf9dfb32f0e23ed3afd02d30ab066ed53e2c4d 7568d1a2b206331412106ea66da3f871025e0c3c Georgi Gerganov ggerganov@gmail.com 2023-07-18T11:50:49+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-18T11:50:49+03:00 llama : shorten quantization descriptions
7568d1a2b206331412106ea66da3f871025e0c3c b7647436ccc80970b44a270f70f4f2ea139054d1 Jiahao Li liplus17@163.com 2023-07-18T01:39:29+08:00 GitHub noreply@github.com 2023-07-17T20:39:29+03:00 Support dup & cont ops on CUDA (#2242)
b7647436ccc80970b44a270f70f4f2ea139054d1 672dda10e4d8ac79df5d5970da7fb69d242ca9a7 Alex Klinkhamer git@grencez.dev 2023-07-16T14:01:45-07:00 GitHub noreply@github.com 2023-07-17T00:01:45+03:00 llama : fix t_start_sample_us initialization warning (#2238)
672dda10e4d8ac79df5d5970da7fb69d242ca9a7 27ab66e437797aedbb23b3599385756b6c26ac39 Qingyou Meng meng.qingyou@gmail.com 2023-07-17T03:57:28+08:00 GitHub noreply@github.com 2023-07-16T22:57:28+03:00 ggml : fixed runtime bugs and compile errors related to GGML_PERF and GGML_DEBUG (#2219)
27ab66e437797aedbb23b3599385756b6c26ac39 6e7cca404748dd4b1a3affd0d1296e37f4ac0a6f Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2023-07-16T21:54:47+02:00 GitHub noreply@github.com 2023-07-16T22:54:47+03:00 py : turn verify-checksum-models.py into executable (#2245)
6e7cca404748dd4b1a3affd0d1296e37f4ac0a6f a6803cab946c817fb7aaf2a40b317f5d3e373bd1 Xiao-Yong Jin jinxiaoyong@gmail.com 2023-07-15T06:34:16-04:00 GitHub noreply@github.com 2023-07-15T13:34:16+03:00 llama : add custom RoPE (#2054)
a6803cab946c817fb7aaf2a40b317f5d3e373bd1 7dabc66f3c63f8ea0f61bac346fa138e01df675f Dave Della Costa ddellacosta+github@gmail.com 2023-07-14T15:13:38-04:00 GitHub noreply@github.com 2023-07-14T22:13:38+03:00 flake : add runHook preInstall/postInstall to installPhase so hooks function (#2224)
7dabc66f3c63f8ea0f61bac346fa138e01df675f 7cdd30bf1f84339c55a5e3de29384f6bbdebb61c wzy 32936898+Freed-Wu@users.noreply.github.com 2023-07-15T03:05:08+08:00 GitHub noreply@github.com 2023-07-14T22:05:08+03:00 make : use pkg-config for OpenBLAS (#2222)
7cdd30bf1f84339c55a5e3de29384f6bbdebb61c e8035f141e1f71d739fa5cfc9c01531cdee6fc16 Bach Le bach@bullno1.com 2023-07-15T03:00:58+08:00 GitHub noreply@github.com 2023-07-14T22:00:58+03:00 cuda : allocate all temporary ggml_tensor_extra_gpu from a fixed-size buffer (#2220)
e8035f141e1f71d739fa5cfc9c01531cdee6fc16 7513b7b0a1c11faa00ad5a34d22681e5f07d32e4 Evan Miller emmiller@gmail.com 2023-07-14T14:55:56-04:00 GitHub noreply@github.com 2023-07-14T21:55:56+03:00 ggml : fix static_assert with older compilers #2024 (#2218)
7513b7b0a1c11faa00ad5a34d22681e5f07d32e4 de8342423d9600cf6e15455c1a27bae441262b45 Bach Le bach@bullno1.com 2023-07-15T02:55:24+08:00 GitHub noreply@github.com 2023-07-14T21:55:24+03:00 llama : add functions that work directly on model (#2197)
de8342423d9600cf6e15455c1a27bae441262b45 c48c525f8711780f3f7c59bf92f1760f38317218 Ali Chraghi 63465728+alichraghi@users.noreply.github.com 2023-07-14T11:50:58-07:00 GitHub noreply@github.com 2023-07-14T21:50:58+03:00 build.zig : install config header (#2216)
c48c525f8711780f3f7c59bf92f1760f38317218 206e01de117cc65ed8713ac9fdebc57ba4532ec3 Shangning Xu 32517059+xushangning@users.noreply.github.com 2023-07-15T02:40:05+08:00 GitHub noreply@github.com 2023-07-14T21:40:05+03:00 examples : fixed path typos in embd-input (#2214)
206e01de117cc65ed8713ac9fdebc57ba4532ec3 4304bd3cded73c867a882ea5ca4517e3995cc996 Jiahao Li liplus17@163.com 2023-07-15T02:38:24+08:00 GitHub noreply@github.com 2023-07-14T21:38:24+03:00 cuda : support broadcast add & mul (#2192)
4304bd3cded73c867a882ea5ca4517e3995cc996 229aab351c375899debad45fcb213bf0565bba4e Johannes Gäßler johannesg@5d6.de 2023-07-14T19:44:08+02:00 GitHub noreply@github.com 2023-07-14T19:44:08+02:00 CUDA: mul_mat_vec_q kernels for k-quants (#2203)
229aab351c375899debad45fcb213bf0565bba4e 697966680b27d9b4f05668605b863cb9aea3e15f James Reynolds magnusviri@users.noreply.github.com 2023-07-14T11:34:40-06:00 GitHub noreply@github.com 2023-07-14T20:34:40+03:00 make : fix combination of LLAMA_METAL and LLAMA_MPI (#2208)
697966680b27d9b4f05668605b863cb9aea3e15f 27ad57a69b85bf12420a27e9945e580cc280be57 Georgi Gerganov ggerganov@gmail.com 2023-07-14T16:36:41+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-14T16:36:41+03:00 ggml : sync (ggml_conv_2d, fix mul_mat bug, CUDA GLM rope)
27ad57a69b85bf12420a27e9945e580cc280be57 32c54116318929c90fd7ae814cf9b5232cd44c36 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-07-14T12:46:21+03:00 GitHub noreply@github.com 2023-07-14T11:46:21+02:00 Metal: faster Q4_0 and Q4_1 matrix x vector kernels (#2212)
32c54116318929c90fd7ae814cf9b5232cd44c36 ff5d58faecf1f02b05bd015bdfc6a394cf2bc9ba Howard Su howard0su@gmail.com 2023-07-13T21:58:25+08:00 GitHub noreply@github.com 2023-07-13T21:58:25+08:00 Revert "Support using mmap when applying LoRA (#2095)" (#2206)
ff5d58faecf1f02b05bd015bdfc6a394cf2bc9ba b782422a3e090d0aeab84bfa03ba008dcd1c2a3d Howard Su howard0su@gmail.com 2023-07-13T21:58:09+08:00 GitHub noreply@github.com 2023-07-13T21:58:09+08:00 Fix compile error on Windows CUDA (#2207)
b782422a3e090d0aeab84bfa03ba008dcd1c2a3d 1cbf561466e957b25f0e8163c2386683f8674369 Bodo Graumann mail@bodograumann.de 2023-07-13T15:49:14+02:00 GitHub noreply@github.com 2023-07-13T16:49:14+03:00 devops : add missing quotes to bash script (#2193)
1cbf561466e957b25f0e8163c2386683f8674369 975221e9548ef6d9f4af8d39cdffc4811c050beb Shouzheng Liu 61452103+lshzh-ww@users.noreply.github.com 2023-07-12T16:10:55-04:00 GitHub noreply@github.com 2023-07-12T23:10:55+03:00 metal : new q4_0 matrix-vector kernel (#2188)
975221e9548ef6d9f4af8d39cdffc4811c050beb 4523d10d0cf8c088f1b26c76d38d73290eb3b444 Georgi Gerganov ggerganov@gmail.com 2023-07-12T20:51:29+03:00 GitHub noreply@github.com 2023-07-12T20:51:29+03:00 ggml : broadcast mul_mat + conv batch support (#2199)
4523d10d0cf8c088f1b26c76d38d73290eb3b444 680e6f91775f972f0df34f56807f30826370db59 Georgi Gerganov ggerganov@gmail.com 2023-07-12T20:27:03+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-12T20:32:15+03:00 ggml : add ggml_pool_1d and ggml_pool_2d
680e6f91775f972f0df34f56807f30826370db59 4e7464ef88885cb3532738b03cac890f4077fa20 Georgi Gerganov ggerganov@gmail.com 2023-07-12T20:26:18+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-12T20:32:15+03:00 cuda : add gelu support
4e7464ef88885cb3532738b03cac890f4077fa20 2b5eb72e109577ed84e44bb8fa47e4956f337300 Howard Su howard0su@gmail.com 2023-07-12T20:18:40+08:00 GitHub noreply@github.com 2023-07-12T20:18:40+08:00 FP16 is supported in CM=6.0 (#2177)
2b5eb72e109577ed84e44bb8fa47e4956f337300 f7d278faf308cb989c221895968f2a26f14b2155 Johannes Gäßler johannesg@5d6.de 2023-07-12T10:38:52+02:00 GitHub noreply@github.com 2023-07-12T10:38:52+02:00 Fixed __dp4a compute capability: 6.0 -> 6.1 (#2189)
f7d278faf308cb989c221895968f2a26f14b2155 20d7740a9b45f6e5b247fa3738fdda35e18c2e8a Georgi Gerganov ggerganov@gmail.com 2023-07-12T10:54:19+03:00 GitHub noreply@github.com 2023-07-12T10:54:19+03:00 ggml : revert CUDA broadcast changes from #2183 (#2191)
20d7740a9b45f6e5b247fa3738fdda35e18c2e8a 5bf2a2771886ee86137e01dbc7492f78fb392066 Georgi Gerganov ggerganov@gmail.com 2023-07-11T22:53:34+03:00 GitHub noreply@github.com 2023-07-11T22:53:34+03:00 ggml : sync (abort callback, mul / add broadcast, fix alibi) (#2183)
5bf2a2771886ee86137e01dbc7492f78fb392066 c9c74b4e3f9dcfab8b0032749ff8a579ab4e4d8d Spencer Sutton spencersutton@users.noreply.github.com 2023-07-11T12:31:10-04:00 GitHub noreply@github.com 2023-07-11T19:31:10+03:00 ggml : remove src0 and src1 from ggml_tensor and rename opt to src (#2178)
c9c74b4e3f9dcfab8b0032749ff8a579ab4e4d8d 3ec7e596b2ba3f43c22f441254ca2bcfa91102ba Bach Le bach@bullno1.com 2023-07-12T00:18:43+08:00 GitHub noreply@github.com 2023-07-11T19:18:43+03:00 llama : add classifier-free guidance (#2135)
3ec7e596b2ba3f43c22f441254ca2bcfa91102ba 917831c63a4138814d23da1917bf2b5d5b9faa6c Jinwoo Jeong 33892306+williamjeong2@users.noreply.github.com 2023-07-12T01:12:35+09:00 GitHub noreply@github.com 2023-07-11T19:12:35+03:00 docker : add '--server' option (#2174)
917831c63a4138814d23da1917bf2b5d5b9faa6c 2347463201a9f4159ae95b737e1544dd300569c8 Chad Brewbaker crb002@gmail.com 2023-07-11T11:03:06-05:00 GitHub noreply@github.com 2023-07-11T19:03:06+03:00 readme : fix zig build instructions (#2171)
2347463201a9f4159ae95b737e1544dd300569c8 bbef28218fe827265716b66977719b9ee2b21165 Howard Su howard0su@gmail.com 2023-07-11T22:37:01+08:00 GitHub noreply@github.com 2023-07-11T22:37:01+08:00 Support using mmap when applying LoRA (#2095)
bbef28218fe827265716b66977719b9ee2b21165 5656d10599bd756dc0f17284e418e704200b43f3 LostRuins 39025047+LostRuins@users.noreply.github.com 2023-07-11T22:01:08+08:00 GitHub noreply@github.com 2023-07-11T22:01:08+08:00 Possible solution to allow K-quants on models with n_vocab!=32000 (#2148)
5656d10599bd756dc0f17284e418e704200b43f3 1d1630996920f889cdc08de26cebf2415958540e Evan Miller emmiller@gmail.com 2023-07-10T11:49:56-04:00 GitHub noreply@github.com 2023-07-10T18:49:56+03:00 mpi : add support for distributed inference via MPI (#2099)
1d1630996920f889cdc08de26cebf2415958540e db4047ad5cd8eae04db3b2efe0245e69a376601a oobabooga 112222186+oobabooga@users.noreply.github.com 2023-07-09T05:59:53-03:00 GitHub noreply@github.com 2023-07-09T11:59:53+03:00 llama : remove "first token must be BOS" restriction (#2153)
db4047ad5cd8eae04db3b2efe0245e69a376601a 18780e0a5e17348236230bbe891901b9b5718709 Nigel Bosch pnigelb@gmail.com 2023-07-09T03:56:18-05:00 GitHub noreply@github.com 2023-07-09T11:56:18+03:00 main : escape prompt prefix/suffix (#2151)
18780e0a5e17348236230bbe891901b9b5718709 3bbc1a11f04a9adc0d0e08c2940ba4d2978755ab JackJollimore 130917767+JackJollimore@users.noreply.github.com 2023-07-09T05:20:43-03:00 GitHub noreply@github.com 2023-07-09T11:20:43+03:00 readme : update Termux instructions (#2147)
3bbc1a11f04a9adc0d0e08c2940ba4d2978755ab 2492a53fd0d8372ecc67f49f07b581905175eea8 clyang clyang@clyang.net 2023-07-09T16:12:20+08:00 GitHub noreply@github.com 2023-07-09T11:12:20+03:00 ggml : fix buidling with Intel MKL but ask for "cblas.h" issue (#2104) (#2115)
2492a53fd0d8372ecc67f49f07b581905175eea8 64639555ff93c8ead2b80becb49cc6b60aeac240 rankaiyx rankaiyx@rankaiyx.com 2023-07-09T15:38:42+08:00 GitHub noreply@github.com 2023-07-09T10:38:42+03:00 readme : add more docs indexes (#2127)
64639555ff93c8ead2b80becb49cc6b60aeac240 061f5f8d2109bb7adcbd40f1b456d887c5a1df25 Johannes Gäßler johannesg@5d6.de 2023-07-08T20:01:44+02:00 GitHub noreply@github.com 2023-07-08T20:01:44+02:00 Fixed OpenLLaMA 3b CUDA mul_mat_vec_q (#2144)
061f5f8d2109bb7adcbd40f1b456d887c5a1df25 84525e7962bee0abef91108948bbf7f7bfdcf421 Johannes Gäßler johannesg@5d6.de 2023-07-08T00:25:15+02:00 GitHub noreply@github.com 2023-07-08T00:25:15+02:00 CUDA: add __restrict__ to mul mat vec kernels (#2140)
84525e7962bee0abef91108948bbf7f7bfdcf421 a7e20edf2266169ccd97a4eb949a593d628fbd64 dylan canardleteer@users.noreply.github.com 2023-07-07T11:25:25-07:00 GitHub noreply@github.com 2023-07-07T21:25:25+03:00 docker : add support for CUDA in docker (#1461)
a7e20edf2266169ccd97a4eb949a593d628fbd64 1d656d6360359cfdaaf5d64ed9690047b600dbcb Georgi Gerganov ggerganov@gmail.com 2023-07-07T21:23:57+03:00 GitHub noreply@github.com 2023-07-07T21:23:57+03:00 ci : switch threads to 1 (#2138)
1d656d6360359cfdaaf5d64ed9690047b600dbcb 72421402834141df6cbdcf595fe46dbd11874dce Qingyou Meng meng.qingyou@gmail.com 2023-07-08T00:24:01+08:00 GitHub noreply@github.com 2023-07-07T19:24:01+03:00 ggml : change ggml_graph_compute() API to not require context (#1999)
72421402834141df6cbdcf595fe46dbd11874dce 3e08ae99ceb143d67f9273fda47541e9d98ff23f Georgi Gerganov ggerganov@gmail.com 2023-07-07T18:36:37+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-07T18:37:10+03:00 ggml : remove sched_yield() call in ggml_graph_compute_thread() (#2134)
3e08ae99ceb143d67f9273fda47541e9d98ff23f 481f793acc3882a09d45d8d2c3076ad3d1c60cfc Aarni Koskela akx@iki.fi 2023-07-07T16:12:49+03:00 GitHub noreply@github.com 2023-07-07T09:12:49-04:00 convert.py: add mapping for safetensors bf16 (#1598)
481f793acc3882a09d45d8d2c3076ad3d1c60cfc dfd9fce6d65599bf33df43e616e85aa639bdae4c Howard Su howard0su@gmail.com 2023-07-07T11:34:18+08:00 GitHub noreply@github.com 2023-07-07T05:34:18+02:00 Fix opencl by wrap #if-else-endif with \n (#2086)
dfd9fce6d65599bf33df43e616e85aa639bdae4c 36680f6e40e4440c3ec3385d0b7e5ca8bb6c37f7 Georgi Gerganov ggerganov@gmail.com 2023-07-06T19:41:31+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-06T19:41:31+03:00 ggml : fix restrict usage
36680f6e40e4440c3ec3385d0b7e5ca8bb6c37f7 a17a2683d8fdb899ba497d0c28ccafb28c62efb6 Judd foldl@users.noreply.github.com 2023-07-07T00:23:49+08:00 GitHub noreply@github.com 2023-07-06T19:23:49+03:00 convert : update for baichuan (#2081)
a17a2683d8fdb899ba497d0c28ccafb28c62efb6 31cfbb1013a482e89c72146e2063ac4362becae7 tslmy tslmy@users.noreply.github.com 2023-07-06T09:17:50-07:00 GitHub noreply@github.com 2023-07-06T19:17:50+03:00 alpaca.sh : update model file name (#2074)
31cfbb1013a482e89c72146e2063ac4362becae7 983b555e9ddb36703cee4d22642afe958de093b7 Tobias Lütke tobi@shopify.com 2023-07-05T16:51:13-04:00 GitHub noreply@github.com 2023-07-05T16:51:13-04:00 Expose generation timings from server & update completions.js (#2116)
983b555e9ddb36703cee4d22642afe958de093b7 ec326d350c72afd23709a409944728a607188cc0 Jesse Jojo Johnson williamsaintgeorge@gmail.com 2023-07-05T18:03:19Z GitHub noreply@github.com 2023-07-05T21:03:19+03:00 Update Server Instructions (#2113)
ec326d350c72afd23709a409944728a607188cc0 1b6efeab829f3eeda5b39bd47624bb60b3531b88 Georgi Gerganov ggerganov@gmail.com 2023-07-05T20:44:11+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-05T20:44:11+03:00 ggml : fix bug introduced in #1237
1b6efeab829f3eeda5b39bd47624bb60b3531b88 1b107b8550dced48dc5f41184640061354226b96 Georgi Gerganov ggerganov@gmail.com 2023-07-05T20:20:05+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-05T20:20:25+03:00 tests : fix test-grad0
1b107b8550dced48dc5f41184640061354226b96 8567c76b5326e862be0755a8dc1dd988223fcae3 Stephan Walter stephan@walter.name 2023-07-05T16:13:06Z GitHub noreply@github.com 2023-07-05T19:13:06+03:00 ggml : generalize `quantize_fns` for simpler FP16 handling (#1237)
8567c76b5326e862be0755a8dc1dd988223fcae3 924dd22fd3ba93e097f8d19ba5cda919ca2fe2fb Jesse Jojo Johnson williamsaintgeorge@gmail.com 2023-07-05T15:13:35Z GitHub noreply@github.com 2023-07-05T18:13:35+03:00 Update server instructions for web front end (#2103)
924dd22fd3ba93e097f8d19ba5cda919ca2fe2fb 051c70dcd55709c9cbbfa849af035951fe720433 Johannes Gäßler johannesg@5d6.de 2023-07-05T14:19:42+02:00 GitHub noreply@github.com 2023-07-05T14:19:42+02:00 Quantized dot products for CUDA mul mat vec (#2067)
051c70dcd55709c9cbbfa849af035951fe720433 9e4475f5cf639315f61ed7b8da6258bb0c7c5ca9 Howard Su howard0su@gmail.com 2023-07-05T18:31:23+08:00 GitHub noreply@github.com 2023-07-05T18:31:23+08:00 llama: Don't double count the sampling time (#2107)
9e4475f5cf639315f61ed7b8da6258bb0c7c5ca9 7f0e9a775ecc4c6ade271c217f63d6dc93e79eaa Johannes Gäßler johannesg@5d6.de 2023-07-05T08:58:05+02:00 GitHub noreply@github.com 2023-07-05T08:58:05+02:00 Fixed OpenCL offloading prints (#2082)
7f0e9a775ecc4c6ade271c217f63d6dc93e79eaa b472f3fca558b6335adbd87210ed58cfb5da37cb Nigel Bosch pnigelb@gmail.com 2023-07-04T18:33:33-05:00 GitHub noreply@github.com 2023-07-05T07:33:33+08:00 embd-input: Fix input embedding example unsigned int seed (#2105)
b472f3fca558b6335adbd87210ed58cfb5da37cb ed9a54e5129a11c2a5b555e1dc65e875e3c37b4f Georgi Gerganov ggerganov@gmail.com 2023-07-04T22:25:22+03:00 GitHub noreply@github.com 2023-07-04T22:25:22+03:00 readme : add link web chat PR
ed9a54e5129a11c2a5b555e1dc65e875e3c37b4f f257fd255044decffad93dee2502875ce66ad80c Georgi Gerganov ggerganov@gmail.com 2023-07-04T21:54:11+03:00 GitHub noreply@github.com 2023-07-04T21:54:11+03:00 ggml : sync latest (new ops, macros, refactoring) (#2106)
f257fd255044decffad93dee2502875ce66ad80c 7ee76e45afae7f9a7a53e93393accfb5b36684e1 jwj7140 32943891+jwj7140@users.noreply.github.com 2023-07-05T03:06:12+09:00 GitHub noreply@github.com 2023-07-04T21:06:12+03:00 Add an API example using server.cpp similar to OAI. (#2009)
7ee76e45afae7f9a7a53e93393accfb5b36684e1 acc111caf93fc6681450924df9f99679c384c59e Tobias Lütke tobi@shopify.com 2023-07-04T10:05:27-04:00 GitHub noreply@github.com 2023-07-04T16:05:27+02:00 Simple webchat for server (#1998)
acc111caf93fc6681450924df9f99679c384c59e 23c7c6fc9182b041f006b86ea1e7f99911ecf344 Henri Vasserman henv@hot.ee 2023-07-04T15:38:04+03:00 GitHub noreply@github.com 2023-07-04T15:38:04+03:00 Allow old Make to build server. (#2098)
23c7c6fc9182b041f006b86ea1e7f99911ecf344 698efad5fbbf326f01288649b123eff5f79b417e ZhouYuChen zhouyuchen@naver.com 2023-07-04T20:15:16+08:00 GitHub noreply@github.com 2023-07-04T14:15:16+02:00 Update Makefile: clean simple (#2097)
698efad5fbbf326f01288649b123eff5f79b417e 14a2cc71f62e45803ae70890ffbdeb0a172e6210 Erik Scholz Green-Sky@users.noreply.github.com 2023-07-04T01:50:12+02:00 GitHub noreply@github.com 2023-07-04T01:50:12+02:00 CI: make the brew update temporarily optional. (#2092)
14a2cc71f62e45803ae70890ffbdeb0a172e6210 1cf14ccef12e19c5a5b0b17ab456242d1f8c7fdd Govlzkoy gotope@users.noreply.github.com 2023-07-04T07:50:00+08:00 GitHub noreply@github.com 2023-07-04T07:50:00+08:00 [ggml] fix index for ne03 value in ggml_cl_mul_f32 (#2088)
1cf14ccef12e19c5a5b0b17ab456242d1f8c7fdd cc45a7feb8412e84ff292207621412fffc0d3d51 Henri Vasserman henv@hot.ee 2023-07-04T00:05:23+03:00 GitHub noreply@github.com 2023-07-04T00:05:23+03:00 fix server crashes (#2076)
cc45a7feb8412e84ff292207621412fffc0d3d51 55dbb915cc2a95048f56e667b09dfad38d840421 Howard Su howard0su@gmail.com 2023-07-04T02:43:55+08:00 GitHub noreply@github.com 2023-07-03T20:43:55+02:00 Fix crash of test-tokenizer-0 under Debug build (#2064)
55dbb915cc2a95048f56e667b09dfad38d840421 d7d2e6a0f0c74f7a570dae384dfff371ac744d2a Howard Su howard0su@gmail.com 2023-07-03T19:58:58+08:00 GitHub noreply@github.com 2023-07-03T19:58:58+08:00 [llama] No need to check file version when loading vocab score (#2079)
d7d2e6a0f0c74f7a570dae384dfff371ac744d2a 46088f72318981341a2d646f12f6eee6aec06d65 WangHaoranRobin 56047610+WangHaoranRobin@users.noreply.github.com 2023-07-03T05:38:44+08:00 GitHub noreply@github.com 2023-07-03T00:38:44+03:00 server: add option to output probabilities for completion (#1962)
46088f72318981341a2d646f12f6eee6aec06d65 0bc2cdfc875fa7877d8e01c8bb17066f99c08f21 Georgi Gerganov ggerganov@gmail.com 2023-07-02T09:46:46+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-02T09:46:46+03:00 ggml : fix build with OpenBLAS (close #2066)
0bc2cdfc875fa7877d8e01c8bb17066f99c08f21 befb3a35627432473f143c90994557d78ff5bc67 Johannes Gäßler johannesg@5d6.de 2023-07-01T21:49:44+02:00 GitHub noreply@github.com 2023-07-01T21:49:44+02:00 Better CUDA synchronization logic (#2057)
befb3a35627432473f143c90994557d78ff5bc67 b2132270678c473f7cd9ba871b03d694126bc33a Johannes Gäßler johannesg@5d6.de 2023-07-01T21:47:26+02:00 GitHub noreply@github.com 2023-07-01T21:47:26+02:00 Test-based VRAM scratch size + context adjustment (#2056)
b2132270678c473f7cd9ba871b03d694126bc33a 2f8cd979ecd1fa582852e7136e92ff8990b98fd8 Daniel Drake drake@endlessos.org 2023-07-01T20:31:44+02:00 GitHub noreply@github.com 2023-07-01T21:31:44+03:00 cmake : don't force -mcpu=native on aarch64 (#2063)
2f8cd979ecd1fa582852e7136e92ff8990b98fd8 471aab6e4cb89d8ef6d043f1bc93acb6eb78ab67 Aaron Miller apage43@ninjawhale.com 2023-07-01T11:14:59-07:00 GitHub noreply@github.com 2023-07-01T21:14:59+03:00 metal : release buffers when freeing metal context (#2062)
471aab6e4cb89d8ef6d043f1bc93acb6eb78ab67 463f2f4c4f8dd5ca9594b7d65849f346f0effe05 Judd foldl@users.noreply.github.com 2023-07-02T01:00:25+08:00 GitHub noreply@github.com 2023-07-01T20:00:25+03:00 convert : add support of baichuan-7b (#2055)
463f2f4c4f8dd5ca9594b7d65849f346f0effe05 cb44dbc7de287b3d17772cfb1aa49d55e082ce5b Georgi Gerganov ggerganov@gmail.com 2023-07-01T19:05:09+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-01T19:05:09+03:00 llama : fix return value of llama_load_session_file_internal (#2022)
cb44dbc7de287b3d17772cfb1aa49d55e082ce5b 79f634a19d1c32a6cfb1befc21551ee684fced6b Rand Xie randxiexyy29@gmail.com 2023-07-02T00:02:58+08:00 GitHub noreply@github.com 2023-07-01T19:02:58+03:00 llama : catch llama_load_session_file_internal exceptions (#2022)
79f634a19d1c32a6cfb1befc21551ee684fced6b 04606a159947566b27810508433e6ca5dbc684ba Georgi Gerganov ggerganov@gmail.com 2023-07-01T18:46:00+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-01T18:46:00+03:00 embd-input : fix returning ptr to temporary
04606a159947566b27810508433e6ca5dbc684ba b1ca8f36a9cdbcee5f5c425df717611a1040a897 Georgi Gerganov ggerganov@gmail.com 2023-07-01T18:45:44+03:00 Georgi Gerganov ggerganov@gmail.com 2023-07-01T18:45:44+03:00 train : fix compile warning
b1ca8f36a9cdbcee5f5c425df717611a1040a897 b8c8dda75fdf5fdea49c80af36818e7c30fe0ddf Qingyou Meng meng.qingyou@gmail.com 2023-07-01T23:42:43+08:00 GitHub noreply@github.com 2023-07-01T18:42:43+03:00 ggml : disable GGML_TASK_INIT and GGML_TASK_FINALIZE by default (#1995)
b8c8dda75fdf5fdea49c80af36818e7c30fe0ddf 96a712ca1b7f427e3bd7ffc0c70b2105cfc7fbf1 Howard Su howard0su@gmail.com 2023-06-29T21:15:15+08:00 GitHub noreply@github.com 2023-06-29T06:15:15-07:00 Use unsigned for random seed (#2006)
96a712ca1b7f427e3bd7ffc0c70b2105cfc7fbf1 d3494bb86bf7ad5b0b60aae0220ea576f273b5c0 LostRuins 39025047+LostRuins@users.noreply.github.com 2023-06-29T11:56:43+08:00 GitHub noreply@github.com 2023-06-29T05:56:43+02:00 Porting the improved K-Quant CUDA kernels to OpenCL (#1966)
d3494bb86bf7ad5b0b60aae0220ea576f273b5c0 5b351e94d041742cd50ffcf2d44718d63bab398a m3ndax adrian.goessl@outlook.com 2023-06-28T20:39:08+02:00 GitHub noreply@github.com 2023-06-28T21:39:08+03:00 llama : replacing auto &kv with const auto &kv (#2041)
5b351e94d041742cd50ffcf2d44718d63bab398a 6432aabb6dc887436e4d57414b63116189c3b13b Salvador E. Tropea stropea@inti.gob.ar 2023-06-28T14:27:31-03:00 GitHub noreply@github.com 2023-06-28T20:27:31+03:00 cuda : remove nchannels_x argument from mul_mat_vec_nc_f16_f32 (#2028)
6432aabb6dc887436e4d57414b63116189c3b13b b922bc351b69770cec2d35d2aa50fa052b95ca93 Salvador E. Tropea stropea@inti.gob.ar 2023-06-28T14:26:26-03:00 GitHub noreply@github.com 2023-06-28T20:26:26+03:00 cuda : fix missing const qualifier in casts (#2027)
b922bc351b69770cec2d35d2aa50fa052b95ca93 7f9753fa1263c4eded9a3de19778562f0e1093d7 Howard Su howard0su@gmail.com 2023-06-28T10:13:02-07:00 GitHub noreply@github.com 2023-06-28T20:13:02+03:00 llama : remove shards weight file support (#2000)
7f9753fa1263c4eded9a3de19778562f0e1093d7 cfa0750bc9dbc2d957a91b8ed09ab0035d8f3d4e Johannes Gäßler johannesg@5d6.de 2023-06-28T18:35:54+02:00 GitHub noreply@github.com 2023-06-28T18:35:54+02:00 CUDA GPU acceleration for LoRAs + f16 models (#1970)
cfa0750bc9dbc2d957a91b8ed09ab0035d8f3d4e 9d23589d638dc74577d5ff880e6d4248b795f12e ningshanwutuobang ningshanwutuobang@gmail.com 2023-06-28T23:53:37+08:00 GitHub noreply@github.com 2023-06-28T18:53:37+03:00 llama : support input embeddings directly (#1910)
9d23589d638dc74577d5ff880e6d4248b795f12e 0be54f75a6c3e9a09ea71bdfcdabf9a996a0549b Erik Scholz Green-Sky@users.noreply.github.com 2023-06-27T19:06:33+02:00 GitHub noreply@github.com 2023-06-27T19:06:33+02:00 fix pthreads setaffinity usage on android (#2020)
0be54f75a6c3e9a09ea71bdfcdabf9a996a0549b 181e8d975528a4e27eabb8ae6e9865f9ceae4b37 Howard Su howard0su@gmail.com 2023-06-27T13:07:13+08:00 GitHub noreply@github.com 2023-06-27T08:07:13+03:00 baby-llama : fix build after ggml_rope change (#2016)
181e8d975528a4e27eabb8ae6e9865f9ceae4b37 d9779021bd59ed96daae75e820a5ac5da47ca8ff Georgi Gerganov ggerganov@gmail.com 2023-06-27T00:37:13+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-27T00:37:33+03:00 llama : fix rope usage after ChatGLM change
d9779021bd59ed96daae75e820a5ac5da47ca8ff d38e45157862b58a1824387e64860d68ca3533a7 Georgi Gerganov ggerganov@gmail.com 2023-06-27T00:06:51+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-27T00:06:51+03:00 ggml : add support for ChatGLM RoPE
d38e45157862b58a1824387e64860d68ca3533a7 eaa6ca5a61b8c9501df9ebe3d264f45b75a5f8aa Roman Parykin donderom@gmail.com 2023-06-26T22:47:59+03:00 GitHub noreply@github.com 2023-06-26T22:47:59+03:00 readme : add Scala 3 bindings repo (#2010)
eaa6ca5a61b8c9501df9ebe3d264f45b75a5f8aa aa777abbb73655c4e1e9237b7c0ad66745e8e48c David Yang davidyang6us@gmail.com 2023-06-27T03:45:32+08:00 GitHub noreply@github.com 2023-06-26T22:45:32+03:00 ggml : increase max tensor name + clean up compiler warnings in train-text (#1988)
aa777abbb73655c4e1e9237b7c0ad66745e8e48c c824d2e368d193d9f564ff29880a51cda9f90527 Gustavo Rocha Dias 91472747+gustrd@users.noreply.github.com 2023-06-26T16:34:45-03:00 GitHub noreply@github.com 2023-06-26T22:34:45+03:00 readme : LD_LIBRARY_PATH complement for some Android devices when building with CLBlast inside Termux (#2007)
c824d2e368d193d9f564ff29880a51cda9f90527 b853d456018b10820686362af41b2f2f75f1eec6 Georgi Gerganov ggerganov@gmail.com 2023-06-26T21:03:59+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-26T21:03:59+03:00 ggml : avoid conv 2d kernel round up
b853d456018b10820686362af41b2f2f75f1eec6 9225baef71407d799a6f7f563b77fd7f82791416 zrm trustiosity.zrm@gmail.com 2023-06-26T13:57:59-04:00 GitHub noreply@github.com 2023-06-26T20:57:59+03:00 ggml : add NUMA support (#1556)
9225baef71407d799a6f7f563b77fd7f82791416 a84ab1da8dc6a59a5b67420ae1322f09503ffc72 Georgi Gerganov ggerganov@gmail.com 2023-06-26T20:10:52+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-26T20:10:52+03:00 k-quants : fix indentation
a84ab1da8dc6a59a5b67420ae1322f09503ffc72 5743ca80928d8410754ec64a5673d5c2dd6cfbb7 katsu560 118887472+katsu560@users.noreply.github.com 2023-06-27T01:47:02+09:00 GitHub noreply@github.com 2023-06-26T19:47:02+03:00 tests : fix quantize perf (#1990)
5743ca80928d8410754ec64a5673d5c2dd6cfbb7 412c60e4739367144e51e59add5dc7749d084115 katsu560 118887472+katsu560@users.noreply.github.com 2023-06-27T01:46:07+09:00 GitHub noreply@github.com 2023-06-26T19:46:07+03:00 k-quants : add AVX support to dot functions (#1916)
412c60e4739367144e51e59add5dc7749d084115 6769e944c727c63612dcafbef52009d21ae00fff Georgi Gerganov ggerganov@gmail.com 2023-06-26T19:45:09+03:00 GitHub noreply@github.com 2023-06-26T19:45:09+03:00 readme : add link to new k-quants for visibility
6769e944c727c63612dcafbef52009d21ae00fff cbebf61ca7584e9709265395f0127ae7fc0f1882 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-26T19:43:07+03:00 GitHub noreply@github.com 2023-06-26T19:43:07+03:00 k-quants : support for super-block size of 64 (#2001)
cbebf61ca7584e9709265395f0127ae7fc0f1882 447ccbe8c39332fcdd0d98a041b6e2ff6f06219d Howard Su howard0su@gmail.com 2023-06-26T23:15:47+08:00 GitHub noreply@github.com 2023-06-26T23:15:47+08:00 Fix assert when free invalid cuda pointer (#2005)
447ccbe8c39332fcdd0d98a041b6e2ff6f06219d bd34cdde38f8fd661890ddd5f57ca30bf279877b Georgi Gerganov ggerganov@gmail.com 2023-06-25T16:08:12+03:00 GitHub noreply@github.com 2023-06-25T16:08:12+03:00 readme : add new roadmap + manifesto
bd34cdde38f8fd661890ddd5f57ca30bf279877b c2a08f87b8d180115d04b8688f383d1b2761b16d Georgi Gerganov ggerganov@gmail.com 2023-06-25T14:25:08+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-25T14:25:08+03:00 ggml : sync latest ggml (custom operators)
c2a08f87b8d180115d04b8688f383d1b2761b16d 66a2555ba6cab954c56d653b29c27bfbbacfbfb1 anon998 131767832+anon998@users.noreply.github.com 2023-06-25T08:48:36Z GitHub noreply@github.com 2023-06-25T10:48:36+02:00 fix server sampling: top k sampler first (#1977)
66a2555ba6cab954c56d653b29c27bfbbacfbfb1 e65ca7e14ac76c4046091da39d41a9017abaa9b3 Georgi Gerganov ggerganov@gmail.com 2023-06-25T09:07:03+03:00 GitHub noreply@github.com 2023-06-25T09:07:03+03:00 readme : add Azure CI discussion link
e65ca7e14ac76c4046091da39d41a9017abaa9b3 5ec8dd5a3c6a9a109351d2257bb9d53869bd0a94 sjinzh sjinzh@gmail.com 2023-06-25T13:45:44+08:00 GitHub noreply@github.com 2023-06-25T08:45:44+03:00 zig : upgrade build system support (#1981)
5ec8dd5a3c6a9a109351d2257bb9d53869bd0a94 65bdd52a867539691007f85c5508146d507f72c1 Robyn robyngraf@users.noreply.github.com 2023-06-25T04:10:29+10:00 GitHub noreply@github.com 2023-06-24T20:10:29+02:00 #1869 Fix null reference errors when training from scratch with CUDA (#1907)
65bdd52a867539691007f85c5508146d507f72c1 fdd18609113862dc6eb34dfc44a093d54c59ff1f Georgi Gerganov ggerganov@gmail.com 2023-06-24T19:40:18+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-24T19:40:18+03:00 tests : sync test-grad0 from ggml
fdd18609113862dc6eb34dfc44a093d54c59ff1f c943d823c14cef33092205ca3944de6fdf7abf99 Rowan Hart rowanbhart@gmail.com 2023-06-24T04:07:08-07:00 GitHub noreply@github.com 2023-06-24T14:07:08+03:00 flake : fix ggml-metal.metal path and run nixfmt (#1974)
c943d823c14cef33092205ca3944de6fdf7abf99 f2c754e1c38936fdde74e4848ac468a696eb73c6 AN Long aisk@users.noreply.github.com 2023-06-24T19:02:06+08:00 GitHub noreply@github.com 2023-06-24T14:02:06+03:00 convert : fix invalid params in write_vocab_only (#1975)
f2c754e1c38936fdde74e4848ac468a696eb73c6 11da1a85cd69af84b5861134738c7e9e20907470 slaren slarengh@gmail.com 2023-06-24T12:57:18+02:00 GitHub noreply@github.com 2023-06-24T13:57:18+03:00 ggml : improve ggml_graph_dump_dot, add ggml_format_name (#1978)
11da1a85cd69af84b5861134738c7e9e20907470 235b610d650cbfed6dbd5d671f750d35fc18cd7d Georgi Gerganov ggerganov@gmail.com 2023-06-24T13:38:18+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-24T13:38:18+03:00 readme : fix whitespaces
235b610d650cbfed6dbd5d671f750d35fc18cd7d b061ba9e2a7a2c335a200df8c11aed5e31e4ccbb Alberto 57916483+albbus-stack@users.noreply.github.com 2023-06-24T12:32:13+02:00 GitHub noreply@github.com 2023-06-24T13:32:13+03:00 readme : fixed termux instructions (#1973)
b061ba9e2a7a2c335a200df8c11aed5e31e4ccbb 527b6fba1d237befb324fd846bda7418c0fa394d Alex Renda alexrenda@users.noreply.github.com 2023-06-24T03:15:01-07:00 GitHub noreply@github.com 2023-06-24T13:15:01+03:00 llama : fix top-p sampling to match the canonical definition (#1953)
527b6fba1d237befb324fd846bda7418c0fa394d d7b7484f74d486f77feb4c0b7af7e1718ed91651 Didzis Gosko didzis@users.noreply.github.com 2023-06-24T11:47:58+03:00 GitHub noreply@github.com 2023-06-24T11:47:58+03:00 llama : make model stateless and context stateful (llama_state) (#1797)
d7b7484f74d486f77feb4c0b7af7e1718ed91651 7487137227eb32ed9b12156338b865cb29b2dfd1 eiery 19350831+eiery@users.noreply.github.com 2023-06-23T04:38:01-04:00 GitHub noreply@github.com 2023-06-23T10:38:01+02:00 Add OpenLLaMA instructions to the README (#1954)
7487137227eb32ed9b12156338b865cb29b2dfd1 bbca06e26949686d61a5126332680ba3cccf235c Erik Scholz Green-Sky@users.noreply.github.com 2023-06-22T14:20:47+02:00 GitHub noreply@github.com 2023-06-22T14:20:47+02:00 rework convert.py to read hyper-parameters from config.json (#1958)
bbca06e26949686d61a5126332680ba3cccf235c fb98254f99d769fcbbf20966ef386abdb48ef601 Johannes Gäßler johannesg@5d6.de 2023-06-21T23:49:25+02:00 GitHub noreply@github.com 2023-06-21T23:49:25+02:00 cmake: revert CUDA arch default to 52, 61 if f16 (#1959)
fb98254f99d769fcbbf20966ef386abdb48ef601 049aa16b8c5c6d086246e4e6b9feb18de4fbd663 Rahul Vivek Nair 68507071+RahulVivekNair@users.noreply.github.com 2023-06-22T03:18:43+05:30 GitHub noreply@github.com 2023-06-21T23:48:43+02:00 Fix typo in README.md (#1961)
049aa16b8c5c6d086246e4e6b9feb18de4fbd663 2322ec223a21625dfe9bd73ee677444a98a24ac9 Georgi Gerganov ggerganov@gmail.com 2023-06-20T19:05:54+03:00 GitHub noreply@github.com 2023-06-20T19:05:54+03:00 readme : add link to p1
2322ec223a21625dfe9bd73ee677444a98a24ac9 aacdbd40562684665b6f7b8ba6695b7a2088bbb0 Xiake Sun xiake.sun@intel.com 2023-06-20T05:42:40-07:00 GitHub noreply@github.com 2023-06-20T15:42:40+03:00 Fix typo (#1949)
aacdbd40562684665b6f7b8ba6695b7a2088bbb0 20568fe60f00155fa25e92eb3a7f6b911d557967 Ettore Di Giacinto mudler@users.noreply.github.com 2023-06-20T03:24:39+02:00 GitHub noreply@github.com 2023-06-20T04:24:39+03:00 llama : fix params struct slignment (#1936)
20568fe60f00155fa25e92eb3a7f6b911d557967 18b35625c3c19c64b7818a12460ba5ddb006dfdc Henri Vasserman henv@hot.ee 2023-06-20T01:12:39+03:00 GitHub noreply@github.com 2023-06-20T01:12:39+03:00 [Fix] Reenable server embedding endpoint (#1937)
18b35625c3c19c64b7818a12460ba5ddb006dfdc ba4e85a8339b9dd7cdffad31838235f2fe45a8ea Georgi Gerganov ggerganov@gmail.com 2023-06-19T20:43:30+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-19T20:43:30+03:00 ggml : fix bug in LBFGS optimizer (found by ggml tests)
ba4e85a8339b9dd7cdffad31838235f2fe45a8ea 23fc5c219a9aebd57c8af3fac454062cc4622980 l3utterfly gc.pthzfoldr@gmail.com 2023-06-19T23:20:06+08:00 GitHub noreply@github.com 2023-06-19T18:20:06+03:00 llama : use aligned memory during ggml_init call from loading saved sessions (#1934)
23fc5c219a9aebd57c8af3fac454062cc4622980 cb40dfca694b5cb849837548fd69932117c78362 Georgi Gerganov ggerganov@gmail.com 2023-06-19T18:18:34+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-19T18:18:34+03:00 cmake : fix trailing whitespaces
cb40dfca694b5cb849837548fd69932117c78362 ca7c3f4da5d144d4cd1dd44903552e6ba49b8ec8 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-19T18:17:03+03:00 GitHub noreply@github.com 2023-06-19T18:17:03+03:00 llama : only use Q6_K for output weights if tensor size is multiple of 256 (#1932)
ca7c3f4da5d144d4cd1dd44903552e6ba49b8ec8 b97ca431db35ec96a339a721acb1219c1dd78bed Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-19T18:14:09+03:00 GitHub noreply@github.com 2023-06-19T18:14:09+03:00 cuda : faster k-quants on older GPUs (#1930)
b97ca431db35ec96a339a721acb1219c1dd78bed 1e3abfcef073e73c2b31e8570cb06c5cb2fd1f55 Georgi Gerganov ggerganov@gmail.com 2023-06-19T18:12:33+03:00 GitHub noreply@github.com 2023-06-19T18:12:33+03:00 ggml : sync latest ggml repo (#1924)
1e3abfcef073e73c2b31e8570cb06c5cb2fd1f55 16b9cd193965769089881bb8ec012fccca7b37b6 Howard Su howard0su@gmail.com 2023-06-19T23:10:37+08:00 GitHub noreply@github.com 2023-06-19T18:10:37+03:00 cmake : fix build shared ggml when CUDA is enabled (#1929)
16b9cd193965769089881bb8ec012fccca7b37b6 b24c3049d96557c24782e4d32feaae65f47277af Johannes Gäßler johannesg@5d6.de 2023-06-19T10:23:56+02:00 GitHub noreply@github.com 2023-06-19T10:23:56+02:00 Convert vector to f16 for dequantize mul mat vec (#1913)
b24c3049d96557c24782e4d32feaae65f47277af 0ede372a51fd8160688e01b587582666c14e94e5 Johannes Gäßler johannesg@5d6.de 2023-06-18T17:41:26+02:00 GitHub noreply@github.com 2023-06-18T17:41:26+02:00 Added tokens per second to info prints (#1928)
0ede372a51fd8160688e01b587582666c14e94e5 8596af427722775f0df4a7c90b9af067ba90d4ef Johannes Gäßler johannesg@5d6.de 2023-06-18T16:07:09+02:00 GitHub noreply@github.com 2023-06-18T16:07:09+02:00 Fixed incorrectly applying RMS norm twice (#1925)
8596af427722775f0df4a7c90b9af067ba90d4ef e1886cf4fe0d0f31661dda52a4a9f34bd9b9009a l3utterfly gc.pthzfoldr@gmail.com 2023-06-18T19:19:16+08:00 GitHub noreply@github.com 2023-06-18T14:19:16+03:00 ggml : fix bug in ggml_compute_forward_add_q_f32 (#1918)
e1886cf4fe0d0f31661dda52a4a9f34bd9b9009a 8ab8ba62eb27cc340be2edf3418e051b1d967416 Mike ytianhui2004@gmail.com 2023-06-18T16:28:26+08:00 GitHub noreply@github.com 2023-06-18T11:28:26+03:00 readme : update Android build instructions (#1922)
8ab8ba62eb27cc340be2edf3418e051b1d967416 90cc59d6ab1363a5c69c60c4b94db647d3a54a18 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-18T11:13:43+03:00 GitHub noreply@github.com 2023-06-18T11:13:43+03:00 llama : prevent usage of k-quants when tensor size is not a multiple of 256 (#1921)
90cc59d6ab1363a5c69c60c4b94db647d3a54a18 ce2c7d72e2d06988b5ddec6811ab923254542077 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-18T10:52:10+03:00 GitHub noreply@github.com 2023-06-18T10:52:10+03:00 examples : fix examples/metal (#1920)
ce2c7d72e2d06988b5ddec6811ab923254542077 57cd69460f736031a3fc54af1e97c03f80128478 Georgi Gerganov ggerganov@gmail.com 2023-06-18T09:09:47+03:00 GitHub noreply@github.com 2023-06-18T09:09:47+03:00 metal : handle buffers larger than device's maxBufferLength (#1826)
57cd69460f736031a3fc54af1e97c03f80128478 b2416493ab3ab21686d47c96669da6d6c6af08a4 Howard Su howard0su@gmail.com 2023-06-18T12:29:47+08:00 GitHub noreply@github.com 2023-06-18T07:29:47+03:00 cmake : add CUDA_ARCHITECTURES to new target ggml_static (#1917)
b2416493ab3ab21686d47c96669da6d6c6af08a4 4f9c43e3bd488b7561119785485e1155dba338d7 Georgi Gerganov ggerganov@gmail.com 2023-06-17T20:55:03+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-17T20:55:03+03:00 make : do not print help for simple example
4f9c43e3bd488b7561119785485e1155dba338d7 2c9380dd2f77e41149340f3ecb09764d793b16db Georgi Gerganov ggerganov@gmail.com 2023-06-17T20:24:11+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-17T20:24:11+03:00 minor : warning fixes
2c9380dd2f77e41149340f3ecb09764d793b16db 051e1b0e6a6e3aee7d989b47760980e6fda5861c Johannes Gäßler johannesg@5d6.de 2023-06-17T19:15:02+02:00 GitHub noreply@github.com 2023-06-17T19:15:02+02:00 Only one CUDA stream per device for async compute (#1898)
051e1b0e6a6e3aee7d989b47760980e6fda5861c 86c7571864ff331f8cdb9e092f3abeb123729a56 Georgi Gerganov ggerganov@gmail.com 2023-06-17T19:30:22+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-17T19:31:20+03:00 llama : fix kv_cache `n` init (close #1903)
86c7571864ff331f8cdb9e092f3abeb123729a56 3d59ec5935ea1d33e9d51060a8dd737169b9b89b DaniAndTheWeb 57776841+DaniAndTheWeb@users.noreply.github.com 2023-06-17T18:17:22+02:00 GitHub noreply@github.com 2023-06-17T19:17:22+03:00 make : update for latest Arch (#1701)
3d59ec5935ea1d33e9d51060a8dd737169b9b89b 0711a5f6dce7f04c2a791b14bc47f7d4cb545408 Howard Su howard0su@gmail.com 2023-06-17T23:46:15+08:00 GitHub noreply@github.com 2023-06-17T18:46:15+03:00 ggml : fix warnings under MSVC (#1908)
0711a5f6dce7f04c2a791b14bc47f7d4cb545408 fc45a81bc642b9ef33d9004f2b363d558438a6c9 Aaron Miller apage43@ninjawhale.com 2023-06-17T07:37:49-07:00 GitHub noreply@github.com 2023-06-17T17:37:49+03:00 metal : add norm, cpy f16->f16, alibi kernels (#1823)
fc45a81bc642b9ef33d9004f2b363d558438a6c9 794db3e7b982fee37e3995db9c3a216a57ff65e3 Faez Shakil faez.shakil@gmail.com 2023-06-17T17:13:05+05:00 GitHub noreply@github.com 2023-06-17T14:13:05+02:00 exposed modules so that they can be invoked by nix run github:ggerganov/llama.cpp#server etc (#1863)
794db3e7b982fee37e3995db9c3a216a57ff65e3 5ddf7ea1fb42bac21026de2f77e0f9c069b92234 Randall Fitzgerald randall@dasaku.net 2023-06-17T07:53:04-04:00 GitHub noreply@github.com 2023-06-17T14:53:04+03:00 Server Example Refactor and Improvements (#1570)
5ddf7ea1fb42bac21026de2f77e0f9c069b92234 bac19927c302737465a1deb14ac0943a221863e8 Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2023-06-17T12:32:48+02:00 GitHub noreply@github.com 2023-06-17T13:32:48+03:00 hooks : setting up flake8 and pre-commit hooks (#1681)
bac19927c302737465a1deb14ac0943a221863e8 b4c6f46f17b6e02f1cd55a81339e7e64f3aaa688 Gustavo Rocha Dias 91472747+gustrd@users.noreply.github.com 2023-06-17T06:01:06-03:00 GitHub noreply@github.com 2023-06-17T12:01:06+03:00 readme : alternative way to build for Android with CLBlast. (#1828)
b4c6f46f17b6e02f1cd55a81339e7e64f3aaa688 92f20d9942c86daeb78637bdad7296a572f4da28 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-06-17T01:49:42-06:00 GitHub noreply@github.com 2023-06-17T01:49:42-06:00 Allow cmake to build ggml as a library (#1896)
92f20d9942c86daeb78637bdad7296a572f4da28 d411968e990c37f51328849c96a743dd78f3c3dd David Yang davidyang6us@gmail.com 2023-06-17T14:51:54+08:00 GitHub noreply@github.com 2023-06-17T09:51:54+03:00 train : get raw text instead of page with html (#1905)
d411968e990c37f51328849c96a743dd78f3c3dd b41b4cad6f956b5f501db0711dd7007c32b5eee5 0cc4m picard12@live.de 2023-06-16T20:59:49+02:00 GitHub noreply@github.com 2023-06-16T21:59:49+03:00 opencl : support k-quants (#1836)
b41b4cad6f956b5f501db0711dd7007c32b5eee5 13fe9d2d84f30cab613c960bf66ac83916006694 SuperUserNameMan yoann@terminajones.com 2023-06-16T20:58:09+02:00 GitHub noreply@github.com 2023-06-16T21:58:09+03:00 examples : add "simple" (#1840)
13fe9d2d84f30cab613c960bf66ac83916006694 ac3b8869538c7fbdb48ff141d78c4dea091789f0 Zenix zenixls2@gmail.com 2023-06-17T03:53:04+09:00 GitHub noreply@github.com 2023-06-16T21:53:04+03:00 cmake : add auto detection of BLAS_INCLUDE_DIRS (#1886)
ac3b8869538c7fbdb48ff141d78c4dea091789f0 5b9ccaf104cc1054d4f8f17bc8a4b8dc949e5527 Johannes Gäßler johannesg@5d6.de 2023-06-16T20:25:51+02:00 GitHub noreply@github.com 2023-06-16T21:25:51+03:00 llama : fix embd when offloading non-repeating layers (#1891)
5b9ccaf104cc1054d4f8f17bc8a4b8dc949e5527 9cbf50c041a525d781c7764f493a5443924e4e38 FrankHB frankhb1989@gmail.com 2023-06-17T02:25:01+08:00 GitHub noreply@github.com 2023-06-16T21:25:01+03:00 Fixed possible macro redefinition (#1892)
9cbf50c041a525d781c7764f493a5443924e4e38 3d0112261042b356621e93db3fa4c6798a5d098f Borislav Stanimirov b.stanimirov@abv.bg 2023-06-16T21:23:53+03:00 GitHub noreply@github.com 2023-06-16T21:23:53+03:00 build : fix and ignore MSVC warnings (#1889)
3d0112261042b356621e93db3fa4c6798a5d098f 602c748863e15270d80d74aa2c3bf86ab8139e07 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-16T20:08:44+03:00 GitHub noreply@github.com 2023-06-16T20:08:44+03:00 CUDA : faster k-quant dot kernels (#1862)
602c748863e15270d80d74aa2c3bf86ab8139e07 a09f9195be39afb4b023b646c0a6ec8a86915174 Borislav Stanimirov b.stanimirov@abv.bg 2023-06-16T09:58:11+03:00 GitHub noreply@github.com 2023-06-16T09:58:11+03:00 gitignore : add several entries specific to Visual Studio (#1888)
a09f9195be39afb4b023b646c0a6ec8a86915174 bed92756172d4514b23aaf9744cf8e2dc892fc7b Johannes Gäßler johannesg@5d6.de 2023-06-15T21:49:08+02:00 GitHub noreply@github.com 2023-06-15T21:49:08+02:00 Fixed CUDA runtime version check (#1879)
bed92756172d4514b23aaf9744cf8e2dc892fc7b c36e81da62ebfe09a768201cc44fa8d712dd00ed Georgi Gerganov ggerganov@gmail.com 2023-06-15T21:56:50+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-15T21:56:50+03:00 cmake : remove whitespaces
c36e81da62ebfe09a768201cc44fa8d712dd00ed 3559433fecedf365e7aba2fe3d5f89d9abb817c1 yangli2 yangli2@gmail.com 2023-06-15T11:05:53-07:00 GitHub noreply@github.com 2023-06-15T21:05:53+03:00 examples : add chat-vicuna.sh (#1854)
3559433fecedf365e7aba2fe3d5f89d9abb817c1 69b34a0e80300bfb3e996983ac3ea075f5526675 Igor Okulist okigan@gmail.com 2023-06-15T12:51:26-05:00 GitHub noreply@github.com 2023-06-15T20:51:26+03:00 cmake : set include path for OpenBlas (#1830)
69b34a0e80300bfb3e996983ac3ea075f5526675 cf267d1c71a781700698f8518e903239c3bcc929 Frederik Vogel Schaltfehler@users.noreply.github.com 2023-06-16T02:47:04+09:00 GitHub noreply@github.com 2023-06-15T20:47:04+03:00 swift : Package compile breaks due to ggml-metal.metal (#1831)
cf267d1c71a781700698f8518e903239c3bcc929 9dda13e5e1f70bdfc25fbc0f0378f27c8b67e983 daboe01 daboe01@googlemail.com 2023-06-15T19:42:48+02:00 GitHub noreply@github.com 2023-06-15T20:42:48+03:00 make : add train-text-from-scratch (#1850)
9dda13e5e1f70bdfc25fbc0f0378f27c8b67e983 37e257c48e350cf03c353c10d31e777f8d00123d Srinivas Billa nivibilla@gmail.com 2023-06-15T18:36:38+01:00 GitHub noreply@github.com 2023-06-15T20:36:38+03:00 readme : server compile flag (#1874)
37e257c48e350cf03c353c10d31e777f8d00123d 64cc19b4fe3df03bc20e520aa111c30cff3a655e sandyiscool sandyiscool@gmail.com 2023-06-15T23:06:06+05:30 GitHub noreply@github.com 2023-06-15T20:36:06+03:00 make : clean *.so files (#1857)
64cc19b4fe3df03bc20e520aa111c30cff3a655e 4bfcc855abdb2c9fcc3c5a84747974521909fa41 Howard Su howard0su@gmail.com 2023-06-16T01:29:59+08:00 GitHub noreply@github.com 2023-06-15T19:29:59+02:00 Fix the validation of main device (#1872)
4bfcc855abdb2c9fcc3c5a84747974521909fa41 6b8312e7979b852f6b6ac9d29cd51fda16c17948 Georgi Gerganov ggerganov@gmail.com 2023-06-15T20:29:48+03:00 GitHub noreply@github.com 2023-06-15T20:29:48+03:00 metal : parallel command buffer encoding (#1860)
6b8312e7979b852f6b6ac9d29cd51fda16c17948 254a7a7a5ff4c874ff8488f1f5cbdd7e9c89d682 Johannes Gäßler johannesg@5d6.de 2023-06-15T19:06:46+02:00 GitHub noreply@github.com 2023-06-15T19:06:46+02:00 Better error when using both LoRA + GPU layers (#1861)
254a7a7a5ff4c874ff8488f1f5cbdd7e9c89d682 92549202659fc23ba9fec5e688227d0da9b06b40 Johannes Gäßler johannesg@5d6.de 2023-06-14T19:47:19+02:00 GitHub noreply@github.com 2023-06-14T19:47:19+02:00 CUDA full GPU acceleration, KV cache in VRAM (#1827)
92549202659fc23ba9fec5e688227d0da9b06b40 e32089b2c20b1b87b22912f4a8b93fe01647d5b9 0xspringtime 110655352+0xspringtime@users.noreply.github.com 2023-06-13T15:37:54-04:00 GitHub noreply@github.com 2023-06-13T22:37:54+03:00 baby-llama : fix operator!= (#1821)
e32089b2c20b1b87b22912f4a8b93fe01647d5b9 2347e45e7bdb09c9a7d74b2c0bc86c2b65f0c343 xaedes xaedes@gmail.com 2023-06-13T21:04:40+02:00 GitHub noreply@github.com 2023-06-13T22:04:40+03:00 train : improved training-from-scratch example (#1652)
2347e45e7bdb09c9a7d74b2c0bc86c2b65f0c343 74d4cfa3438cb58bd177eed30014e6588694aaa8 Georgi Gerganov ggerganov@gmail.com 2023-06-13T20:20:07+03:00 GitHub noreply@github.com 2023-06-13T20:20:07+03:00 llama : do a warm-up eval at start for better timings (#1824)
74d4cfa3438cb58bd177eed30014e6588694aaa8 74a6d922f12ccfe16b0c265f43be8978c6f25e98 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-06-13T04:23:23-06:00 GitHub noreply@github.com 2023-06-13T04:23:23-06:00 Allow "quantizing" to f16 and f32 (#1787)
74a6d922f12ccfe16b0c265f43be8978c6f25e98 e4caa8da59c1c97dc23fa336f4d726984a20560f Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-12T22:39:21+03:00 GitHub noreply@github.com 2023-06-12T22:39:21+03:00 Metal implementation for all k_quants (#1807)
e4caa8da59c1c97dc23fa336f4d726984a20560f 58970a4c39124a647ac2a640d9e178ea6c961e65 slaren slarengh@gmail.com 2023-06-12T19:12:47+02:00 GitHub noreply@github.com 2023-06-12T20:12:47+03:00 ci : run when changing only the CUDA sources (#1800)
58970a4c39124a647ac2a640d9e178ea6c961e65 8c0a10e64dbf60fd9946c0cd5e6f59690800b123 Howard Su howard0su@gmail.com 2023-06-12T20:44:16+08:00 GitHub noreply@github.com 2023-06-12T14:44:16+02:00 Leverage mmap for offloading tensors to GPU (#1597)
8c0a10e64dbf60fd9946c0cd5e6f59690800b123 fa84c4b3e80199a5683438f062009c031a06c4fa Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-12T14:31:36+03:00 GitHub noreply@github.com 2023-06-12T14:31:36+03:00 metal : fix failure to load model (#1817)
fa84c4b3e80199a5683438f062009c031a06c4fa 12b063f0ecf280e98028e444fc492ee6222cdcdc Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-06-11T08:19:17-06:00 GitHub noreply@github.com 2023-06-11T08:19:17-06:00 Fix issue where interactive mode crashes when input exceeds ctx size (#1789)
12b063f0ecf280e98028e444fc492ee6222cdcdc 31d2b5f4a4bae081e59b36ab37c6ff6f5b5940ad Kyle Liang liangmanlai@gmail.com 2023-06-11T21:20:52+08:00 GitHub noreply@github.com 2023-06-11T15:20:52+02:00 Fixed WSL cuda's OOM error (#1594)
31d2b5f4a4bae081e59b36ab37c6ff6f5b5940ad 4de0334f5cabf4696eced2e5d6e279fdfaa6c0f2 Ryan Landay rlanday@gmail.com 2023-06-11T17:38:53+08:00 GitHub noreply@github.com 2023-06-11T12:38:53+03:00 Update SHA256SUMS with current hashes for models quantized using q4_0 (#1798)
4de0334f5cabf4696eced2e5d6e279fdfaa6c0f2 3f1223155a462477ac933474ebc4eab0ce3ca264 Georgi Gerganov ggerganov@gmail.com 2023-06-10T22:56:53+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-10T22:56:53+03:00 cmake : fix Metal build (close #1791)
3f1223155a462477ac933474ebc4eab0ce3ca264 303f5809f1b4ec49823dbe70cacd2124ec1d0df0 Artyom Lebedev vagran.ast@gmail.com 2023-06-10T22:51:36+03:00 GitHub noreply@github.com 2023-06-10T22:51:36+03:00 k-quants : GCC12 compilation fix (#1792)
303f5809f1b4ec49823dbe70cacd2124ec1d0df0 059e99066d95d73d1ca26c3375d47c0e35596229 Andrei abetlen@gmail.com 2023-06-10T10:47:34-04:00 GitHub noreply@github.com 2023-06-10T17:47:34+03:00 metal : fix issue with ggml-metal.metal path. Closes #1769 (#1782)
059e99066d95d73d1ca26c3375d47c0e35596229 17c10acfb44ecb7af25e37fb67b9501cbc0034d2 Aisuko urakiny@gmail.com 2023-06-11T00:08:11+10:00 GitHub noreply@github.com 2023-06-10T17:08:11+03:00 doc : fix wrong address of BLIS.md (#1772)
17c10acfb44ecb7af25e37fb67b9501cbc0034d2 e9b66ee9829039d4ab54550d6222e42a0b31e52a Georgi Gerganov ggerganov@gmail.com 2023-06-10T12:06:45+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-10T12:08:15+03:00 ggml : force no_alloc == false when creating opt tensors (close #1699)
e9b66ee9829039d4ab54550d6222e42a0b31e52a 4f0154b0bad775ac4651bf73b5c216eb43c45cdc Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-10T11:28:11+03:00 GitHub noreply@github.com 2023-06-10T11:28:11+03:00 metal : add Q4_1 implementation (#1785)
4f0154b0bad775ac4651bf73b5c216eb43c45cdc ef3171d16241c18581d4d08374f0b9e396ade6b7 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-06-10T01:59:17-06:00 GitHub noreply@github.com 2023-06-10T10:59:17+03:00 llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691)
ef3171d16241c18581d4d08374f0b9e396ade6b7 555275a693843273759230547001f9ae07fb537e Xingchen Song(宋星辰) xingchensong1996@163.com 2023-06-10T15:49:40+08:00 GitHub noreply@github.com 2023-06-10T10:49:40+03:00 ggml : workaround for missing _mm256_setr_m128i in GCC < 8 (#1638)
555275a693843273759230547001f9ae07fb537e 98ed16557432d7a5179c57eddcc3a08a7ae6d54d rankaiyx rankaiyx@rankaiyx.com 2023-06-10T14:41:59+08:00 GitHub noreply@github.com 2023-06-10T09:41:59+03:00 make : add SSSE3 compilation use case (#1659)
98ed16557432d7a5179c57eddcc3a08a7ae6d54d ae9663f1887513e152839e91f61c513075a19422 Robert Sung-wook Shin edp1096@users.noreply.github.com 2023-06-10T01:24:40+09:00 GitHub noreply@github.com 2023-06-09T18:24:40+02:00 OpenCL: Add release memory (#1741)
ae9663f1887513e152839e91f61c513075a19422 b33dee282f5d8032b5f780152732dc45cbf2d349 Johannes Gäßler johannesg@5d6.de 2023-06-09T13:58:15+02:00 GitHub noreply@github.com 2023-06-09T13:58:15+02:00 Windows nvcc workaround (#1753)
b33dee282f5d8032b5f780152732dc45cbf2d349 92f44ff7f778ef1b94028b2ba6d39943b5ca0ada Georgi Gerganov ggerganov@gmail.com 2023-06-09T11:11:04+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-09T11:11:04+03:00 metal : fix build "tanhf" -> "tanh"
92f44ff7f778ef1b94028b2ba6d39943b5ca0ada 245fc3c37da5ac5963f9f11a9f4f2ac08d96afc6 AT manyoso@users.noreply.github.com 2023-06-09T04:00:51-04:00 GitHub noreply@github.com 2023-06-09T11:00:51+03:00 metal : add GELU implementation (#1770)
245fc3c37da5ac5963f9f11a9f4f2ac08d96afc6 72ff5282bf0388c60821f504c4c8cc2b1f491aa6 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-09T10:39:59+03:00 GitHub noreply@github.com 2023-06-09T10:39:59+03:00 metal : faster q4_0 (#1775)
72ff5282bf0388c60821f504c4c8cc2b1f491aa6 0bf7cf1b296fc9fca05411b37afdf08a531487d2 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-08T22:28:21+03:00 GitHub noreply@github.com 2023-06-08T22:28:21+03:00 metal : add Q2_K implementation (#1762)
0bf7cf1b296fc9fca05411b37afdf08a531487d2 8432d4d9f716b25133e3ed671d91e21f6f3be867 Georgi Gerganov ggerganov@gmail.com 2023-06-08T20:48:14+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-08T20:48:14+03:00 Revert "ggml : load data into int8x16x4_t using vld4q_s8 on arm64 (#1738)"
8432d4d9f716b25133e3ed671d91e21f6f3be867 0f291e1f65c1d68201e71ce99c89562a36686b6d le.chang cljs118@126.com 2023-06-09T00:47:56+08:00 GitHub noreply@github.com 2023-06-08T19:47:56+03:00 ggml : load data into int8x16x4_t using vld4q_s8 on arm64 (#1738)
0f291e1f65c1d68201e71ce99c89562a36686b6d 8fc8179919a11738910db07a800f2b176f8adf09 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-08T19:46:22+03:00 GitHub noreply@github.com 2023-06-08T19:46:22+03:00 metal : Q6_K implementation (#1752)
8fc8179919a11738910db07a800f2b176f8adf09 b50b570ed9d699d3d126d72fc02de92926bcd937 qingfengfenga 41416092+qingfengfenga@users.noreply.github.com 2023-06-08T15:58:53+08:00 GitHub noreply@github.com 2023-06-08T00:58:53-07:00 Add llama.cpp docker support for non-latin languages (#1673)
b50b570ed9d699d3d126d72fc02de92926bcd937 53aba3f393f2e02a78ddaba2e934893a8bbf3246 Steven Roussey sroussey@gmail.com 2023-06-08T00:12:28-07:00 GitHub noreply@github.com 2023-06-08T10:12:28+03:00 ggml : fix fprintf warnings (#1720)
53aba3f393f2e02a78ddaba2e934893a8bbf3246 4161bdc04debb70bf5f275492b4d89fd9330087c Georgi Gerganov ggerganov@gmail.com 2023-06-08T10:09:08+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-08T10:09:08+03:00 clang-tidy : restore dot file from accidental deletion
4161bdc04debb70bf5f275492b4d89fd9330087c 0035858273ebe0694926bf4414d279f3e1cd109d Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-08T10:08:23+03:00 GitHub noreply@github.com 2023-06-08T10:08:23+03:00 metal : add Q4_K implementation (#1733)
0035858273ebe0694926bf4414d279f3e1cd109d 5c64a0952ee58b2d742ee84e8e3d43cce5d366db johnson442 56517414+johnson442@users.noreply.github.com 2023-06-08T08:02:48+01:00 GitHub noreply@github.com 2023-06-08T10:02:48+03:00 k-quants : add missing compile definition to CMakeLists (#1748)
5c64a0952ee58b2d742ee84e8e3d43cce5d366db 5b57a5b72676540b6a45a3f527126299969ad241 Georgi Gerganov ggerganov@gmail.com 2023-06-07T10:59:52+03:00 GitHub noreply@github.com 2023-06-07T10:59:52+03:00 k-quants : allow to optionally disable at compile time (#1734)
5b57a5b72676540b6a45a3f527126299969ad241 4dc62c545df0af60635d579e9e4dd91bc5afff51 jacobi petrucciani 8117202+jpetrucciani@users.noreply.github.com 2023-06-07T00:15:31-04:00 GitHub noreply@github.com 2023-06-07T07:15:31+03:00 flake : update to support metal on m1/m2 (#1724)
4dc62c545df0af60635d579e9e4dd91bc5afff51 35a84916fb029905c44746127026079268216e7a Georgi Gerganov ggerganov@gmail.com 2023-06-07T07:15:08+03:00 GitHub noreply@github.com 2023-06-07T07:15:08+03:00 readme : add June roadmap
35a84916fb029905c44746127026079268216e7a 2d7bf110edd8c49209401a16132052cba706ffd0 Willy Tarreau w@1wt.eu 2023-06-07T04:10:17+02:00 GitHub noreply@github.com 2023-06-06T22:10:17-04:00 main: add the possibility to open the prompt cache read-only (#1640)
2d7bf110edd8c49209401a16132052cba706ffd0 2a4e41a086ce80da68c402457c75c77e52dcc698 Georgi Gerganov ggerganov@gmail.com 2023-06-06T22:54:39+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-06T22:54:39+03:00 llama : fix vram_scratch var
2a4e41a086ce80da68c402457c75c77e52dcc698 17366df842e358768c0df7024484fffecfc7865b Georgi Gerganov ggerganov@gmail.com 2023-06-06T22:41:53+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-06T22:41:53+03:00 llama : fix compile warnings
17366df842e358768c0df7024484fffecfc7865b 44f906e8537fcec965e312d621c80556d6aa9bec Johannes Gäßler johannesg@5d6.de 2023-06-06T21:33:23+02:00 GitHub noreply@github.com 2023-06-06T21:33:23+02:00 Multi GPU support, CUDA refactor, CUDA scratch buffer (#1703)
44f906e8537fcec965e312d621c80556d6aa9bec d5b111f53d14972669eb52055f9df2567663ad8b Georgi Gerganov ggerganov@gmail.com 2023-06-06T20:16:57+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-06T20:21:56+03:00 metal : add f16 support
d5b111f53d14972669eb52055f9df2567663ad8b 2d43387dafe9c60f15f57aa23ee0b37864b98b32 LostRuins 39025047+LostRuins@users.noreply.github.com 2023-06-07T01:00:01+08:00 GitHub noreply@github.com 2023-06-06T19:00:01+02:00 Clblast fixes + enhancements to save VRAM and offload more layers (#1675)
2d43387dafe9c60f15f57aa23ee0b37864b98b32 7ad7750c5c9f6bcea73a1895ffc57e7d21f2ab95 Georgi Gerganov ggerganov@gmail.com 2023-06-06T10:18:03+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-06T10:18:03+03:00 ggml : fix builds, add ggml-quants-k.o (close #1712, close #1710)
7ad7750c5c9f6bcea73a1895ffc57e7d21f2ab95 7a74dee6b4e0e80862191141c0037abe28967d5c Georgi Gerganov ggerganov@gmail.com 2023-06-06T09:55:10+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-06T09:55:25+03:00 gitignore : add .clang-tidy
7a74dee6b4e0e80862191141c0037abe28967d5c 590250f7a9847bc9c83aa063dbaac8fa0fea27c8 Georgi Gerganov ggerganov@gmail.com 2023-06-06T09:39:38+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-06T09:39:38+03:00 llama : temporary disable Q6_K output quantization (#1711)
590250f7a9847bc9c83aa063dbaac8fa0fea27c8 f4c55d3bd7e124b101bc974cbbf0e0dbbc32d5a3 Spencer Sutton spencersutton@users.noreply.github.com 2023-06-05T23:28:17-04:00 GitHub noreply@github.com 2023-06-06T06:28:17+03:00 metal : add checks for buffer size (#1706)
f4c55d3bd7e124b101bc974cbbf0e0dbbc32d5a3 f1465624c2cbc8ee65b566e3d87f2af27796d4c4 Yuval Peled 31162840+Yuval-Peled@users.noreply.github.com 2023-06-05T23:32:36+03:00 GitHub noreply@github.com 2023-06-05T23:32:36+03:00 docs : add performance troubleshoot + example benchmark documentation (#1674)
f1465624c2cbc8ee65b566e3d87f2af27796d4c4 c2df36d60dc0ff1576541b965d751eadbacbeada Foul-Tarnished 107711110+Foul-Tarnished@users.noreply.github.com 2023-06-05T22:28:37+02:00 GitHub noreply@github.com 2023-06-05T23:28:37+03:00 readme : fix typo (#1700)
c2df36d60dc0ff1576541b965d751eadbacbeada 9d0693bce38013364b1042568d9083353bfff48f mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2023-06-05T22:24:29+02:00 GitHub noreply@github.com 2023-06-05T23:24:29+03:00 llama : consistently catch and throw only exceptions deriving from std::exception (#1599)
9d0693bce38013364b1042568d9083353bfff48f efe05076323f5c6bafece109e21cce046f5e4b07 kiltyj kiltyj@gmail.com 2023-06-05T13:24:04-07:00 GitHub noreply@github.com 2023-06-05T23:24:04+03:00 metal : use shared buffers between CPU and GPU (#1696)
efe05076323f5c6bafece109e21cce046f5e4b07 e7fe66e670537990ccc075cce9286df88bba052a grahameth 96447521+grahameth@users.noreply.github.com 2023-06-05T22:11:49+02:00 GitHub noreply@github.com 2023-06-05T23:11:49+03:00 ggml : fix internal overflow in ggml_time_us on Windows (#1702)
e7fe66e670537990ccc075cce9286df88bba052a 99009e72f8072fa552eb02efee436be596c71cdd Georgi Gerganov ggerganov@gmail.com 2023-06-05T23:05:05+03:00 GitHub noreply@github.com 2023-06-05T23:05:05+03:00 ci : disable auto tidy (#1705)
99009e72f8072fa552eb02efee436be596c71cdd 5220a991a5e92bddad9542267ab445a2c033681c Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-06-05T22:56:18+03:00 GitHub noreply@github.com 2023-06-05T22:56:18+03:00 ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684)
5220a991a5e92bddad9542267ab445a2c033681c d1f563a743a83dabc11e125d4a7d64189c16498c Henri Vasserman henv@hot.ee 2023-06-05T13:43:08+03:00 GitHub noreply@github.com 2023-06-05T13:43:08+03:00 Increase 3B scratch buffers. (#1698)
d1f563a743a83dabc11e125d4a7d64189c16498c 827f5eda91e5b7299848ee2c7179d873bdee0f7b Georgi Gerganov ggerganov@gmail.com 2023-06-05T10:19:03+03:00 Georgi Gerganov ggerganov@gmail.com 2023-06-05T10:19:03+03:00 llama : fix Metal KV cache sync (close #1695)
827f5eda91e5b7299848ee2c7179d873bdee0f7b ecb217db4fcfa3880300ad08531a5fb6bb142d45 Georgi Gerganov ggerganov@gmail.com 2023-06-04T23:38:19+03:00 GitHub noreply@github.com 2023-06-04T23:38:19+03:00 readme : update hot topics
ecb217db4fcfa3880300ad08531a5fb6bb142d45 dcb2ed48268e421baf25adc00d602dad0f415564 Georgi Gerganov ggerganov@gmail.com 2023-06-04T23:34:30+03:00 GitHub noreply@github.com 2023-06-04T23:34:30+03:00 llama : Metal inference (#1642)
dcb2ed48268e421baf25adc00d602dad0f415564 d8bd0013e8768aaa3dc9cfc1ff01499419d5348e 0cc4m picard12@live.de 2023-06-04T08:12:05+02:00 GitHub noreply@github.com 2023-06-04T08:12:05+02:00 OpenCL: Fix duplication of layers in VRAM and RAM, add GPU mul kernel (#1653)
d8bd0013e8768aaa3dc9cfc1ff01499419d5348e b5c85468a3eadf424420af5bf11c2353ff828cda Henri Vasserman henv@hot.ee 2023-06-03T16:35:20+03:00 GitHub noreply@github.com 2023-06-03T16:35:20+03:00 Add info about CUDA_VISIBLE_DEVICES (#1682)
b5c85468a3eadf424420af5bf11c2353ff828cda 136476e898fb96c302b0829ee3e79267ae12660f Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2023-06-03T14:11:53+02:00 GitHub noreply@github.com 2023-06-03T15:11:53+03:00 Docker: change to calling convert.py (#1641)
136476e898fb96c302b0829ee3e79267ae12660f ffb06a345e3a9e30d39aaa5b46a23201a74be6de Evan Jones evan.q.jones@gmail.com 2023-06-03T07:28:45-04:00 GitHub noreply@github.com 2023-06-03T07:28:45-04:00 Fix prompt cache saving and chat-persistent rollover (#1678)
ffb06a345e3a9e30d39aaa5b46a23201a74be6de 7552ac586380f202b75b18aa216ecfefbd438d94 Henri Vasserman henv@hot.ee 2023-05-30T21:24:22+03:00 GitHub noreply@github.com 2023-05-30T21:24:22+03:00 OpenLLaMA 3B support (#1588)
7552ac586380f202b75b18aa216ecfefbd438d94 5d1830b99dfd85bb6279adb4dd94aa444afd5b5e Georgi Gerganov ggerganov@gmail.com 2023-05-29T19:31:44+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-29T19:31:44+03:00 ggml : sync cgraph import / export API
5d1830b99dfd85bb6279adb4dd94aa444afd5b5e 248367605ead6fb7c36d2bfb1ebd8f00a23f7c71 Georgi Gerganov ggerganov@gmail.com 2023-05-29T19:30:49+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-29T19:30:49+03:00 ggml : fix bug in ggml_alibi
248367605ead6fb7c36d2bfb1ebd8f00a23f7c71 0e730dd23b0fb5f93dba574e0a48d9a69dc5dbae DannyDaemonic DannyDaemonic@gmail.com 2023-05-29T05:13:40-07:00 GitHub noreply@github.com 2023-05-29T05:13:40-07:00 Work around for recalculating logits in cached prompts (Fixes #1585) (#1609)
0e730dd23b0fb5f93dba574e0a48d9a69dc5dbae 3b126f654fceb4f5f195a1c2e825bb18e101188c Jiří Podivín 66251151+jpodivin@users.noreply.github.com 2023-05-29T06:45:50+02:00 GitHub noreply@github.com 2023-05-28T21:45:50-07:00 Adding git in container package dependencies (#1621)
3b126f654fceb4f5f195a1c2e825bb18e101188c 1b78ed20818b72306edc7208b9bfb69a1a0d3297 Johannes Gäßler johannesg@5d6.de 2023-05-28T21:01:02+02:00 GitHub noreply@github.com 2023-05-28T21:01:02+02:00 LLAMA_DEBUG adds debug symbols (#1617)
1b78ed20818b72306edc7208b9bfb69a1a0d3297 337aea11390221bc925e4acb1f603f1649af2735 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-05-28T11:48:57-06:00 GitHub noreply@github.com 2023-05-28T11:48:57-06:00 Only show -ngl option when relevant + other doc/arg handling updates (#1625)
337aea11390221bc925e4acb1f603f1649af2735 bb051d9723d628414b9e929e5264e23262a2f1b2 Vladimir Zorin vladimir@deviant.guru 2023-05-28T20:14:24+03:00 GitHub noreply@github.com 2023-05-28T20:14:24+03:00 examples : add --alias option to gpt_params to set use friendly model name (#1614)
bb051d9723d628414b9e929e5264e23262a2f1b2 ca74884f6625b15ef69832f07fc60fe00db5f90c Howard Su howard0su@gmail.com 2023-05-29T01:13:36+08:00 GitHub noreply@github.com 2023-05-28T20:13:36+03:00 opencl : no need to allocate cl_mem on heap (#1612)
ca74884f6625b15ef69832f07fc60fe00db5f90c a6704643b62243bc4b6bbcd727d63d44e01a1002 Howard Su howard0su@gmail.com 2023-05-29T01:09:56+08:00 GitHub noreply@github.com 2023-05-28T20:09:56+03:00 opencl : use strstr to check if fp16 supported (#1611)
a6704643b62243bc4b6bbcd727d63d44e01a1002 0df7d63e5ba0ab8856476e121a03b985d6f15c9d apcameron 37645737+apcameron@users.noreply.github.com 2023-05-27T21:03:25+01:00 GitHub noreply@github.com 2023-05-27T23:03:25+03:00 ggml : add support for the RISCV architecture (#1616)
0df7d63e5ba0ab8856476e121a03b985d6f15c9d 97c9b77c4fc5e2283755c4418759cfc5fc73ad05 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-05-27T11:04:14-06:00 GitHub noreply@github.com 2023-05-27T11:04:14-06:00 Include server in releases + other build system cleanups (#1610)
97c9b77c4fc5e2283755c4418759cfc5fc73ad05 0ecb1bbbeb16e36a2ea7a5ce525c6c59ef74312b Henri Vasserman henv@hot.ee 2023-05-27T18:47:55+03:00 GitHub noreply@github.com 2023-05-27T18:47:55+03:00 Add documentation about CLBlast (#1604)
0ecb1bbbeb16e36a2ea7a5ce525c6c59ef74312b 93618031c7ccdd949d976370f24953d261048575 Henri Vasserman henv@hot.ee 2023-05-27T17:24:06+03:00 GitHub noreply@github.com 2023-05-27T17:24:06+03:00 [CI] Fix openblas (#1613)
93618031c7ccdd949d976370f24953d261048575 83c54e6da58f1970556741b143bd26e30b1f46af Georgi Gerganov ggerganov@gmail.com 2023-05-27T16:19:56+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-27T16:19:56+03:00 ggml : add ggml_tensor_overhead()
83c54e6da58f1970556741b143bd26e30b1f46af bdbda1b17afb78e8613d03c8210a57fac632397b Henri Vasserman henv@hot.ee 2023-05-27T15:18:25+03:00 GitHub noreply@github.com 2023-05-27T14:18:25+02:00 [CI] CLBlast: Fix directory name (#1606)
bdbda1b17afb78e8613d03c8210a57fac632397b 66874d4fbcc7866377246efbcee938e8cc9c7d76 Georgi Gerganov ggerganov@gmail.com 2023-05-27T12:22:05+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-27T12:23:16+03:00 ggml : sync ggml core (minor additions, e.g. ggml_get_tensor_by_name())
66874d4fbcc7866377246efbcee938e8cc9c7d76 1fcdcc28b119a6608774d52de905931bd5f8a43d Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-05-25T20:18:01-06:00 GitHub noreply@github.com 2023-05-25T20:18:01-06:00 Some improvements to loading the session with --prompt-cache (#1550)
1fcdcc28b119a6608774d52de905931bd5f8a43d ac7876ac20124a15a44fd6317721ff1aa2538806 Johannes Gäßler johannesg@5d6.de 2023-05-25T23:07:29+02:00 GitHub noreply@github.com 2023-05-26T00:07:29+03:00 cuda : performance optimizations (#1530)
ac7876ac20124a15a44fd6317721ff1aa2538806 c31bbe934b9666af42f32ce12d32cae9160e5dc4 Henri Vasserman henv@hot.ee 2023-05-24T10:30:09+03:00 GitHub noreply@github.com 2023-05-24T10:30:09+03:00 Update CLBlast to 1.6.0 (#1580)
c31bbe934b9666af42f32ce12d32cae9160e5dc4 1359b6aba55d5b0410f6adaa0aa2e49bbfd01d84 Evan Jones evan.q.jones@gmail.com 2023-05-24T02:24:01-04:00 GitHub noreply@github.com 2023-05-24T09:24:01+03:00 readme : add docs for chat-persistent.sh (#1568)
1359b6aba55d5b0410f6adaa0aa2e49bbfd01d84 7d873811f31d4d8c909015c946a862c0089cda7d Senemu 10880819+Senemu@users.noreply.github.com 2023-05-24T06:16:22Z GitHub noreply@github.com 2023-05-24T09:16:22+03:00 chat-persistent.sh : use bracket expressions in grep (#1564)
7d873811f31d4d8c909015c946a862c0089cda7d 2e6cd4b02549e343bef3768e6b946f999c82e823 Maarten ter Huurne maarten@treewalker.org 2023-05-23T18:01:15+02:00 GitHub noreply@github.com 2023-05-23T19:01:15+03:00 Fix handling of "invalid property" when creating OpenCL command queue (#1565)
2e6cd4b02549e343bef3768e6b946f999c82e823 7e4ea5beff567f53be92f75f9089e6f11fa5dabd 0cc4m picard12@live.de 2023-05-22T23:33:24+02:00 GitHub noreply@github.com 2023-05-23T00:33:24+03:00 OpenCL Token Generation Acceleration (#1459)
7e4ea5beff567f53be92f75f9089e6f11fa5dabd 7780e4f479dc5af106287c164b8e186cd9b6215c Steward Garcia 57494570+FSSRepo@users.noreply.github.com 2023-05-21T11:51:18-06:00 GitHub noreply@github.com 2023-05-21T20:51:18+03:00 examples : add server example with REST API (#1443)
7780e4f479dc5af106287c164b8e186cd9b6215c 265db9834e761b7c8210ea1888117efcd3262f52 Stefan Sydow stefan@sydow.email 2023-05-21T16:03:44+02:00 GitHub noreply@github.com 2023-05-21T17:03:44+03:00 make : .PHONY clean (#1553)
265db9834e761b7c8210ea1888117efcd3262f52 fab49c685e09d95942de34e3eadd72f880de21d5 Georgi Gerganov ggerganov@gmail.com 2023-05-21T11:56:23+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-21T11:56:23+03:00 ggml : output 3d sizes in ggml_graph_dump_dot()
fab49c685e09d95942de34e3eadd72f880de21d5 b8ee340abe4a46143eb8cc4a135b976856c9136c Georgi Gerganov ggerganov@gmail.com 2023-05-20T20:00:41+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-20T20:00:41+03:00 ggml : update WASM SIMD
b8ee340abe4a46143eb8cc4a135b976856c9136c 9ecb30f9594f222d8318fb1e803a4c363b0c39e5 Zenix zenixls2@gmail.com 2023-05-20T23:58:31+09:00 GitHub noreply@github.com 2023-05-20T17:58:31+03:00 feature : support blis and other blas implementation (#1536)
9ecb30f9594f222d8318fb1e803a4c363b0c39e5 29cf5596fe0c37213f9b74e80d8f631193a93f0f Henri Vasserman henv@hot.ee 2023-05-20T17:57:39+03:00 GitHub noreply@github.com 2023-05-20T17:57:39+03:00 OpenCL: Fixes for older devices. (#1435)
29cf5596fe0c37213f9b74e80d8f631193a93f0f 3de84b26066d95068409c1dc79bcc41c1eea2a03 Juuso Alasuutari juuso.alasuutari@gmail.com 2023-05-20T15:58:15+03:00 GitHub noreply@github.com 2023-05-20T15:58:15+03:00 llama : define magic numbers as integer constants (#1518) (#1520)
3de84b26066d95068409c1dc79bcc41c1eea2a03 affc76edfdefa7b326f526e463cc65ff13fcfb92 Georgi Gerganov ggerganov@gmail.com 2023-05-20T15:34:45+03:00 GitHub noreply@github.com 2023-05-20T15:34:45+03:00 ggml : add ggml_clamp() (#1539)
affc76edfdefa7b326f526e463cc65ff13fcfb92 ea600071cb005267e9e8f2629c1e406dd5fde083 Johannes Gäßler johannesg@5d6.de 2023-05-20T14:19:28+02:00 GitHub noreply@github.com 2023-05-20T15:19:28+03:00 cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483)
ea600071cb005267e9e8f2629c1e406dd5fde083 07e9ace0f9da424d82e75df969642522880feb92 Georgi Gerganov ggerganov@gmail.com 2023-05-20T12:03:48+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-20T12:03:48+03:00 Revert "feature : add blis and other BLAS implementation support (#1502)"
07e9ace0f9da424d82e75df969642522880feb92 ec2e10c4443209da56b431b24dd0845b60e757fb Zenix zenixls2@gmail.com 2023-05-20T18:02:48+09:00 GitHub noreply@github.com 2023-05-20T12:02:48+03:00 feature : add blis and other BLAS implementation support (#1502)
ec2e10c4443209da56b431b24dd0845b60e757fb d2c59b8ba498ab01e65203dde6fe95236d20f6e7 Georgi Gerganov ggerganov@gmail.com 2023-05-20T11:06:11+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-20T11:06:37+03:00 llama : add llama_init_backend() API (close #1527)
d2c59b8ba498ab01e65203dde6fe95236d20f6e7 503db28849d8641d66244385e7e9649608a2e4d0 DannyDaemonic DannyDaemonic@gmail.com 2023-05-20T00:40:02-07:00 GitHub noreply@github.com 2023-05-20T00:40:02-07:00 Fix for mingw (#1462)
503db28849d8641d66244385e7e9649608a2e4d0 8a203f9fa1b24e010be8f35ebbbd6786293684cb Maxime 672982+maximegmd@users.noreply.github.com 2023-05-20T09:22:37+02:00 GitHub noreply@github.com 2023-05-20T10:22:37+03:00 llama : fix name shadowing and C4146 (#1526)
8a203f9fa1b24e010be8f35ebbbd6786293684cb 4fd3e29297e3246a7be291932c115636fadb0f52 Georgi Gerganov ggerganov@gmail.com 2023-05-20T10:14:31+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-20T10:14:43+03:00 llama : fix compile warnings in llama_set_state_data()
4fd3e29297e3246a7be291932c115636fadb0f52 2d5db48371052087a83974abda3767d1aedec598 Georgi Gerganov ggerganov@gmail.com 2023-05-20T10:13:19+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-20T10:13:19+03:00 ggml : fix scalar implementation of Q4_1 dot
2d5db48371052087a83974abda3767d1aedec598 6986c7835adc13ba3f9d933b95671bb1f3984dc6 Georgi Gerganov ggerganov@gmail.com 2023-05-19T22:17:18+03:00 GitHub noreply@github.com 2023-05-19T22:17:18+03:00 ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508)
6986c7835adc13ba3f9d933b95671bb1f3984dc6 943e6081cc939df7584f8f0ab7057a39c2ef3271 Georgi Gerganov ggerganov@gmail.com 2023-05-19T21:17:28+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-19T21:17:28+03:00 tests : add missing header
943e6081cc939df7584f8f0ab7057a39c2ef3271 7694b52b9a206b93d59139c3c7c9b55da0f5aa59 Evan Jones evan.q.jones@gmail.com 2023-05-19T13:39:51-04:00 GitHub noreply@github.com 2023-05-19T20:39:51+03:00 examples : add persistent chat (#1495)
7694b52b9a206b93d59139c3c7c9b55da0f5aa59 79e3efb0e97b65b6cc72cd9ee970fa8189ad79a4 Jason McCartney jmac@theroot.org 2023-05-19T10:24:59-07:00 GitHub noreply@github.com 2023-05-19T20:24:59+03:00 main : make reverse prompt option act as a stop token in non-interactive mode (#1032)
79e3efb0e97b65b6cc72cd9ee970fa8189ad79a4 4b7e245adf63db675c3daab4a9bfddd451ef4097 David Kennedy dakennedyd@gmail.com 2023-05-19T13:16:30-04:00 GitHub noreply@github.com 2023-05-19T20:16:30+03:00 readme : adds WizardLM to the list of supported models (#1485)
4b7e245adf63db675c3daab4a9bfddd451ef4097 5ea43392731040b454c293123839b90e159cbb99 Georgi Gerganov ggerganov@gmail.com 2023-05-19T20:14:51+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-19T20:14:51+03:00 minor : fix compile warnings
5ea43392731040b454c293123839b90e159cbb99 ee9654138ab0ae5f138f4abddf56ca234ea3c352 Erik Scholz Green-Sky@users.noreply.github.com 2023-05-18T19:31:01+02:00 GitHub noreply@github.com 2023-05-18T19:31:01+02:00 make kv_f16 the default for api users (#1517)
ee9654138ab0ae5f138f4abddf56ca234ea3c352 dc271c52ed65e7c8dfcbaaf84dabb1f788e4f3d0 DannyDaemonic DannyDaemonic@gmail.com 2023-05-18T10:30:40-07:00 GitHub noreply@github.com 2023-05-18T19:30:40+02:00 Fixes #1511 lambda issue for w64devkit (mingw) (#1513)
dc271c52ed65e7c8dfcbaaf84dabb1f788e4f3d0 c238b5873a1ea496db03ffcfe124c9d0d83afbc6 Stephan Walter stephan@walter.name 2023-05-17T22:12:01Z GitHub noreply@github.com 2023-05-17T22:12:01Z Remove unused n_parts parameter (#1509)
c238b5873a1ea496db03ffcfe124c9d0d83afbc6 2b2646931bd2a2eb3e21c6f3733cc0e090b2e24b rankaiyx rankaiyx@rankaiyx.com 2023-05-17T22:47:58+08:00 GitHub noreply@github.com 2023-05-17T16:47:58+02:00 benchmark-matmul: Print the average of the test results (#1490)
2b2646931bd2a2eb3e21c6f3733cc0e090b2e24b 42627421ece816e632e6a0d757fa75150c687f87 Tom Jobbins 784313+TheBloke@users.noreply.github.com 2023-05-16T23:04:35+01:00 GitHub noreply@github.com 2023-05-17T00:04:35+02:00 convert.py: Support models which are stored in a single pytorch_model.bin (#1469)
42627421ece816e632e6a0d757fa75150c687f87 9560655409dc80771a9b19e838ff47c5c1df6483 Ilya Kurdyukov 59548320+ilyakurdyukov@users.noreply.github.com 2023-05-17T01:36:47+07:00 GitHub noreply@github.com 2023-05-16T18:36:47Z ~7% faster Q5_1 AVX2 code (#1477)
9560655409dc80771a9b19e838ff47c5c1df6483 2a5ee023ad3022bc0b505343394b9754587fb731 András Salamon ott2@users.noreply.github.com 2023-05-16T16:46:34+01:00 GitHub noreply@github.com 2023-05-16T17:46:34+02:00 define default model path once, sync path with readme (#1366)
2a5ee023ad3022bc0b505343394b9754587fb731 63d20469b85467c5729cc9a97bd44cc3da63423f sandyiscool sandyiscool@gmail.com 2023-05-16T14:00:15+05:30 GitHub noreply@github.com 2023-05-16T10:30:15+02:00 Add alternate include path for openblas (#1476)
63d20469b85467c5729cc9a97bd44cc3da63423f b5c9295eef2b56e307393b35b3a923e3518d226e zrm trustiosity.zrm@gmail.com 2023-05-14T22:25:42-04:00 GitHub noreply@github.com 2023-05-15T04:25:42+02:00 fix get_num_physical_cores() (#1436)
b5c9295eef2b56e307393b35b3a923e3518d226e eb363627fda5f47de8ab5e9be8abd426049d00df slaren slarengh@gmail.com 2023-05-14T22:46:00+02:00 GitHub noreply@github.com 2023-05-14T22:46:00+02:00 benchmark-matmul: fix clang-tidy issues, report results in GFLOPS (#1458)
eb363627fda5f47de8ab5e9be8abd426049d00df 79b2d5b69d80be0bf29312fb9a95854876b0a8a5 Johannes Gäßler johannesg@5d6.de 2023-05-14T20:53:23+02:00 GitHub noreply@github.com 2023-05-14T21:53:23+03:00 cuda : deduplicated dequantization code (#1453)
79b2d5b69d80be0bf29312fb9a95854876b0a8a5 13c351ad7292c5b5ab35db25c7a4f993e75d9cfd xaedes xaedes@gmail.com 2023-05-14T17:55:02+02:00 GitHub noreply@github.com 2023-05-14T18:55:02+03:00 ggml : alternative fix for race condition bug in non-inplace ggml_compute_forward_diag_mask_f32 (#1454)
13c351ad7292c5b5ab35db25c7a4f993e75d9cfd 60f8c361ca26328ef8523dfb08077fe2f1034490 Georgi Gerganov ggerganov@gmail.com 2023-05-14T18:22:50+03:00 GitHub noreply@github.com 2023-05-14T18:22:50+03:00 ggml : various fixes (#1450)
60f8c361ca26328ef8523dfb08077fe2f1034490 601a033475645370483973817d987928ea95f36c katsu560 118887472+katsu560@users.noreply.github.com 2023-05-14T19:03:51+09:00 GitHub noreply@github.com 2023-05-14T10:03:51Z ggml : add AVX support based on AVX2 code (#1430)
601a033475645370483973817d987928ea95f36c 08737ef720f0510c7ec2aa84d7f70c691073c35d Georgi Gerganov ggerganov@gmail.com 2023-05-14T10:20:19+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-14T10:20:19+03:00 ggml : add GGML_QNT_VERSION to track quantization format changes
08737ef720f0510c7ec2aa84d7f70c691073c35d bda4d7c215aa16b2a78e522521dfc0e1c2e8b194 Georgi Gerganov ggerganov@gmail.com 2023-05-13T17:40:58+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-13T17:40:58+03:00 cuda : fix convert function (#1412)
bda4d7c215aa16b2a78e522521dfc0e1c2e8b194 5a5aeb1e91009c72bf816400b758bb8a305616d7 Georgi Gerganov ggerganov@gmail.com 2023-05-13T17:25:09+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-13T17:25:09+03:00 make : fix PERF build with cuBLAS
5a5aeb1e91009c72bf816400b758bb8a305616d7 66841fdb0eaf0cc210757cc7f683d0f4eebadc21 Georgi Gerganov ggerganov@gmail.com 2023-05-13T16:55:14+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-13T16:55:14+03:00 llama : fix unused warning
66841fdb0eaf0cc210757cc7f683d0f4eebadc21 905d87b70aa189623d500a28602d7a3a755a4769 Georgi Gerganov ggerganov@gmail.com 2023-05-13T16:48:03+03:00 GitHub noreply@github.com 2023-05-13T16:48:03+03:00 ggml : multi-thread mul and diag_mask ops (#1428)
905d87b70aa189623d500a28602d7a3a755a4769 f954edda935a70a14cf0cc45ecc7fe7d60cf3e4b Johannes Gäßler johannesg@5d6.de 2023-05-13T15:38:36+02:00 GitHub noreply@github.com 2023-05-13T16:38:36+03:00 ggml : GPU-accelerated token generation (#1412)
f954edda935a70a14cf0cc45ecc7fe7d60cf3e4b f048af0230ad5381bb20b9e3c1467ec6fc7debdf xaedes xaedes@gmail.com 2023-05-13T14:56:40+02:00 GitHub noreply@github.com 2023-05-13T15:56:40+03:00 ggml : implement backward pass for llama + small training-llama-from-scratch example (#1360)
f048af0230ad5381bb20b9e3c1467ec6fc7debdf ac0cd259d54be7e45ffa2c7b2508812cf4f83ccf Georgi Gerganov ggerganov@gmail.com 2023-05-13T11:54:33+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-13T11:54:33+03:00 ggml : sync alibi fix from ggml repo
ac0cd259d54be7e45ffa2c7b2508812cf4f83ccf 0cd22e190aeaef867fa5db025b4d274f2fcfdcf6 3ooabkhxtn 31479382+3ooabkhxtn@users.noreply.github.com 2023-05-13T10:43:33+02:00 GitHub noreply@github.com 2023-05-13T08:43:33Z Adding SSE instructions to ggml_vec_dot_q4_0_q8_0 (#1413)
0cd22e190aeaef867fa5db025b4d274f2fcfdcf6 6456a4eb9f9c1f4de8f6908ef100daa78802ad00 Georgi Gerganov ggerganov@gmail.com 2023-05-13T11:23:15+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-13T11:23:15+03:00 llama : fix various warnings
6456a4eb9f9c1f4de8f6908ef100daa78802ad00 cdd5350892b1d4e521e930c77341f858fcfcd433 Rinne AsakusaRinne@gmail.com 2023-05-13T15:24:20+08:00 GitHub noreply@github.com 2023-05-13T10:24:20+03:00 embedding : remove unused code (#1426)
cdd5350892b1d4e521e930c77341f858fcfcd433 738ace394a6f8cf0174e90a97185d9e512c0e200 Georgi Gerganov ggerganov@gmail.com 2023-05-13T09:12:44+03:00 GitHub noreply@github.com 2023-05-13T09:12:44+03:00 readme : update Q4_0 perplexities
738ace394a6f8cf0174e90a97185d9e512c0e200 699b1ad7fe6f7b9e41d3cb41e61a8cc3ea5fc6b5 Georgi Gerganov ggerganov@gmail.com 2023-05-13T09:08:52+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-13T09:08:52+03:00 llama : free ggml context in set / copy state data (close #1425)
699b1ad7fe6f7b9e41d3cb41e61a8cc3ea5fc6b5 fb62f924336c9746da9976c6ab3c2e6460258d54 Henri Vasserman henv@hot.ee 2023-05-13T09:01:15+03:00 GitHub noreply@github.com 2023-05-13T09:01:15+03:00 opencl : fix kernels for the new formats (#1422)
fb62f924336c9746da9976c6ab3c2e6460258d54 773ee249fb6c14f791ac39f6ec05536f40dedc54 Georgi Gerganov ggerganov@gmail.com 2023-05-12T21:44:20+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-12T21:44:20+03:00 llama : fix --mtest option (close #1414)
773ee249fb6c14f791ac39f6ec05536f40dedc54 553fd4d4b5f3314f338be8006f62a4fdf7670186 Johannes Gäßler johannesg@5d6.de 2023-05-12T16:34:55+02:00 GitHub noreply@github.com 2023-05-12T14:34:55Z CLI args use - instead of _, backwards compatible (#1416)
553fd4d4b5f3314f338be8006f62a4fdf7670186 089b1c93ba2b93bc9a605af293730a028fad2c4e slaren slarengh@gmail.com 2023-05-12T15:40:53+02:00 GitHub noreply@github.com 2023-05-12T15:40:53+02:00 Add clang-tidy reviews to CI (#1407)
089b1c93ba2b93bc9a605af293730a028fad2c4e b9fd7eee57df101d4a3e3eabc9fd6c2cb13c9ca1 Rinne liu_yaohui1998@126.com 2023-05-12T13:39:40+08:00 GitHub noreply@github.com 2023-05-12T08:39:40+03:00 readme : add C#/.NET bindings repo (#1409)
b9fd7eee57df101d4a3e3eabc9fd6c2cb13c9ca1 b608b55a3ea8e4760c617418538465449175bdb8 Georgi Gerganov ggerganov@gmail.com 2023-05-12T00:23:08+03:00 GitHub noreply@github.com 2023-05-12T00:23:08+03:00 ggml : remove bit shuffling (#1405)
b608b55a3ea8e4760c617418538465449175bdb8 cf348a60e0af3905acd1d297cb064b918265b7ac CRD716 crd716@gmail.com 2023-05-11T10:10:19-05:00 GitHub noreply@github.com 2023-05-11T18:10:19+03:00 prompts : model agnostic DAN (#1304)
cf348a60e0af3905acd1d297cb064b918265b7ac e6a46b0ed1884c77267dc70693183e3b7164e0e0 Evan Jones evan.q.jones@gmail.com 2023-05-10T11:37:14-04:00 GitHub noreply@github.com 2023-05-10T11:37:14-04:00 main : add option to save full output to session (#1338)
e6a46b0ed1884c77267dc70693183e3b7164e0e0 9f8dbc4787f32cd9c13b5c647d497d13c1a06db2 DannyDaemonic DannyDaemonic@gmail.com 2023-05-09T10:53:28-07:00 GitHub noreply@github.com 2023-05-09T19:53:28+02:00 Locale fix for Windows (#1379)
9f8dbc4787f32cd9c13b5c647d497d13c1a06db2 41654efea879bbdf4fd794e13335929d4cf0eb90 Sami Farin 3876865+Safari77@users.noreply.github.com 2023-05-09T15:29:20+03:00 GitHub noreply@github.com 2023-05-09T14:29:20+02:00 use pause asm insn in busyloop to run the CPU (13600K) 10 °C cooler (#1314)
41654efea879bbdf4fd794e13335929d4cf0eb90 56551bc11f46b2716fdf61bb48ac28414889dc0a DannyDaemonic DannyDaemonic@gmail.com 2023-05-08T19:45:48-07:00 GitHub noreply@github.com 2023-05-08T19:45:48-07:00 Interface improvements and `--multiline-input` (previously `--author-mode`) (#1040)
56551bc11f46b2716fdf61bb48ac28414889dc0a fe60904eef4b504685fa0406cb19864ae619fb4f Georgi Gerganov ggerganov@gmail.com 2023-05-08T22:52:18+03:00 GitHub noreply@github.com 2023-05-08T22:52:18+03:00 readme : add notice about upcoming breaking change
fe60904eef4b504685fa0406cb19864ae619fb4f 003ba2fb4309e2339487564bd249e4fcc8d7ea01 AlpinDale 52078762+AlpinDale@users.noreply.github.com 2023-05-08T21:03:30+04:30 GitHub noreply@github.com 2023-05-08T19:33:30+03:00 readme : add TOC and Pygmalion instructions (#1359)
003ba2fb4309e2339487564bd249e4fcc8d7ea01 f9a6364912fd0463fddfdbc9ef9f79fdc281570d Pavol Rusnak pavol@rusnak.io 2023-05-08T16:48:21+02:00 GitHub noreply@github.com 2023-05-08T17:48:21+03:00 llama : fix hparams shadow (#1367)
f9a6364912fd0463fddfdbc9ef9f79fdc281570d 95078cc554fe03d4512363c7e4dec963f0047c72 Georgi Gerganov ggerganov@gmail.com 2023-05-08T17:41:54+03:00 GitHub noreply@github.com 2023-05-08T17:41:54+03:00 llama : require first token to be BOS (#1303)
95078cc554fe03d4512363c7e4dec963f0047c72 1f48b0abcfbd6cc99571e42348e0ec97e4be8b93 ubik2 ubik2@users.noreply.github.com 2023-05-08T04:54:26-07:00 GitHub noreply@github.com 2023-05-08T13:54:26+02:00 convert: add ability to convert safetensors files (#1276)
1f48b0abcfbd6cc99571e42348e0ec97e4be8b93 e1295513a48ae8254d8af5ec0250b56d6eaffefd Johannes Gäßler johannesg@5d6.de 2023-05-08T02:42:01+02:00 GitHub noreply@github.com 2023-05-08T02:42:01+02:00 Documented CUDA reproducibility, added warning (#1346)
e1295513a48ae8254d8af5ec0250b56d6eaffefd 1b0fd454650ef4d68a980e3225488b79e6e9af25 Henri Vasserman henv@hot.ee 2023-05-07T14:20:09+03:00 GitHub noreply@github.com 2023-05-07T13:20:09+02:00 CI: add Windows CLBlast and OpenBLAS builds (#1277)
1b0fd454650ef4d68a980e3225488b79e6e9af25 3924088512d9e12e90ed6dbf28a6c5712481d33e swittk switt1995@gmail.com 2023-05-07T10:03:23+07:00 GitHub noreply@github.com 2023-05-06T23:03:23-04:00 ggml : Allow usage of CLBlast alongside Accelerate.framework (#1336)
3924088512d9e12e90ed6dbf28a6c5712481d33e 173d0e6419e8f8f3c1f4f13201b777f4c60629f3 Jed Fox git@jedfox.com 2023-05-06T17:01:47-04:00 GitHub noreply@github.com 2023-05-06T17:01:47-04:00 Remove default arguments from sampling functions (#1343)
173d0e6419e8f8f3c1f4f13201b777f4c60629f3 a3b85b28da84c67c3406807aef5e0457bcc4b00f DaniAndTheWeb 57776841+DaniAndTheWeb@users.noreply.github.com 2023-05-05T23:57:14+02:00 GitHub noreply@github.com 2023-05-05T23:57:14+02:00 makefile: automatic Arch Linux detection (#1332)
a3b85b28da84c67c3406807aef5e0457bcc4b00f 921dcee00a55d9aba3b3026d0509d31ac8386e2a Erik Scholz Green-Sky@users.noreply.github.com 2023-05-05T22:56:09+02:00 GitHub noreply@github.com 2023-05-05T22:56:09+02:00 ci : add cublas to windows release (#1271)
921dcee00a55d9aba3b3026d0509d31ac8386e2a 2d13786e91ec9fd28ddf737053822042a824da78 Pavol Rusnak pavol@rusnak.io 2023-05-05T16:43:36+02:00 GitHub noreply@github.com 2023-05-05T16:43:36+02:00 readme: add missing info (#1324)
2d13786e91ec9fd28ddf737053822042a824da78 a90e96b266873ebb5e947c9864b12193bdada0fb Ionoclast Laboratories brigham@ionoclast.com 2023-05-05T08:18:21-04:00 GitHub noreply@github.com 2023-05-05T14:18:21+02:00 Fix for OpenCL / clbast builds on macOS. (#1329)
a90e96b266873ebb5e947c9864b12193bdada0fb 94c5652fc0f4d04ac54412c4d81e2ebcdafb6ede Benjamin Lecaillon 84293038+blecaillon@users.noreply.github.com 2023-05-05T02:17:07+02:00 GitHub noreply@github.com 2023-05-05T03:17:07+03:00 Convert.py @staticmethod (#1327)
94c5652fc0f4d04ac54412c4d81e2ebcdafb6ede 34d9f22f44c42d345cc72c8f3aa4cb71c5df0acb slaren slarengh@gmail.com 2023-05-05T00:58:56+02:00 GitHub noreply@github.com 2023-05-05T00:58:56+02:00 quantize: make output filename optional, default to ggml-model-<ftype>.bin (#1301)
34d9f22f44c42d345cc72c8f3aa4cb71c5df0acb d3e8093e9b5845514b049ede3b12728c8f013eba Ivan Stepanov ivanstepanovftw@gmail.com 2023-05-04T19:56:27+03:00 GitHub noreply@github.com 2023-05-04T18:56:27+02:00 Wrap exceptions in std::exception to verbose output on exception. (#1316)
d3e8093e9b5845514b049ede3b12728c8f013eba 360cfe5bec852805b84eec799102fc6f45df9fef Ivan Stepanov ivanstepanovftw@gmail.com 2023-05-04T19:54:37+03:00 GitHub noreply@github.com 2023-05-04T18:54:37+02:00 convert: support DT_BF16 tensors (#1309)
360cfe5bec852805b84eec799102fc6f45df9fef 2edbdb0f99336cb41f0995061c7602ed54beb863 44670 44670@users.noreply.github.com 2023-05-05T00:33:31+08:00 GitHub noreply@github.com 2023-05-04T19:33:31+03:00 readme : add OpenBuddy link (#1321)
2edbdb0f99336cb41f0995061c7602ed54beb863 20fbf2a2a08d8edefe9b3435fa86f8b2f63f8588 44670 44670@users.noreply.github.com 2023-05-04T23:41:12+08:00 GitHub noreply@github.com 2023-05-04T18:41:12+03:00 main : add --in-suffix option (#1318)
20fbf2a2a08d8edefe9b3435fa86f8b2f63f8588 db1080876a62ec3bb4119d90b16e7dce7594b733 Ron Jailall rojailal@gmail.com 2023-05-04T11:05:59-04:00 GitHub noreply@github.com 2023-05-04T18:05:59+03:00 ggml : change immintrin.h to intrin.h for compatibility (#1307)
db1080876a62ec3bb4119d90b16e7dce7594b733 c65a7fbfa9c736416a25369cc05d356789df4c15 DannyDaemonic DannyDaemonic@gmail.com 2023-05-04T05:08:25-07:00 GitHub noreply@github.com 2023-05-04T05:08:25-07:00 Only escape prompts when used with `-e` (#1311)
c65a7fbfa9c736416a25369cc05d356789df4c15 f647ce040ff06348d2ceaa5443a6a7a8b80c70c9 DannyDaemonic DannyDaemonic@gmail.com 2023-05-04T03:02:59-07:00 GitHub noreply@github.com 2023-05-04T03:02:59-07:00 Update main's README.md with new features (#1296)
f647ce040ff06348d2ceaa5443a6a7a8b80c70c9 799fdc1b5d888b8a8682baf112e1c2a2df0df1c4 Tomas tom.tomas.36478119@gmail.com 2023-05-04T17:02:30+07:00 GitHub noreply@github.com 2023-05-04T03:02:30-07:00 fix #1224 reverse prompt and multi line (#1297)
799fdc1b5d888b8a8682baf112e1c2a2df0df1c4 6daa09d87926fe654385c2887e39ec3eeaa58120 Georgi Gerganov ggerganov@gmail.com 2023-05-03T23:24:20+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-03T23:24:20+03:00 ggml : vectorize Q8_0 quantization
6daa09d87926fe654385c2887e39ec3eeaa58120 bca9ad938a2a43621cf406d993b755cc91728dd5 khimaros me@khimaros.com 2023-05-03T10:58:11-07:00 GitHub noreply@github.com 2023-05-03T20:58:11+03:00 examples : read chat prompts from a template file (#1196)
bca9ad938a2a43621cf406d993b755cc91728dd5 e2a937ca6abadc7e01e139db31e6db9dce16e3e9 Georgi Gerganov ggerganov@gmail.com 2023-05-03T20:09:42+03:00 GitHub noreply@github.com 2023-05-03T20:09:42+03:00 minor : fix whitespaces (#1302)
e2a937ca6abadc7e01e139db31e6db9dce16e3e9 b0c71c7b6dc0c0adb507d78f401e95e7ab0f5a38 Georgi Gerganov ggerganov@gmail.com 2023-05-03T18:43:23+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-03T18:43:23+03:00 minor : fix trailing whitespaces
b0c71c7b6dc0c0adb507d78f401e95e7ab0f5a38 a8a2efdc8161d4f69a0dd863e741c11fbd5df85c KASR karim.asrih@gmail.com 2023-05-03T17:31:28+02:00 GitHub noreply@github.com 2023-05-03T18:31:28+03:00 scripts : platform independent script to verify sha256 checksums (#1203)
a8a2efdc8161d4f69a0dd863e741c11fbd5df85c e216aa04633892b972d013719e38b59fd4917341 CRD716 crd716@gmail.com 2023-05-03T10:26:47-05:00 GitHub noreply@github.com 2023-05-03T18:26:47+03:00 examples : various prompt and example fixes (#1298)
e216aa04633892b972d013719e38b59fd4917341 2485d7a4d39406cd0f468e35551b472cceb5bd61 Evan Jones evan.q.jones@gmail.com 2023-05-02T22:26:13-04:00 GitHub noreply@github.com 2023-05-02T22:26:13-04:00 llama : only copy used KV cache in get / set state (#1272)
2485d7a4d39406cd0f468e35551b472cceb5bd61 13b0c68ed7a9948db0720f7393df094ab1005b14 DannyDaemonic DannyDaemonic@gmail.com 2023-05-02T18:46:20-07:00 GitHub noreply@github.com 2023-05-02T18:46:20-07:00 Process escape sequences given in prompts (#1173)
13b0c68ed7a9948db0720f7393df094ab1005b14 55bc5f0900d925c539488901c5538b637d68665c DannyDaemonic DannyDaemonic@gmail.com 2023-05-02T18:01:57-07:00 GitHub noreply@github.com 2023-05-02T18:01:57-07:00 Handle signals properly on Windows (#1123)
55bc5f0900d925c539488901c5538b637d68665c 9daff419f6be818595ddbea293a0ea7283af726e DannyDaemonic DannyDaemonic@gmail.com 2023-05-02T17:52:35-07:00 GitHub noreply@github.com 2023-05-02T17:52:35-07:00 Call sh on build-info.sh (#1294)
9daff419f6be818595ddbea293a0ea7283af726e bf4b22ffe433dc5e2fba7588c4485a7e51b1a30d kuvaus 22169537+kuvaus@users.noreply.github.com 2023-05-03T03:43:43+03:00 GitHub noreply@github.com 2023-05-03T02:43:43+02:00 fix build-info.h for git submodules (#1289)
bf4b22ffe433dc5e2fba7588c4485a7e51b1a30d 67c77799e025a8425c23a6a0599c007f46ded590 slaren slarengh@gmail.com 2023-05-03T01:36:45+02:00 GitHub noreply@github.com 2023-05-03T01:36:45+02:00 fix missing parameters in `llama_init_from_gpt_params` (#1293)
67c77799e025a8425c23a6a0599c007f46ded590 0e6cbff1b7509628c588e661166f6e187137734d Ron Evans ron@hybridgroup.com 2023-05-02T22:39:51+02:00 GitHub noreply@github.com 2023-05-02T23:39:51+03:00 examples : add llama_init_from_gpt_params() common function (#1290)
0e6cbff1b7509628c588e661166f6e187137734d 5d5817ca603d4cb451bed26594aa3dcd93f4ec56 Georgi Gerganov ggerganov@gmail.com 2023-05-02T23:09:08+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-02T23:09:08+03:00 llama : fix compile warnings
5d5817ca603d4cb451bed26594aa3dcd93f4ec56 8c9be35ff998cbb4178b0fedcb9afd85cb6852e2 Georgi Gerganov ggerganov@gmail.com 2023-05-02T22:14:50+03:00 Georgi Gerganov ggerganov@gmail.com 2023-05-02T22:14:50+03:00 ggml : fix 32-bit ARM
8c9be35ff998cbb4178b0fedcb9afd85cb6852e2 cc0bb7235c72e50a621800e366d0e4fe315f0e11 Ron Evans ron@hybridgroup.com 2023-05-02T19:53:52+02:00 GitHub noreply@github.com 2023-05-02T20:53:52+03:00 examples : improve vertical alignment of a few variables (#1286)
cc0bb7235c72e50a621800e366d0e4fe315f0e11 2bb992f034689e2ddd7b9ac05163b0359a5957b3 Marvin Gießing marvin.giessing@gmail.com 2023-05-02T18:42:16+02:00 GitHub noreply@github.com 2023-05-02T19:42:16+03:00 ggml : fix ppc64le build error and make cmake detect Power processors (#1284)
2bb992f034689e2ddd7b9ac05163b0359a5957b3 e2cd5069999181a9e4a22cf420e0491878b3062f Robert Brisita 986796+rbrisita@users.noreply.github.com 2023-05-02T12:23:44-04:00 GitHub noreply@github.com 2023-05-02T19:23:44+03:00 llama : allow 0 as a seed number. (#1275)
e2cd5069999181a9e4a22cf420e0491878b3062f 2d099e5193d73f800b646c39e2fad08c1c1f1096 Ron Evans ron@hybridgroup.com 2023-05-02T18:13:26+02:00 GitHub noreply@github.com 2023-05-02T19:13:26+03:00 main : switch input_noecho to input_echo to remove negation (#979)
2d099e5193d73f800b646c39e2fad08c1c1f1096 f4cef87edfd1b2f8d5befd4fde54ca2e03987bea slaren slarengh@gmail.com 2023-05-02T16:03:00+02:00 GitHub noreply@github.com 2023-05-02T16:03:00+02:00 ggml: add names to tensors (#1268)
f4cef87edfd1b2f8d5befd4fde54ca2e03987bea 58b367c2d757c0ea12aec672382462b42204c724 DannyDaemonic DannyDaemonic@gmail.com 2023-05-01T09:23:47-07:00 GitHub noreply@github.com 2023-05-01T18:23:47+02:00 Add git-based build information for better issue tracking (#1232)
58b367c2d757c0ea12aec672382462b42204c724 ea3a0ad6b6b5ca4693b94acd4cb32e2803f66fae slaren 2141330+slaren@users.noreply.github.com 2023-05-01T18:11:07+02:00 GitHub noreply@github.com 2023-05-01T18:11:07+02:00 cuBLAS: refactor and optimize f16 mat mul performance (#1259)
ea3a0ad6b6b5ca4693b94acd4cb32e2803f66fae 2bdc09646d8c6cb74a6f573e9081586b4b83b9d1 xloem 0xloem@gmail.com 2023-05-01T08:58:51-04:00 GitHub noreply@github.com 2023-05-01T15:58:51+03:00 llama : update stubs for systems without mmap and mlock (#1266)
2bdc09646d8c6cb74a6f573e9081586b4b83b9d1 70269cae37538461ff816e714afbb3ebcdcdc26b Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-05-01T05:56:07-06:00 GitHub noreply@github.com 2023-05-01T14:56:07+03:00 ggml : fix ggml_used_mem() (#1264)
70269cae37538461ff816e714afbb3ebcdcdc26b b925f1f1b082319ee69943f8d1a83ac9b6ff09ca Georgi Gerganov ggerganov@gmail.com 2023-05-01T14:54:59+03:00 GitHub noreply@github.com 2023-05-01T14:54:59+03:00 llama : fix session load / save (#1263)
b925f1f1b082319ee69943f8d1a83ac9b6ff09ca 90b19bd6eee943832584f9cac0b6f9ea29cc42a4 slaren 2141330+slaren@users.noreply.github.com 2023-05-01T13:32:22+02:00 GitHub noreply@github.com 2023-05-01T13:32:22+02:00 cuBLAS: fall back to pageable memory if pinned alloc fails (#1233)
90b19bd6eee943832584f9cac0b6f9ea29cc42a4 7ff0dcd32091c703a12adb0c57c32c565ce17664 Alex Klinkhamer git@grencez.dev 2023-05-01T00:24:20-07:00 GitHub noreply@github.com 2023-05-01T10:24:20+03:00 llama : let context be const when accessing const data (#1261)
7ff0dcd32091c703a12adb0c57c32c565ce17664 6f796992869f306c48484d62a39f2a181ae2fd6f Georgi Gerganov ggerganov@gmail.com 2023-04-30T22:28:51+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-30T22:28:51+03:00 ggml : fix UB (int << 31)
6f796992869f306c48484d62a39f2a181ae2fd6f a5d30b1f53677cb50791fec41c43e93274347303 Pavol Rusnak pavol@rusnak.io 2023-04-30T20:48:38+02:00 GitHub noreply@github.com 2023-04-30T20:48:38+02:00 build: add armv{6,7,8} support to cmake (#1251)
a5d30b1f53677cb50791fec41c43e93274347303 76a884920aa1d2fc0dc7a7ac12dfc5ec5816377c jon-chuang 9093549+jon-chuang@users.noreply.github.com 2023-04-30T14:41:35-04:00 GitHub noreply@github.com 2023-04-30T21:41:35+03:00 common : better default number of threads (#934)
76a884920aa1d2fc0dc7a7ac12dfc5ec5816377c 6bc4400e67e6bc4faad3ad3d5e9d8a6576a9752d 0cc4m picard12@live.de 2023-04-30T20:34:52+02:00 GitHub noreply@github.com 2023-04-30T21:34:52+03:00 ggml : add CLBlast q5_0, q5_1, q8_0 dequant kernels (#1225)
6bc4400e67e6bc4faad3ad3d5e9d8a6576a9752d f0d70f147d969e41fa410b8af2965a27aa901eb9 Georgi Gerganov ggerganov@gmail.com 2023-04-30T19:07:00+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-30T19:07:43+03:00 ggml : add Q5 WASM SIMD + GGML_FTYPE
f0d70f147d969e41fa410b8af2965a27aa901eb9 3e5aa8a1c44051153d6d7b3eeca2f4b4e5fb310c Stephan Walter stephan@walter.name 2023-04-30T12:32:37Z GitHub noreply@github.com 2023-04-30T12:32:37Z Various fixes to mat_mul benchmark (#1253)
3e5aa8a1c44051153d6d7b3eeca2f4b4e5fb310c c3ca7a5f0546c561eb278be3f2fe335795679e01 Georgi Gerganov ggerganov@gmail.com 2023-04-30T10:25:46+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-30T10:25:46+03:00 ggml : fix labels for GGML_OP_ALIBI
c3ca7a5f0546c561eb278be3f2fe335795679e01 e8c051611abfc9a7f37fd4bba48217180893bd68 Georgi Gerganov ggerganov@gmail.com 2023-04-29T21:34:23+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-29T21:34:23+03:00 ggml : fix 32-bit ARM NEON
e8c051611abfc9a7f37fd4bba48217180893bd68 0b5a9350993e6fc8be45dc2a3eafc1fd0812d392 Georgi Gerganov ggerganov@gmail.com 2023-04-29T21:12:56+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-29T21:12:56+03:00 ggml : use vzip instead of vuzp for consistency
0b5a9350993e6fc8be45dc2a3eafc1fd0812d392 ec728e44d7488c2da3560970317708b2b12b9c04 Georgi Gerganov ggerganov@gmail.com 2023-04-29T19:28:36+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-29T19:28:36+03:00 ggml : fix visibility and unused warnings
ec728e44d7488c2da3560970317708b2b12b9c04 214b6a35702a489e3738acd81fad6d46182d3036 Georgi Gerganov ggerganov@gmail.com 2023-04-29T18:43:42+03:00 GitHub noreply@github.com 2023-04-29T18:43:42+03:00 ggml : fix #if for f32_f32 mul_mat (CLBlast) (#1229)
214b6a35702a489e3738acd81fad6d46182d3036 305eb5afd51325e3142c01c17431febb7c67de87 Georgi Gerganov ggerganov@gmail.com 2023-04-29T18:43:28+03:00 GitHub noreply@github.com 2023-04-29T18:43:28+03:00 ggml : adjust mul_mat_f16 work memory (#1226)
305eb5afd51325e3142c01c17431febb7c67de87 84ca9c2ecf3391d911589d0fe2b483cbfb4b82a6 Georgi Gerganov ggerganov@gmail.com 2023-04-29T13:53:12+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-29T13:53:12+03:00 build : fix reference to old llama_util.h
84ca9c2ecf3391d911589d0fe2b483cbfb4b82a6 334637e43e3a0529b4b50e2c22968b1ed1633353 Georgi Gerganov ggerganov@gmail.com 2023-04-29T13:48:11+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-29T13:48:11+03:00 examples : fix save-load-state + rename llama-util.h
334637e43e3a0529b4b50e2c22968b1ed1633353 dd7eff57d8491792010b1002b8de6a4b54912e5c Georgi Gerganov ggerganov@gmail.com 2023-04-29T09:51:06+03:00 GitHub noreply@github.com 2023-04-29T09:51:06+03:00 common : change default parameters to pre-#1126 (#1223)
dd7eff57d8491792010b1002b8de6a4b54912e5c 7fc50c051ae8a78e9643fdf172d12e20f2dd9b6c Ivan Stepanov ivanstepanovftw@gmail.com 2023-04-29T08:34:41+03:00 GitHub noreply@github.com 2023-04-29T08:34:41+03:00 llama : new sampling algorithms (#1126)
7fc50c051ae8a78e9643fdf172d12e20f2dd9b6c b1ee8f59b4101b46999a0995d9a34506f7285466 slaren 2141330+slaren@users.noreply.github.com 2023-04-29T02:04:18+02:00 GitHub noreply@github.com 2023-04-29T02:04:18+02:00 cuBLAS: use host pinned memory and dequantize while copying (#1207)
b1ee8f59b4101b46999a0995d9a34506f7285466 36d19a603b221d1bd7897fcb10e823e2103b052d Henri Vasserman henv@hot.ee 2023-04-29T02:31:56+03:00 GitHub noreply@github.com 2023-04-29T01:31:56+02:00 cuBLAS: non-contiguous tensor support (#1215)
36d19a603b221d1bd7897fcb10e823e2103b052d 7f15c5c477d9933689a9d1c40794483e350c2f19 Stephan Walter stephan@walter.name 2023-04-28T23:10:43Z GitHub noreply@github.com 2023-04-28T23:10:43Z Remove Q4_3 which is no better than Q5 (#1218)
7f15c5c477d9933689a9d1c40794483e350c2f19 55390bcaf2579a5435564d7267ae3ed367837fd6 Georgi Gerganov ggerganov@gmail.com 2023-04-28T21:32:52+03:00 GitHub noreply@github.com 2023-04-28T21:32:52+03:00 readme : update hot topics
55390bcaf2579a5435564d7267ae3ed367837fd6 5fba3c016bfd1d73a070e7c93dac14162ce118d0 Georgi Gerganov ggerganov@gmail.com 2023-04-28T20:37:43+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-28T20:51:05+03:00 ggml : sync ggml (ggml_alibi)
5fba3c016bfd1d73a070e7c93dac14162ce118d0 1481a9cf25ea2e4abef6b13a57660a35f3e66af1 CRD716 crd716@gmail.com 2023-04-28T11:13:33-05:00 GitHub noreply@github.com 2023-04-28T19:13:33+03:00 examples : add Jeopardy example (#1168)
1481a9cf25ea2e4abef6b13a57660a35f3e66af1 11d902364b0e3b503a02a4e757ee2dc38aacb68f Evan Jones evan.q.jones@gmail.com 2023-04-28T11:59:37-04:00 GitHub noreply@github.com 2023-04-28T18:59:37+03:00 llama : add session file format and saved sessions in main (#1169)
11d902364b0e3b503a02a4e757ee2dc38aacb68f 7296c961d9303010a2b98379f738da2a8a55aa1b Georgi Gerganov ggerganov@gmail.com 2023-04-28T17:58:44+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-28T17:59:08+03:00 ggml : add helper debug printf in soft_max
7296c961d9303010a2b98379f738da2a8a55aa1b 78ec543733d10a1629f984fd0302fdaa4e87fe66 0cc4m picard12@live.de 2023-04-28T16:57:16+02:00 GitHub noreply@github.com 2023-04-28T17:57:16+03:00 ggml : add CLBlast support (#1164)
78ec543733d10a1629f984fd0302fdaa4e87fe66 92a6e13a31ba052abd9062af6cb8df2a293ce661 Folko-Ven 71110216+Folko-Ven@users.noreply.github.com 2023-04-28T19:22:48+05:00 GitHub noreply@github.com 2023-04-28T16:22:48+02:00 Correcting link to w64devkit (#1214)
92a6e13a31ba052abd9062af6cb8df2a293ce661 04aaae1d79482cad2564412f3b32e70298ac7789 Johannes Gäßler johannesg@5d6.de 2023-04-28T15:40:32+02:00 GitHub noreply@github.com 2023-04-28T15:40:32+02:00 Add Manjaro CUDA include and lib dirs to Makefile (#1212)
04aaae1d79482cad2564412f3b32e70298ac7789 0b2da20538d01926b77ea237dd1c930c4d20b686 Yann Follet 131855179+YannFollet@users.noreply.github.com 2023-04-28T19:59:48+08:00 GitHub noreply@github.com 2023-04-28T11:59:48Z add avx2 for dot_q8_0_q8_0, 2x faster than scalar (#1211)
0b2da20538d01926b77ea237dd1c930c4d20b686 f9be42add0bf3ce61814b7ede0e6d0dda9ff22c6 Stephan Walter stephan@walter.name 2023-04-26T20:26:42Z GitHub noreply@github.com 2023-04-26T23:26:42+03:00 ggml : slightly faster AVX2 implementation for Q5 (#1197)
f9be42add0bf3ce61814b7ede0e6d0dda9ff22c6 574406dc7e350ddbffaeca33bf0392b7bfeb1436 Georgi Gerganov ggerganov@gmail.com 2023-04-26T23:24:42+03:00 GitHub noreply@github.com 2023-04-26T23:24:42+03:00 readme : add quantization info
574406dc7e350ddbffaeca33bf0392b7bfeb1436 87a6f846d3e929632c45916dd08f1e2a9c72d2a3 Georgi Gerganov ggerganov@gmail.com 2023-04-26T23:14:13+03:00 GitHub noreply@github.com 2023-04-26T23:14:13+03:00 ggml : add Q5_0 and Q5_1 quantization (#1187)
87a6f846d3e929632c45916dd08f1e2a9c72d2a3 ea3ad7eb60cfb44526a58122e8019850f437cd1b Ásgeir Bjarni Ingvarsson asgeir@fundinn.org 2023-04-26T20:08:43Z GitHub noreply@github.com 2023-04-26T22:08:43+02:00 Allow setting the rng seed after initialization. (#1184)
ea3ad7eb60cfb44526a58122e8019850f437cd1b 859fee6dfb00fab7ce6bc215b4adae78d82f4759 DaniAndTheWeb 57776841+DaniAndTheWeb@users.noreply.github.com 2023-04-26T22:03:03+02:00 GitHub noreply@github.com 2023-04-26T22:03:03+02:00 Updating build instructions to include BLAS support (#1183)
859fee6dfb00fab7ce6bc215b4adae78d82f4759 4afcc378698e057fcde64e23eb664e5af8dd6956 Pavol Rusnak pavol@rusnak.io 2023-04-26T18:43:27+02:00 GitHub noreply@github.com 2023-04-26T18:43:27+02:00 quantize : use `map` to assign quantization type from `string` (#1191)
4afcc378698e057fcde64e23eb664e5af8dd6956 667c501334ace706e3abc3f7a37cf1d6b4228745 Stephan Walter stephan@walter.name 2023-04-25T21:41:56Z GitHub noreply@github.com 2023-04-25T23:41:56+02:00 Update SHA256SUMS after quantization change (#1181)
667c501334ace706e3abc3f7a37cf1d6b4228745 bb98e77be704584fb40b0400394b4c16ae75f8e2 ostix360 55257054+ostix360@users.noreply.github.com 2023-04-25T23:33:08+02:00 GitHub noreply@github.com 2023-04-25T23:33:08+02:00 py : cast lora_alpha to int in convert-lora-to-ggml (#1170)
bb98e77be704584fb40b0400394b4c16ae75f8e2 7a32fcb3b29f4db8aed8a85dc58eb958fb118153 Pavol Rusnak pavol@rusnak.io 2023-04-25T23:19:57+02:00 GitHub noreply@github.com 2023-04-25T23:19:57+02:00 nix: use convert.py instead of legacy wrapper convert-pth-to-ggml.py (#981)
7a32fcb3b29f4db8aed8a85dc58eb958fb118153 dd0eabc049fb1efc631cab8eb0a646808d704e18 Georgi Gerganov ggerganov@gmail.com 2023-04-25T23:40:51+03:00 GitHub noreply@github.com 2023-04-25T23:40:51+03:00 ggml : add Q8_0 quantization format (rename the old one to Q8_1) (ARM NEON) (#1179)
dd0eabc049fb1efc631cab8eb0a646808d704e18 54bb60e26858be251a0eb3cb70f80322aff804a0 unbounded haakon@likedan.net 2023-04-25T19:20:46+02:00 GitHub noreply@github.com 2023-04-25T20:20:46+03:00 ggml : use full range for Q4_0 and Q4_2 quantization (#729)
54bb60e26858be251a0eb3cb70f80322aff804a0 8a0f8673ba1cdc6aa6df27a9fbc698431ca70e8d xaedes xaedes@gmail.com 2023-04-24T23:02:02+02:00 GitHub noreply@github.com 2023-04-24T23:02:02+02:00 ggml : fix bug in ggml_compute_forward_sum_f32 (#1162)
8a0f8673ba1cdc6aa6df27a9fbc698431ca70e8d 0c5692345d5c046dbc6a7d311a00ae5842ac39c3 Georgi Gerganov ggerganov@gmail.com 2023-04-24T22:18:25+03:00 GitHub noreply@github.com 2023-04-24T22:18:25+03:00 ggml : export symbols (#1155)
0c5692345d5c046dbc6a7d311a00ae5842ac39c3 957c8ae21d1e7052ea45a40ee8c0407b909e90cc xaedes xaedes@gmail.com 2023-04-24T18:23:31+02:00 GitHub noreply@github.com 2023-04-24T19:23:31+03:00 examples : add save_load_state example (#1150)
957c8ae21d1e7052ea45a40ee8c0407b909e90cc 9b0a4d421459f4e5e1af735c9784c3247b379025 Georgi Gerganov ggerganov@gmail.com 2023-04-24T18:47:03+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-24T18:47:30+03:00 llama : increase scratch buffer size for 65B (ref #1152)
9b0a4d421459f4e5e1af735c9784c3247b379025 2ec83428de7a876ecbbe484e1de42b73b5a40e25 mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2023-04-24T17:45:32+02:00 GitHub noreply@github.com 2023-04-24T15:45:32Z examples/main README improvements and some light refactoring (#1131)
2ec83428de7a876ecbbe484e1de42b73b5a40e25 e4cf982e0d4fcfbb4b977a52dbeacd115da10c3b Stephan Walter stephan@walter.name 2023-04-24T15:38:26Z GitHub noreply@github.com 2023-04-24T15:38:26Z Fix build for gcc 8 and test in CI (#1154)
e4cf982e0d4fcfbb4b977a52dbeacd115da10c3b c4fe84fb0d28851a5c10e5a633f82ae2ba3b7fae slaren 2141330+slaren@users.noreply.github.com 2023-04-24T17:29:58+02:00 GitHub noreply@github.com 2023-04-24T17:29:58+02:00 Fix cuda compilation (#1128)
c4fe84fb0d28851a5c10e5a633f82ae2ba3b7fae 1d78fecdab4087028a38517e86ed129f077174d8 Georgi Gerganov ggerganov@gmail.com 2023-04-24T07:40:02+03:00 GitHub noreply@github.com 2023-04-24T07:40:02+03:00 llama : refactor get / set state + remove redundant kv cache API (#1143)
1d78fecdab4087028a38517e86ed129f077174d8 284685f1692258c2bcf08b86b723b80ba2e66c7a slaren 2141330+slaren@users.noreply.github.com 2023-04-23T23:03:44+02:00 GitHub noreply@github.com 2023-04-23T23:03:44+02:00 Fix LoRA acronym (#1145)
284685f1692258c2bcf08b86b723b80ba2e66c7a edce63baa9dbd3963c3441bce07ee0acbb635697 Georgi Gerganov ggerganov@gmail.com 2023-04-23T19:57:09+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-23T19:57:09+03:00 scripts : add helper scripts to synch ggml repo
edce63baa9dbd3963c3441bce07ee0acbb635697 ec9cdb6752dd96b3cc74d90ad1adeba5b4fa2b0e DannyDaemonic DannyDaemonic@gmail.com 2023-04-23T08:37:02-07:00 GitHub noreply@github.com 2023-04-23T15:37:02Z Added README.md for main with examples and explanations (#1139)
ec9cdb6752dd96b3cc74d90ad1adeba5b4fa2b0e e4422e299c10c7e84c8e987770ef40d31905a76b Georgi Gerganov ggerganov@gmail.com 2023-04-23T18:32:52+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-23T18:32:52+03:00 ggml : do not print perf ops that have not been used at all
e4422e299c10c7e84c8e987770ef40d31905a76b 53c8434398b3cba7ac6298cdd44abd40f0e640b1 Georgi Gerganov ggerganov@gmail.com 2023-04-23T18:15:39+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-23T18:15:39+03:00 ggml : better PERF prints + support "LLAMA_PERF=1 make"
53c8434398b3cba7ac6298cdd44abd40f0e640b1 c6524f46eb93fdb949330293a8469fd70080bd5a Stephan Walter stephan@walter.name 2023-04-23T11:01:03Z GitHub noreply@github.com 2023-04-23T11:01:03Z Improve AVX2 for vec_dot_q4_3_q8_0 (#1138)
c6524f46eb93fdb949330293a8469fd70080bd5a c9e2c26f413377b352845f442cdab976ce85a05d Pavol Rusnak pavol@rusnak.io 2023-04-23T10:21:26+02:00 GitHub noreply@github.com 2023-04-23T10:21:26+02:00 readme : update gpt4all instructions (#980)
c9e2c26f413377b352845f442cdab976ce85a05d 0e018fe008eacebdbcfa2d61b6c988c245c961cd Yishuo Wang MeouSker77@outlook.com 2023-04-23T15:57:05+08:00 GitHub noreply@github.com 2023-04-23T07:57:05Z A better `packNibbles` and `mul_sum_i8_pairs_float` implementation using AVX512 (#1119)
0e018fe008eacebdbcfa2d61b6c988c245c961cd 857308d1e8fb6afe33edb481d48560eee8fe7d7c Georgi Gerganov ggerganov@gmail.com 2023-04-22T16:31:56+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-22T16:32:07+03:00 ggml : fix Q4_3 cuBLAS
857308d1e8fb6afe33edb481d48560eee8fe7d7c c50b628810f36a3e6e0324371f6db579eacefa0e Stephan Walter stephan@walter.name 2023-04-22T13:12:29Z GitHub noreply@github.com 2023-04-22T16:12:29+03:00 ci : trigger CI for drafts, but not most PR actions (#1125)
c50b628810f36a3e6e0324371f6db579eacefa0e 5f939498d517b4dddbe904f202e895a3ecfb9dc4 Stephan Walter stephan@walter.name 2023-04-22T10:54:13Z GitHub noreply@github.com 2023-04-22T10:54:13Z Fix CI: ARM NEON, quantization unit tests, editorconfig (#1122)
5f939498d517b4dddbe904f202e895a3ecfb9dc4 36b4f7e06406eed8a605cc9f2921d9244ef6a8e5 unbounded haakon@likedan.net 2023-04-22T11:10:39+02:00 GitHub noreply@github.com 2023-04-22T12:10:39+03:00 ggml : unit test for quantization functions (#953)
36b4f7e06406eed8a605cc9f2921d9244ef6a8e5 10f19c1121068ce3dab9bece03a8b9caaea2db36 wbpxre150 100937007+wbpxre150@users.noreply.github.com 2023-04-22T16:56:35+08:00 GitHub noreply@github.com 2023-04-22T11:56:35+03:00 llama : print timings on ctrl+c exit (#1021)
10f19c1121068ce3dab9bece03a8b9caaea2db36 7e312f165c5047d6e16680d1eebc83055e95c313 eiery 19350831+eiery@users.noreply.github.com 2023-04-22T04:27:05-04:00 GitHub noreply@github.com 2023-04-22T11:27:05+03:00 llama : have n_batch default to 512 (#1091)
7e312f165c5047d6e16680d1eebc83055e95c313 872c365a9176a011b13d31269bb3121fa89c37e1 Howard Su howard0su@gmail.com 2023-04-22T16:18:20+08:00 GitHub noreply@github.com 2023-04-22T11:18:20+03:00 cmake : fix build under Windows when enable BUILD_SHARED_LIBS (#1100)
872c365a9176a011b13d31269bb3121fa89c37e1 955ef9a5d53d8f911fe00580ac9bd0caa56430af Georgi Gerganov ggerganov@gmail.com 2023-04-22T11:08:12+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-22T11:08:12+03:00 ggml : fix AVX build + update to new Q8_0 format
955ef9a5d53d8f911fe00580ac9bd0caa56430af c5aa5e577741d0359ad26ec50b9e21a74c65d911 Georgi Gerganov ggerganov@gmail.com 2023-04-22T10:55:35+03:00 GitHub noreply@github.com 2023-04-22T10:55:35+03:00 ggml : alternative Q4_3 implementation using modified Q8_0 (#1109)
c5aa5e577741d0359ad26ec50b9e21a74c65d911 e9a9cb0c54461ffbda75b7b2f99f3ea5562291c2 Stephan Walter stephan@walter.name 2023-04-22T07:37:05Z GitHub noreply@github.com 2023-04-22T10:37:05+03:00 ggml : AVX2 optimization for vec_dot_q4_3_q8_0 and refactoring (#1099)
e9a9cb0c54461ffbda75b7b2f99f3ea5562291c2 b6e7f9b09e9c340ec97a2fae61c1eb8db861f2f9 Clint Herron hanclinto@gmail.com 2023-04-22T02:54:33-04:00 GitHub noreply@github.com 2023-04-22T09:54:33+03:00 examples : Improve Alpaca Default Repeat Penalty: Better Match Alpaca.cpp Experience (#1107)
b6e7f9b09e9c340ec97a2fae61c1eb8db861f2f9 50cb666b8a2e35a49b08c0f6bc81138c8f6f2ac1 xaedes xaedes@gmail.com 2023-04-22T08:21:32+02:00 GitHub noreply@github.com 2023-04-22T09:21:32+03:00 llama : add api for getting/setting the complete state: rng, logits, embedding and kv_cache (#1105)
50cb666b8a2e35a49b08c0f6bc81138c8f6f2ac1 25d7abbd1f73582b7e0fdc422a936e8541c0780b slaren 2141330+slaren@users.noreply.github.com 2023-04-21T21:59:17+02:00 GitHub noreply@github.com 2023-04-21T21:59:17+02:00 Improve cuBLAS performance by using a memory pool (#1094)
25d7abbd1f73582b7e0fdc422a936e8541c0780b 018f2279f5fe3ef743bd8254b23ea8f0efae7e73 apaz aarpazdera@gmail.com 2023-04-21T13:48:06-05:00 GitHub noreply@github.com 2023-04-21T21:48:06+03:00 llama : fixed rlimit error message (#888)
018f2279f5fe3ef743bd8254b23ea8f0efae7e73 9411288271ab548216902a029f42a0a38ebcedb7 源文雨 41315874+fumiama@users.noreply.github.com 2023-04-22T02:27:06+08:00 GitHub noreply@github.com 2023-04-21T21:27:06+03:00 cmake : link threads publicly to ggml (#1042)
9411288271ab548216902a029f42a0a38ebcedb7 8687c1f2581d059cd5b6a9502f89bd343566062a Alex Klinkhamer from.github.com.917@grencez.dev 2023-04-21T11:18:09-07:00 GitHub noreply@github.com 2023-04-21T21:18:09+03:00 main : evaluate tokens in batches after swapping context (#1014)
8687c1f2581d059cd5b6a9502f89bd343566062a 1bfc153e2f35ddd9d64b084e8d1a5e6fa57ad1c9 xaedes xaedes@googlemail.com 2023-04-21T17:25:21+02:00 GitHub noreply@github.com 2023-04-21T18:25:21+03:00 llama : remember and restore kv cache data pointers (#1104)
1bfc153e2f35ddd9d64b084e8d1a5e6fa57ad1c9 3d59769c3bb7e72c915646ddb1e239b1face19f5 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-04-21T17:18:26+02:00 GitHub noreply@github.com 2023-04-21T18:18:26+03:00 ggml : a faster version for Q4_1 x Q8_0 dot products (#1083)
3d59769c3bb7e72c915646ddb1e239b1face19f5 d40fded93e1a533e969768e1e335c15c61c296ce slaren 2141330+slaren@users.noreply.github.com 2023-04-21T14:57:57+02:00 GitHub noreply@github.com 2023-04-21T14:57:57+02:00 Show perplexity ETA in hours and minutes (#1096)
d40fded93e1a533e969768e1e335c15c61c296ce 2510c1831fac874f32e272f6079f01b5461f3986 Georgi Gerganov ggerganov@gmail.com 2023-04-21T10:23:36+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-21T10:24:02+03:00 llama : fix comment for "output.weight" tensor
2510c1831fac874f32e272f6079f01b5461f3986 12b5900dbc9743dee3ce83513cf5c3a44523a1b6 Stephan Walter stephan@walter.name 2023-04-20T21:56:44Z GitHub noreply@github.com 2023-04-20T23:56:44+02:00 Add ggml-model-*.bin checksums for 7B, 13B, 30B, 65B (#1088)
12b5900dbc9743dee3ce83513cf5c3a44523a1b6 9ff334f3c9b960a44c5e149b08c748a2914fb882 Georgi Gerganov ggerganov@gmail.com 2023-04-20T23:32:59+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-20T23:32:59+03:00 ggml : sync ggml (add GPT-NeoX RoPE implementation)
9ff334f3c9b960a44c5e149b08c748a2914fb882 2005469ea130cf920c50175d4f47a87bfd8aaf4d Georgi Gerganov ggerganov@gmail.com 2023-04-20T21:58:05+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-20T21:58:38+03:00 ggml : fix bug in ggml_compute_forward_dup_f32()
2005469ea130cf920c50175d4f47a87bfd8aaf4d 8a1756abdf1f48cb4dcb898bc8fbe9102ef49dc6 slaren 2141330+slaren@users.noreply.github.com 2023-04-20T20:49:53+02:00 GitHub noreply@github.com 2023-04-20T20:49:53+02:00 Add Q4_3 support to cuBLAS (#1086)
8a1756abdf1f48cb4dcb898bc8fbe9102ef49dc6 66aab46079609972ee1f7bd6f319d826205a2fbd Georgi Gerganov ggerganov@gmail.com 2023-04-20T21:43:50+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-20T21:43:50+03:00 ggml : do not break cuBLAS build (Q4_3 is not yet implemented)
66aab46079609972ee1f7bd6f319d826205a2fbd 38de86a7114c97ecf3644e3a60159f1ed893e1b0 Georgi Gerganov ggerganov@gmail.com 2023-04-20T20:44:05+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-20T20:44:05+03:00 ggml : fix Q4_3 quantization
38de86a7114c97ecf3644e3a60159f1ed893e1b0 e0305ead3a072db9c08b35c9600c49273b38a4b5 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-04-20T19:42:27+02:00 GitHub noreply@github.com 2023-04-20T20:42:27+03:00 llama : multi-threaded quantization (#1075)
e0305ead3a072db9c08b35c9600c49273b38a4b5 6a9661ea5ad72166b700ae5e87976e4452499dda Georgi Gerganov ggerganov@gmail.com 2023-04-20T20:35:53+03:00 GitHub noreply@github.com 2023-04-20T20:35:53+03:00 ggml : add Q4_3 quantization (#1082)
6a9661ea5ad72166b700ae5e87976e4452499dda 5addcb120cf2682c7ede0b1c520592700d74c87c Ivan Komarov Ivan.Komarov@dfyz.info 2023-04-20T17:15:18+02:00 GitHub noreply@github.com 2023-04-20T18:15:18+03:00 ci : remove the LLAMA_ACCELERATE matrix dimension from Ubuntu builds in the CI (#1074)
5addcb120cf2682c7ede0b1c520592700d74c87c c8c2c524827be8fd681a63f0e5a697b0bf4c587b 源文雨 41315874+fumiama@users.noreply.github.com 2023-04-20T21:28:43+08:00 GitHub noreply@github.com 2023-04-20T15:28:43+02:00 fix: LLAMA_CUBLAS=1 undefined reference 'shm_open' (#1080)
c8c2c524827be8fd681a63f0e5a697b0bf4c587b 02d6988121510c067e06d498a273a351a888f5b9 Stephan Walter stephan@walter.name 2023-04-20T06:45:41Z GitHub noreply@github.com 2023-04-20T08:45:41+02:00 AVX2 optimization for vec_dot_q4_2_q8_0 (#1068)
02d6988121510c067e06d498a273a351a888f5b9 834695fe3a3ed2a962e774c9615e3f7b41d360a8 slaren 2141330+slaren@users.noreply.github.com 2023-04-20T03:14:14+02:00 GitHub noreply@github.com 2023-04-20T03:14:14+02:00 Improve cuBLAS performance by dequantizing on the GPU (#1065)
834695fe3a3ed2a962e774c9615e3f7b41d360a8 f7d05095b404b5500b4a702ea16f67fc22446e49 CRD716 crd716@gmail.com 2023-04-19T14:52:14-05:00 GitHub noreply@github.com 2023-04-19T19:52:14Z Minor: Readme fixed grammar, spelling, and misc updates (#1071)
f7d05095b404b5500b4a702ea16f67fc22446e49 884e7d7a2bfd7325b107442d6758983f5886ed3d Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-04-19T20:20:14+02:00 GitHub noreply@github.com 2023-04-19T20:20:14+02:00 Q4_2 quantization with rmse-optimized scale and quants (#1062)
884e7d7a2bfd7325b107442d6758983f5886ed3d 7cd5c4a3e9106151d48f328bb3c94c298a211f18 Georgi Gerganov ggerganov@gmail.com 2023-04-19T20:10:08+03:00 GitHub noreply@github.com 2023-04-19T20:10:08+03:00 ggml : use 8-bit precision for Q4_1 intermediate results (#1047)
7cd5c4a3e9106151d48f328bb3c94c298a211f18 f3d4edf504c19ee9d1381c5727fe38667205d979 Georgi Gerganov ggerganov@gmail.com 2023-04-19T19:07:54+03:00 GitHub noreply@github.com 2023-04-19T19:07:54+03:00 readme : add warning about Q4_2 and Q4_3
f3d4edf504c19ee9d1381c5727fe38667205d979 8944a1329648c57bb7d66851170938230587a52c Stephan Walter stephan@walter.name 2023-04-19T16:06:37Z GitHub noreply@github.com 2023-04-19T19:06:37+03:00 ggml : Q4 cleanup - remove 4-bit dot product code (#1061)
8944a1329648c57bb7d66851170938230587a52c 66674012389f9537140044290f8517bc4a5e0b74 slaren 2141330+slaren@users.noreply.github.com 2023-04-19T11:22:45+02:00 GitHub noreply@github.com 2023-04-19T11:22:45+02:00 Add NVIDIA cuBLAS support (#1044)
66674012389f9537140044290f8517bc4a5e0b74 77a73403ca8eaced2590559d0f9cebd2b3649d32 slaren 2141330+slaren@users.noreply.github.com 2023-04-19T00:53:24+02:00 GitHub noreply@github.com 2023-04-19T00:53:24+02:00 Multi-threaded ggml_cpy (#1035)
77a73403ca8eaced2590559d0f9cebd2b3649d32 50a8a2af97cb92e53e7a3195aa201c3d87da5415 Georgi Gerganov ggerganov@gmail.com 2023-04-18T23:54:57+03:00 GitHub noreply@github.com 2023-04-18T23:54:57+03:00 ggml : add new Q4_2 quantization (ARM only) (#1046)
50a8a2af97cb92e53e7a3195aa201c3d87da5415 4caebf6d408b91c2d29d0abc7b1e867b5de64db5 Georgi Gerganov ggerganov@gmail.com 2023-04-18T23:11:23+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-18T23:11:23+03:00 ggml : scratch that - vmlaq_n_f32 is always better
4caebf6d408b91c2d29d0abc7b1e867b5de64db5 dcdd65e2969bc03c91a1ebd1160162d5054e6923 Georgi Gerganov ggerganov@gmail.com 2023-04-18T23:00:08+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-18T23:00:08+03:00 gitignore : vdot
dcdd65e2969bc03c91a1ebd1160162d5054e6923 5ecff35151156118c2df74899637ad34ee384b9b Georgi Gerganov ggerganov@gmail.com 2023-04-18T22:59:17+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-18T22:59:17+03:00 ggml : optimize ggml_vec_dot_q4_0_q8_0() using vectorized accumulators
5ecff35151156118c2df74899637ad34ee384b9b 7faa7460f03bdd88becf1e659cf359f274055404 Kawrakow 48489457+ikawrakow@users.noreply.github.com 2023-04-18T21:00:14+02:00 GitHub noreply@github.com 2023-04-18T19:00:14Z Adding a simple program to measure speed of dot products (#1041)
7faa7460f03bdd88becf1e659cf359f274055404 5af8e32238c7d9c4cdb7fc640472c9a26538b9da Georgi Gerganov ggerganov@gmail.com 2023-04-18T20:10:26+03:00 GitHub noreply@github.com 2023-04-18T20:10:26+03:00 readme : update hot topics about new LoRA functionality
5af8e32238c7d9c4cdb7fc640472c9a26538b9da 42747220b4cac548b6e3059b66b3e960b517cfa4 Georgi Gerganov ggerganov@gmail.com 2023-04-17T18:00:10+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-18T19:57:06+03:00 ci : do not run on drafts
42747220b4cac548b6e3059b66b3e960b517cfa4 e9298af3896b536d0c6d740447dc764e56b22102 Ivan Komarov Ivan.Komarov@dfyz.info 2023-04-18T03:15:50+02:00 GitHub noreply@github.com 2023-04-18T03:15:50+02:00 Do not close file after mmap (Windows version) (#1034)
e9298af3896b536d0c6d740447dc764e56b22102 4ad73137a1aadc40a62f7101aab883f69e7172c6 Atsushi Tatsuma yoshoku@outlook.com 2023-04-18T04:34:35+09:00 GitHub noreply@github.com 2023-04-17T22:34:35+03:00 readme : add Ruby bindings (#1029)
4ad73137a1aadc40a62f7101aab883f69e7172c6 315a95a4d30db726fb7d244dd3b9e90a83fb1616 Cameron csteele@steelecameron.com 2023-04-17T11:26:23-07:00 GitHub noreply@github.com 2023-04-17T20:26:23+02:00 add 4_0 to default outfile namestr dict (#1031)
315a95a4d30db726fb7d244dd3b9e90a83fb1616 efd05648c88a0923a55f56e7ce1b0f9c33410afb slaren 2141330+slaren@users.noreply.github.com 2023-04-17T17:28:55+02:00 GitHub noreply@github.com 2023-04-17T17:28:55+02:00 Add LoRA support (#820)
efd05648c88a0923a55f56e7ce1b0f9c33410afb eb17a026fd23d1c1b612fa4600f7f5c58e501a28 Arik Poznanski arikpoz@users.noreply.github.com 2023-04-17T17:41:53+03:00 GitHub noreply@github.com 2023-04-17T17:41:53+03:00 llama : well-defined static initialization of complex objects (#927)
eb17a026fd23d1c1b612fa4600f7f5c58e501a28 69b740289f9b3756ea9dd2a23f241c6f688d88b9 Georgi Gerganov ggerganov@gmail.com 2023-04-17T17:31:06+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-17T17:31:06+03:00 quantize-stats : fix bug in --type argument
69b740289f9b3756ea9dd2a23f241c6f688d88b9 f266259ad9a2bce5a34d919592310147af23f3dc Georgi Gerganov ggerganov@gmail.com 2023-04-17T16:16:23+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-17T16:16:23+03:00 ggml : avoid using ggml_fp16_to_fp32() and ggml_fp32_to_fp16() in ggml.c
f266259ad9a2bce5a34d919592310147af23f3dc 47f61aaa5f76d04286792e2fbd0c95b659ab2af0 Ivan Komarov Ivan.Komarov@dfyz.info 2023-04-17T15:10:57+02:00 GitHub noreply@github.com 2023-04-17T15:10:57+02:00 Speedup the AVX-512 implementation of ggml_vec_dot_q4_0() (#933)
47f61aaa5f76d04286792e2fbd0c95b659ab2af0 3173a62eb9f90b94fb3184131032c1c8b7aa8d86 slaren 2141330+slaren@users.noreply.github.com 2023-04-16T21:27:38+02:00 GitHub noreply@github.com 2023-04-16T21:27:38+02:00 Fix: do not close file on mmap (#1017)
3173a62eb9f90b94fb3184131032c1c8b7aa8d86 489537e6cf6c93b74a029a11533dbcaa89791dcc Georgi Gerganov ggerganov@gmail.com 2023-04-16T13:58:48+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-16T13:59:27+03:00 stdout : vertical align outputs for better readibility
489537e6cf6c93b74a029a11533dbcaa89791dcc 2d3481c72125cd388258864c7ad8d7d36777bad7 Pavol Rusnak pavol@rusnak.io 2023-04-16T12:13:00+02:00 GitHub noreply@github.com 2023-04-16T10:13:00Z examples: add missing <ctime> include for time() (#1011)
2d3481c72125cd388258864c7ad8d7d36777bad7 74f5899df4a6083fc467b620baa1cf821e37799d nanahi 130121847+na-na-hi@users.noreply.github.com 2023-04-16T17:13:42+08:00 GitHub noreply@github.com 2023-04-16T11:13:42+02:00 Fix msys2 build error and warnings (#1009)
74f5899df4a6083fc467b620baa1cf821e37799d 2f7c8e014e3c0ceaf39688845c2ff6f919fb03b7 comex comexk@gmail.com 2023-04-15T14:53:21-07:00 GitHub noreply@github.com 2023-04-15T23:53:21+02:00 convert.py: Fix loading safetensors and ggml format on Windows (#991)
2f7c8e014e3c0ceaf39688845c2ff6f919fb03b7 0ad964631f9b3970f1936008fcfb1eadef59c7ed Stephan Walter stephan@walter.name 2023-04-15T18:28:56Z GitHub noreply@github.com 2023-04-15T18:28:56Z Fix potential int8 overflow in non-SIMD vec_dot (#986)
0ad964631f9b3970f1936008fcfb1eadef59c7ed e95b6554b493e71a0275764342e09bd5784a7026 Stephan Walter stephan@walter.name 2023-04-15T16:25:38Z GitHub noreply@github.com 2023-04-15T16:25:38Z Refactor ggml.c for future tensor types (#1001)
e95b6554b493e71a0275764342e09bd5784a7026 aa485cee334e84437e21681c14b6f80b65876d8b Georgi Gerganov ggerganov@gmail.com 2023-04-15T17:53:22+03:00 GitHub noreply@github.com 2023-04-15T17:53:22+03:00 ggml : add Q8_0 quantization for intermediate results (#951)
aa485cee334e84437e21681c14b6f80b65876d8b c12b14b77fced0ce9a0e2d81f670c3a746dec251 Georgi Gerganov ggerganov@gmail.com 2023-04-15T14:25:45+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-15T14:25:45+03:00 ggml : use posix_memalign on non-Windows env
c12b14b77fced0ce9a0e2d81f670c3a746dec251 106faaf2971d6c89d6010279a9a95737772470ef Ivan Komarov Ivan.Komarov@dfyz.info 2023-04-15T07:51:54+02:00 GitHub noreply@github.com 2023-04-15T08:51:54+03:00 benchmark : fix result validation in benchmark-q4_0-matmult (#987)
106faaf2971d6c89d6010279a9a95737772470ef c85e03d12e4b8af22cb13aa9c618dcd5935862fd katsu560 118887472+katsu560@users.noreply.github.com 2023-04-15T14:51:11+09:00 GitHub noreply@github.com 2023-04-15T08:51:11+03:00 cmake : add finding the OpenBLAS header file (#992)
c85e03d12e4b8af22cb13aa9c618dcd5935862fd 489093548c89c67520109ab25c4df4a4614a32a0 Pavol Rusnak pavol@rusnak.io 2023-04-14T21:58:43+02:00 GitHub noreply@github.com 2023-04-14T22:58:43+03:00 Revert "main : alternative instruct mode (Vicuna support, etc.) (#863)" (#982)
489093548c89c67520109ab25c4df4a4614a32a0 93265e988af32b8be314bfed334f795a3037555d Pavol Rusnak pavol@rusnak.io 2023-04-14T21:46:49+02:00 GitHub noreply@github.com 2023-04-14T19:46:49Z py : bump sentencepiece to 0.1.98 to support Python 3.11 (#976)
93265e988af32b8be314bfed334f795a3037555d c56b7152690ca25cfd66b20210b3629e6c1e739b Stephan Walter stephan@walter.name 2023-04-14T19:39:48Z GitHub noreply@github.com 2023-04-14T22:39:48+03:00 make : fix dependencies, use auto variables (#983)
c56b7152690ca25cfd66b20210b3629e6c1e739b f4d277ae17247ee51129ef1a9ff74d377cc90b1b Pavol Rusnak pavol@rusnak.io 2023-04-14T20:05:37+02:00 GitHub noreply@github.com 2023-04-14T20:05:37+02:00 Expose type name from ggml (#970)
f4d277ae17247ee51129ef1a9ff74d377cc90b1b c9a59b70a54e0bc05777df287feaea3dbe0310c4 Tomáš Pazdiora tomas.pazdiora@gmail.com 2023-04-14T17:19:17+02:00 GitHub noreply@github.com 2023-04-14T18:19:17+03:00 main : alternative instruct mode (Vicuna support, etc.) (#863)
c9a59b70a54e0bc05777df287feaea3dbe0310c4 a32f7acc9f54dba1c728cb1e596bd00bf3b4eb5f Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-04-14T08:43:55-06:00 GitHub noreply@github.com 2023-04-14T17:43:55+03:00 ggml : add unary and binary map operations (#874)
a32f7acc9f54dba1c728cb1e596bd00bf3b4eb5f 43ffdefb7424f79a3d510c199e2ea86684b4f824 Pavol Rusnak pavol@rusnak.io 2023-04-14T15:37:11+02:00 GitHub noreply@github.com 2023-04-14T15:37:11+02:00 py : cleanup dependencies (#962)
43ffdefb7424f79a3d510c199e2ea86684b4f824 1623a6e9b46453bff30afd7d0f6c3fd188499c2f Pavol Rusnak pavol@rusnak.io 2023-04-14T14:23:21+02:00 GitHub noreply@github.com 2023-04-14T14:23:21+02:00 py : fix flake8 and isort nitpicks (#960)
1623a6e9b46453bff30afd7d0f6c3fd188499c2f c14e0d2f23e6d1e785255f4da8c253c1b4723659 Georgi Gerganov ggerganov@gmail.com 2023-04-14T13:31:29+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-14T13:31:29+03:00 ggml : minor
c14e0d2f23e6d1e785255f4da8c253c1b4723659 723dac55fa2ba7adc6e3fc8609781d1ad0378906 Georgi Gerganov ggerganov@gmail.com 2023-04-14T13:31:15+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-14T13:31:15+03:00 ggml : always allocate buffers with size multiple of GGML_MEM_ALIGN
723dac55fa2ba7adc6e3fc8609781d1ad0378906 0f07cacb05f49704d35a39aa27cfd4b419eb6f8d comex comexk@gmail.com 2023-04-14T00:03:03-07:00 GitHub noreply@github.com 2023-04-14T10:03:03+03:00 py : new conversion script (#545)
0f07cacb05f49704d35a39aa27cfd4b419eb6f8d c5d70f5c9ea5a8f0f6b0d6aa741455978a1dabfd Georgi Gerganov ggerganov@gmail.com 2023-04-14T09:45:42+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-14T09:45:42+03:00 ggml : fix q4_1 dot product types
c5d70f5c9ea5a8f0f6b0d6aa741455978a1dabfd be87b6ed20a5f7528bf491a83e759a9fc6a24fea Howard Su howard0su@gmail.com 2023-04-14T14:24:52+08:00 GitHub noreply@github.com 2023-04-14T09:24:52+03:00 ggml : optimize rope function to avoid call powf in the tight loop (#807)
be87b6ed20a5f7528bf491a83e759a9fc6a24fea 0e07e6a8399fd993739a3ba3c6f95f92bfab6f58 Gary Linscott glinscott@gmail.com 2023-04-13T14:50:42-07:00 GitHub noreply@github.com 2023-04-14T00:50:42+03:00 perplexity : add support for batch size to `--perplexity` (#407)
0e07e6a8399fd993739a3ba3c6f95f92bfab6f58 a3a2a0eda8828b60436e9f69d9ac2c1060d03e7a CRD716 crd716@gmail.com 2023-04-13T10:39:25-05:00 GitHub noreply@github.com 2023-04-13T18:39:25+03:00 common : remove unnecessary includes (#947)
a3a2a0eda8828b60436e9f69d9ac2c1060d03e7a d990e3fffc5b0f5448e90a16c79a4f2675100af0 Georgi Gerganov ggerganov@gmail.com 2023-04-13T18:36:40+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-13T18:36:48+03:00 ggml : add GGML_DEFAULT_N_THREADS
d990e3fffc5b0f5448e90a16c79a4f2675100af0 9190e8eac8bdc108c40d2d7505e9b45fa773251f Georgi Gerganov ggerganov@gmail.com 2023-04-13T18:32:36+03:00 GitHub noreply@github.com 2023-04-13T18:32:36+03:00 ggml : speed-up ggml_vec_dot_q4_1() ARM_NEON + 32-bit ARM support (#900)
9190e8eac8bdc108c40d2d7505e9b45fa773251f c85980acd04631a7c43d13676276f76ec72f5dfe Georgi Gerganov ggerganov@gmail.com 2023-04-13T18:04:45+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-13T18:04:45+03:00 llama : merge llama_internal.h into llama.h
c85980acd04631a7c43d13676276f76ec72f5dfe 6232f2d7fd7a22d5eeb62182b2f21fcf01359754 Georgi Gerganov ggerganov@gmail.com 2023-04-13T18:01:22+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-13T18:01:33+03:00 gitignore : benchmark
6232f2d7fd7a22d5eeb62182b2f21fcf01359754 6c248707f51c8a50f7792e7f7787ec481881db88 Stephan Walter stephan@walter.name 2023-04-13T14:59:50Z GitHub noreply@github.com 2023-04-13T17:59:50+03:00 ggml : optimize non-SIMD Q4_0 vector dot product (#703)
6c248707f51c8a50f7792e7f7787ec481881db88 8cda5c981d0bf4dcb7664194b2cb9a06e2dbdd54 Pavol Rusnak pavol@rusnak.io 2023-04-13T16:08:32+02:00 GitHub noreply@github.com 2023-04-13T17:08:32+03:00 ggml : introduce GGML_ALIGNED_MALLOC/GGML_ALIGNED_FREE macros (#884)
8cda5c981d0bf4dcb7664194b2cb9a06e2dbdd54 ec29272175d7a79681d9919f3e755b1bcefa0478 CRD716 crd716@gmail.com 2023-04-13T09:03:57-05:00 GitHub noreply@github.com 2023-04-13T16:03:57+02:00 fix whitespace (#944)
ec29272175d7a79681d9919f3e755b1bcefa0478 7e941b95eba067cb5b92785e642fd803657376ee CRD716 crd716@gmail.com 2023-04-13T08:59:53-05:00 GitHub noreply@github.com 2023-04-13T16:59:53+03:00 readme : remove python 3.10 warning (#929)
7e941b95eba067cb5b92785e642fd803657376ee c729ff730a46a135817a3d9988a097e3678a9722 Genkagaku.GPT hlhr202@163.com 2023-04-13T21:54:27+08:00 GitHub noreply@github.com 2023-04-13T16:54:27+03:00 readme : llama node binding (#911)
c729ff730a46a135817a3d9988a097e3678a9722 4579af95e8e16910f6dbab0994917a5b3901f0cf Pavol Rusnak pavol@rusnak.io 2023-04-13T15:49:05+02:00 GitHub noreply@github.com 2023-04-13T15:49:05+02:00 flake.nix: add all binaries from bin (#848)
4579af95e8e16910f6dbab0994917a5b3901f0cf 8c3ffc2f048a372639906fb30ec3c2070288d3be Judd foldl@users.noreply.github.com 2023-04-13T21:43:22+08:00 GitHub noreply@github.com 2023-04-13T16:43:22+03:00 zig : update build.zig (#872)
8c3ffc2f048a372639906fb30ec3c2070288d3be 107980d970808c2ccf9334ad033e2782a560b911 Vladimir bogdad@gmail.com 2023-04-13T15:24:30+02:00 GitHub noreply@github.com 2023-04-13T16:24:30+03:00 ggml : update cblas_sgemm columns var to be more reasonable (#838)
107980d970808c2ccf9334ad033e2782a560b911 585d91a156794d30eec16ebe67c8d7a1d41406c1 niansa/tuxifan anton-sa@web.de 2023-04-13T15:03:39+02:00 GitHub noreply@github.com 2023-04-13T16:03:39+03:00 examples : add -n to alpaca and gpt4all scripts (#706)
585d91a156794d30eec16ebe67c8d7a1d41406c1 95ea26f6e92d620a5437f576b80868aee7f808d6 anzz1 anzz1@live.com 2023-04-13T15:48:21+03:00 GitHub noreply@github.com 2023-04-13T15:48:21+03:00 cmake : add explicit F16C option (x86) (#576)
95ea26f6e92d620a5437f576b80868aee7f808d6 82d146df9b43cf677e0dbce20b03cf864958a0cc SebastianApel 13675545+SebastianApel@users.noreply.github.com 2023-04-13T14:46:23+02:00 GitHub noreply@github.com 2023-04-13T15:46:23+03:00 benchmark : add tool for timing q4_0 matrix multiplication (#653)
82d146df9b43cf677e0dbce20b03cf864958a0cc e7f6997f897a18b6372a6460e25c5f89e1469f1d Pavol Rusnak pavol@rusnak.io 2023-04-13T11:33:16+02:00 GitHub noreply@github.com 2023-04-13T11:33:16+02:00 do not force the prompt file to end with a new line (#908)
e7f6997f897a18b6372a6460e25c5f89e1469f1d f76cb3a34d6a6b03afb96650e39495f201eac042 Stephan Walter stephan@walter.name 2023-04-12T15:06:16Z GitHub noreply@github.com 2023-04-12T15:06:16Z Don't crash on ftype (formerly f16) == 4 (#917)
f76cb3a34d6a6b03afb96650e39495f201eac042 782438070f7568380755ffc7bf5e09b20c1e8272 Georgi Gerganov ggerganov@gmail.com 2023-04-12T14:48:57+03:00 GitHub noreply@github.com 2023-04-12T14:48:57+03:00 readme : change "GPU support" link to discussion
782438070f7568380755ffc7bf5e09b20c1e8272 4dbbd407500cf500ca5f47e4e947635797997c05 Georgi Gerganov ggerganov@gmail.com 2023-04-12T14:31:12+03:00 GitHub noreply@github.com 2023-04-12T14:31:12+03:00 readme : update hot topics with link to "GPU support" issue
4dbbd407500cf500ca5f47e4e947635797997c05 8b679987cdce292ff36bd741f6715e4927e26f9b Nicolai Weitkemper kontakt@nicolaiweitkemper.de 2023-04-12T08:46:20+02:00 GitHub noreply@github.com 2023-04-12T08:46:20+02:00 readme: link to sha256sums file (#902)
8b679987cdce292ff36bd741f6715e4927e26f9b 3e6e70d8e8917b5bd14c7c9f9b89a585f1ff0b31 Pavol Rusnak pavol@rusnak.io 2023-04-11T21:45:44+02:00 GitHub noreply@github.com 2023-04-11T19:45:44Z Fix whitespace, add .editorconfig, add GitHub workflow (#883)
3e6e70d8e8917b5bd14c7c9f9b89a585f1ff0b31 2663d2c6784ad7b77998c6874df25648d597f74b Stephan Walter stephan@walter.name 2023-04-11T15:03:51Z GitHub noreply@github.com 2023-04-11T15:03:51Z Add enum llama_ftype, sync ggml_type to model files (#709)
2663d2c6784ad7b77998c6874df25648d597f74b a0caa34b162449b5c13b8d604573053300ff54a1 comex comexk@gmail.com 2023-04-11T06:19:54-07:00 GitHub noreply@github.com 2023-04-11T15:19:54+02:00 Windows fixes (#890)
a0caa34b162449b5c13b8d604573053300ff54a1 461ba9e66ed3885f80680d71495e055580573c74 qouoq qouoq@fastmail.com 2023-04-11T04:41:53+08:00 GitHub noreply@github.com 2023-04-10T22:41:53+02:00 Add BAIR's Koala to supported models (#877)
461ba9e66ed3885f80680d71495e055580573c74 c3ac702e5ee3533457e0489df4906ee112fe88e7 Georgi Gerganov ggerganov@gmail.com 2023-04-10T23:20:01+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-10T23:20:01+03:00 ggml : fix WASM build
c3ac702e5ee3533457e0489df4906ee112fe88e7 9d634ef452d0fc24fcd49592952d13d0ab0f41b7 Georgi Gerganov ggerganov@gmail.com 2023-04-10T22:40:28+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-10T22:42:28+03:00 ggml : add ggml_cont() + optimize ggml_cpy() for contiguous dst
9d634ef452d0fc24fcd49592952d13d0ab0f41b7 d9a239c4104c888eafda672c1e42c9bbc5084cb8 Georgi Gerganov ggerganov@gmail.com 2023-04-10T19:32:45+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-10T22:42:28+03:00 ggml : remove trailing whitespaces
d9a239c4104c888eafda672c1e42c9bbc5084cb8 684da25926e5c505f725b4f10b5485b218fa1fc7 Marco Matthies 71844+marcom@users.noreply.github.com 2023-04-10T19:57:59+02:00 GitHub noreply@github.com 2023-04-10T19:57:59+02:00 Simplify to include lower-case windows.h always, fix compile on mingw32 (#747)
684da25926e5c505f725b4f10b5485b218fa1fc7 180b693a47b6b825288ef9f2c39d24b6eea4eea6 Georgi Gerganov ggerganov@gmail.com 2023-04-10T19:29:48+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-10T19:29:48+03:00 ggml : fix quantize_row_q4_1() ARM_NEON (close #876)
180b693a47b6b825288ef9f2c39d24b6eea4eea6 f963b63afa0e057cfb9eba4d88407c6a0850a0d8 comex comexk@gmail.com 2023-04-08T13:08:21-07:00 Matvey Soloviev blackhole89@gmail.com 2023-04-10T01:10:46+02:00 Print model version.
f963b63afa0e057cfb9eba4d88407c6a0850a0d8 aaf3b23debc1fe1a06733c8c6468fb84233cc44f comex comexk@gmail.com 2023-04-08T12:24:37-07:00 Matvey Soloviev blackhole89@gmail.com 2023-04-10T01:10:46+02:00 Rewrite loading code to try to satisfy everyone:
aaf3b23debc1fe1a06733c8c6468fb84233cc44f f2d1c472946dee2aba9077e8df73346796752b10 Tomáš Pazdiora tomas.pazdiora@gmail.com 2023-04-08T17:49:39+02:00 GitHub noreply@github.com 2023-04-08T17:49:39+02:00 fix for windows utf-8 input (#840)
f2d1c472946dee2aba9077e8df73346796752b10 317fb12fbd7cef5d86476574bffe0e904af884ca eiery 19350831+eiery@users.noreply.github.com 2023-04-08T07:15:17-04:00 GitHub noreply@github.com 2023-04-08T11:15:17Z cmake should link openblas properly with -lopenblas like how it's done in the makefile (#839)
317fb12fbd7cef5d86476574bffe0e904af884ca 62cfc54f77e519057110265b52b0d614fa363e2a lon 114724657+longregen@users.noreply.github.com 2023-04-08T07:04:23-03:00 GitHub noreply@github.com 2023-04-08T12:04:23+02:00 Add new binaries to flake.nix (#847)
62cfc54f77e519057110265b52b0d614fa363e2a 698f7b5d6316a1f8453b3b32fd0d637d24952ffd unbounded haakon@likedan.net 2023-04-08T00:09:18+02:00 GitHub noreply@github.com 2023-04-08T00:09:18+02:00 Add quantize-stats command for testing quantization (#728)
698f7b5d6316a1f8453b3b32fd0d637d24952ffd c1950c343109ab1fd15fc2ae1c83650c85d4eeef bhubbb 79117352+bhubbb@users.noreply.github.com 2023-04-08T02:11:58+10:00 GitHub noreply@github.com 2023-04-07T19:11:58+03:00 make : add libllama.so target for llama-cpp-python (#797)
c1950c343109ab1fd15fc2ae1c83650c85d4eeef 4953e9007f86327aabc8312a7211c18019a3a40e iacore 74560659+iacore@users.noreply.github.com 2023-04-07T16:05:29Z GitHub noreply@github.com 2023-04-07T19:05:29+03:00 zig : don't link examples/common.cpp for non-example (#814)
4953e9007f86327aabc8312a7211c18019a3a40e cc9cee8e9e7598bd280295f6264f36d3a9224006 Ivan Stepanov ivanstepanovftw@gmail.com 2023-04-07T19:02:12+03:00 GitHub noreply@github.com 2023-04-07T19:02:12+03:00 llama : always sort logits before nucleus sampling (#812)
cc9cee8e9e7598bd280295f6264f36d3a9224006 d2beca95dcfcd6f1145886e914b879ffc3604b7a Sergey Alirzaev zl29ah@gmail.com 2023-04-06T17:59:11+02:00 GitHub noreply@github.com 2023-04-06T17:59:11+02:00 Do not crash when it has nothing to say. (#796)
d2beca95dcfcd6f1145886e914b879ffc3604b7a eeaa7b0492fc79baab8bb1fe195d6c87159f2bd3 Pavol Rusnak pavol@rusnak.io 2023-04-06T08:56:58+02:00 GitHub noreply@github.com 2023-04-06T08:56:58+02:00 Make docker instructions more explicit (#785)
eeaa7b0492fc79baab8bb1fe195d6c87159f2bd3 986b6ce9f99503c51ec5afd8a10baa32359434c6 Georgi Gerganov ggerganov@gmail.com 2023-04-05T22:11:03+03:00 GitHub noreply@github.com 2023-04-05T22:11:03+03:00 ggml : multi-thread ggml_rope() (~3-4 times faster on M1) (#781)
986b6ce9f99503c51ec5afd8a10baa32359434c6 34162989297fdfe3ab7305451ce55bc87e3f4c9c Georgi Gerganov ggerganov@gmail.com 2023-04-05T22:07:33+03:00 GitHub noreply@github.com 2023-04-05T22:07:33+03:00 ggml, llama : avoid heavy V transpose + improvements (#775)
34162989297fdfe3ab7305451ce55bc87e3f4c9c 5a8c4f624077373a198cd562146ffa67b02ebc75 Georgi Gerganov ggerganov@gmail.com 2023-04-05T19:54:30+03:00 GitHub noreply@github.com 2023-04-05T19:54:30+03:00 Update README.md
5a8c4f624077373a198cd562146ffa67b02ebc75 ff05d05c960076b42f027e4d318cbd6ea59b3030 Ivan Stepanov ivanstepanovftw@gmail.com 2023-04-05T19:20:05+03:00 GitHub noreply@github.com 2023-04-05T19:20:05+03:00 llama : define non-positive top_k; top_k range check (#779)
ff05d05c960076b42f027e4d318cbd6ea59b3030 62b3e81aaeafb282934de8b21de13b0104f12f8c at8u 129688334+at8u@users.noreply.github.com 2023-04-05T15:59:13Z GitHub noreply@github.com 2023-04-05T18:59:13+03:00 miku.sh : add executable bit (#780)
62b3e81aaeafb282934de8b21de13b0104f12f8c 8d10406d6ee230e6c1a96590cc8273f86ae606f8 Georgi Gerganov ggerganov@gmail.com 2023-04-05T18:58:06+03:00 Georgi Gerganov ggerganov@gmail.com 2023-04-05T18:58:31+03:00 media : add logos and banners
8d10406d6ee230e6c1a96590cc8273f86ae606f8 ed1c214e667d58ac442b3c6664831fe0eed2941d Georgi Gerganov ggerganov@gmail.com 2023-04-05T18:56:20+03:00 GitHub noreply@github.com 2023-04-05T18:56:20+03:00 readme : change logo + add bindings + add uis + add wiki
ed1c214e667d58ac442b3c6664831fe0eed2941d 0c44427df10ee024b4e7ef7bfec56e993daff1db iacore 74560659+iacore@users.noreply.github.com 2023-04-05T15:06:02Z GitHub noreply@github.com 2023-04-05T18:06:02+03:00 zig : add build.zig (#773)
0c44427df10ee024b4e7ef7bfec56e993daff1db 594cc95fabab0b662dabba3ea619ca5dca18bf6b Ivan Stepanov ivanstepanovftw@gmail.com 2023-04-05T17:38:37+03:00 GitHub noreply@github.com 2023-04-05T17:38:37+03:00 make : missing host optimizations in CXXFLAGS (#763)
594cc95fabab0b662dabba3ea619ca5dca18bf6b 88ed5761b869a221bc26847ff3f6977e7ee6425e Adithya Balaji adithya.b94@gmail.com 2023-04-05T16:36:12+02:00 GitHub noreply@github.com 2023-04-05T17:36:12+03:00 readme : update with CMake and windows example (#748)
88ed5761b869a221bc26847ff3f6977e7ee6425e 58c438cf7dfbbef710b1905a453a38a8a9ced07d at8u 129688334+at8u@users.noreply.github.com 2023-04-05T14:32:42Z GitHub noreply@github.com 2023-04-05T17:32:42+03:00 examples : add Miku.sh (#724)
58c438cf7dfbbef710b1905a453a38a8a9ced07d 53dbba769537e894ead5c6913ab2fd3a4658b738 Andrew Duffy a10y@users.noreply.github.com 2023-04-05T11:44:24+01:00 GitHub noreply@github.com 2023-04-05T06:44:24-04:00 Add Accelerate/BLAS when using Swift (#765)
53dbba769537e894ead5c6913ab2fd3a4658b738 437e77855a54e69c86fe03bc501f63d9a3fddb0e mgroeber9110 45620825+mgroeber9110@users.noreply.github.com 2023-04-03T18:00:55+02:00 GitHub noreply@github.com 2023-04-03T18:00:55+02:00 Windows: reactive sigint handler after each Ctrl-C (#736)
437e77855a54e69c86fe03bc501f63d9a3fddb0e cd7fa956904cb8e321b72b3499f4a3a82e43c266 SebastianApel 13675545+SebastianApel@users.noreply.github.com 2023-04-03T09:52:28+02:00 GitHub noreply@github.com 2023-04-03T09:52:28+02:00 10+% performance improvement of ggml_vec_dot_q4_0 on AVX2 (#654)
cd7fa956904cb8e321b72b3499f4a3a82e43c266 a0c05164168297c04737936ad0cad849a512547a Ivan Stepanov ivanstepanovftw@gmail.com 2023-04-03T03:19:04+03:00 GitHub noreply@github.com 2023-04-03T02:19:04+02:00 Define non-positive temperature behavior (#720)
a0c05164168297c04737936ad0cad849a512547a d8d4e865cd481b18f10508ffee35db903767ef5c bsilvereagle bsilvereagle@users.noreply.github.com 2023-04-02T15:13:03-07:00 GitHub noreply@github.com 2023-04-03T00:13:03+02:00 Remove torch GPU dependencies from the Docker.full image (#665)
d8d4e865cd481b18f10508ffee35db903767ef5c e986f94829bae0b9e66b326acbbba179931c84f1 Thatcher Chamberlin j.thatcher.c@gmail.com 2023-04-02T06:48:57-04:00 GitHub noreply@github.com 2023-04-02T12:48:57+02:00 Add a missing step to the gpt4all instructions (#690)
e986f94829bae0b9e66b326acbbba179931c84f1 c0bb1d3ce21005ab21d686626ba87261a6e3a660 Christian Falch 875252+chrfalch@users.noreply.github.com 2023-04-02T12:23:04+02:00 GitHub noreply@github.com 2023-04-02T12:23:04+02:00 Added api for getting/setting the kv_cache (#685)
c0bb1d3ce21005ab21d686626ba87261a6e3a660 6e7801d08d81c931a5427bae46f00763e993f54a Marian Cepok marian.cepok@gmail.com 2023-04-02T12:21:31+02:00 GitHub noreply@github.com 2023-04-02T13:21:31+03:00 ggml : change ne to int64_t (#626)
6e7801d08d81c931a5427bae46f00763e993f54a 81040f10aae3160317c5787c9c59acb219927826 Leonardo Neumann leonardo@neumann.dev.br 2023-04-02T04:56:20-03:00 GitHub noreply@github.com 2023-04-02T10:56:20+03:00 examples : add gpt4all script (#658)
81040f10aae3160317c5787c9c59acb219927826 c4f89d8d73aab4318a6c61e3835135adfcf55407 Stephan Walter stephan@walter.name 2023-04-02T07:18:53Z GitHub noreply@github.com 2023-04-02T10:18:53+03:00 llama : do not allocate KV cache for "vocab_only == true" (#682)
c4f89d8d73aab4318a6c61e3835135adfcf55407 5b70e7de4c0b8186669d0c5609ba61a2d46de562 Fabian cmdrf@users.noreply.github.com 2023-04-02T09:17:05+02:00 GitHub noreply@github.com 2023-04-02T10:17:05+03:00 make : use -march=native -mtune=native on x86 (#609)
5b70e7de4c0b8186669d0c5609ba61a2d46de562 a717cba8440b380f43cd3e2510862fc1ea3de9a2 Murilo Santana mvrilo@gmail.com 2023-04-01T23:41:12-03:00 GitHub noreply@github.com 2023-04-02T04:41:12+02:00 fix default params for examples/main (#697)
a717cba8440b380f43cd3e2510862fc1ea3de9a2 d0a7f742e76bb48c0bd852f0b3bf09ec0b75b200 Ikko Eltociear Ashimine eltociear@gmail.com 2023-04-02T01:38:18+09:00 GitHub noreply@github.com 2023-04-01T18:38:18+02:00 py: huggingface -> Hugging Face (#686)
d0a7f742e76bb48c0bd852f0b3bf09ec0b75b200 0d054e292e5492981867be69c788edd04dc8adeb rimoliga 53384203+rimoliga@users.noreply.github.com 2023-04-01T11:57:30-03:00 GitHub noreply@github.com 2023-04-01T16:57:30+02:00 readme: replace termux links with homepage, play store is deprecated (#680)
0d054e292e5492981867be69c788edd04dc8adeb 3525899277d2e2bdc8ec3f0e6e40c47251608700 Slaren 2141330+slaren@users.noreply.github.com 2023-03-31T20:03:48+02:00 Pavol Rusnak pavol@rusnak.io 2023-04-01T16:08:40+02:00 Show error message when -f fails
3525899277d2e2bdc8ec3f0e6e40c47251608700 1d08882afa647c44195f4f6495a68ea455650cae Stephan Walter stephan@walter.name 2023-03-31T19:19:16Z GitHub noreply@github.com 2023-03-31T19:19:16Z Enable -std= for cmake builds, fix warnings (#598)
1d08882afa647c44195f4f6495a68ea455650cae 02c5b27e91a6d18cf1043d3a2d8dbc59610ac257 slaren 2141330+slaren@users.noreply.github.com 2023-03-31T17:55:52+02:00 GitHub noreply@github.com 2023-03-31T15:55:52Z Optimize AVX2 ggml_vec_dot_q4_0 (#642)
02c5b27e91a6d18cf1043d3a2d8dbc59610ac257 cbef542879962fdc491656cd0c8cadd65a5f1356 perserk perserk@gmail.com 2023-03-31T16:55:44+05:00 GitHub noreply@github.com 2023-03-31T13:55:44+02:00 Add AVX acceleration (#617)
cbef542879962fdc491656cd0c8cadd65a5f1356 9733104be5389ebb1ff05095eca2a70280cd875a Pavol Rusnak pavol@rusnak.io 2023-03-29T21:31:24+02:00 Pavol Rusnak pavol@rusnak.io 2023-03-31T10:32:01+02:00 py : cleanup the code
9733104be5389ebb1ff05095eca2a70280cd875a 3df890aef432ce68143cfafcd7caf828bc4c3e55 Pavol Rusnak pavol@rusnak.io 2023-03-31T00:52:06+02:00 Pavol Rusnak pavol@rusnak.io 2023-03-31T01:07:32+02:00 drop quantize.py (now that models are using a single file)
3df890aef432ce68143cfafcd7caf828bc4c3e55 ee0c40dd6de8c3c658ae43199939ef40bb1cf408 Georgi Gerganov ggerganov@gmail.com 2023-03-30T22:31:54+03:00 GitHub noreply@github.com 2023-03-30T22:31:54+03:00 readme : update supported models
ee0c40dd6de8c3c658ae43199939ef40bb1cf408 6f23ba5ee235cbcb1eedd63b98422dd8d4392a78 Justine Tunney jtunney@gmail.com 2023-03-30T05:42:56-07:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Introduce GGML migration tool for new file format
6f23ba5ee235cbcb1eedd63b98422dd8d4392a78 78ca9838ee36660a776e97e3391b6fb5dcaacf7f Justine Tunney jtunney@gmail.com 2023-03-30T01:53:36-07:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Ensure --mlock works properly with mmap() support
78ca9838ee36660a776e97e3391b6fb5dcaacf7f a017390358cdb23fffb30988dc84bb190d0403ca Justine Tunney jtunney@gmail.com 2023-03-29T13:51:37-07:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Make loading weights 10-100x faster
a017390358cdb23fffb30988dc84bb190d0403ca ac184d514723902f9b05b688703b1be6e8dc65de Slaren 2141330+slaren@users.noreply.github.com 2023-03-29T22:22:36+02:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Initial windows support (untested)
ac184d514723902f9b05b688703b1be6e8dc65de 276e5b781155e3bbe6834472c58f03dfe62efabe Slaren 2141330+slaren@users.noreply.github.com 2023-03-29T08:53:14+02:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Always initialize mm_addr and mm_length in llama_model
276e5b781155e3bbe6834472c58f03dfe62efabe d68c5dc4356c8f49e933df210f2ceca5002a8118 Slaren 2141330+slaren@users.noreply.github.com 2023-03-29T08:31:26+02:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Unmap the file in llama_free
d68c5dc4356c8f49e933df210f2ceca5002a8118 64bde3ffd4aef799acb790a3eedddbd0a0612108 Slaren 2141330+slaren@users.noreply.github.com 2023-03-29T06:18:18+02:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Make mmap_file static
64bde3ffd4aef799acb790a3eedddbd0a0612108 c03ae8dca1d7c451054754979e60a6de1f64c3cd Slaren 2141330+slaren@users.noreply.github.com 2023-03-29T05:38:57+02:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Fix ggml_init_params in quantize
c03ae8dca1d7c451054754979e60a6de1f64c3cd 3bcc129ba881c99795e850b0a23707a4dfdabe9d Slaren 2141330+slaren@users.noreply.github.com 2023-03-29T02:03:43+02:00 Justine Tunney jtunney@gmail.com 2023-03-30T12:28:25-07:00 Add mmap support for model files
3bcc129ba881c99795e850b0a23707a4dfdabe9d a4755cf288deb83df646f91f8fc98613271322db Stephan Walter stephan@walter.name 2023-03-30T17:56:59Z GitHub noreply@github.com 2023-03-30T20:56:59+03:00 cmake : properly invoke CTest (#629)
a4755cf288deb83df646f91f8fc98613271322db 1f0414feecc336482163af6c1e5650f9373ed8c9 Casey Primozic me@ameo.link 2023-03-30T10:53:35-07:00 GitHub noreply@github.com 2023-03-30T17:53:35Z Remove unused variable (#607)
1f0414feecc336482163af6c1e5650f9373ed8c9 77efdf5a501b1140801da5cd8751e9f9b259ec32 david raistrick keen99@users.noreply.github.com 2023-03-30T13:34:45-04:00 GitHub noreply@github.com 2023-03-30T20:34:45+03:00 make : fix darwin f16c flags check (#615)
77efdf5a501b1140801da5cd8751e9f9b259ec32 ed3c680bcd0e8ce6e574573ba95880b694449878 Georgi Gerganov ggerganov@gmail.com 2023-03-30T20:27:32+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-30T20:27:32+03:00 ggml : fix NEON signs (close #620, #622)
ed3c680bcd0e8ce6e574573ba95880b694449878 9cbc404ba6699a9ba4925ea25a60552b13491c7a slaren 2141330+slaren@users.noreply.github.com 2023-03-30T11:16:30+02:00 GitHub noreply@github.com 2023-03-30T11:16:30+02:00 Fix GGML_F32Cx8_STORE in AVX without F16C path (#619)
9cbc404ba6699a9ba4925ea25a60552b13491c7a b51c717d5cf9181c33afcb84554e47f6d539c891 anzz1 anzz1@live.com 2023-03-29T23:44:39+03:00 GitHub noreply@github.com 2023-03-29T23:44:39+03:00 ci : re-enable AVX512 testing (Windows-MSVC) (#584)
b51c717d5cf9181c33afcb84554e47f6d539c891 0ba76c1e73ae21038b80bfb5a746157376c88173 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:15:34+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:15:34+03:00 ggml : init time on first ggml_init() call
0ba76c1e73ae21038b80bfb5a746157376c88173 cea1c859483a5cfc7e2b31a06f8561d7a7604870 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:13:12+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:13:12+03:00 llama : fix compile warnings when reading the vocab
cea1c859483a5cfc7e2b31a06f8561d7a7604870 f202ada131f60059112a948f660b2e0ac93d049a Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:10:01+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:10:01+03:00 ggml : add ARM_NEON dequantize_row_q4_1()
f202ada131f60059112a948f660b2e0ac93d049a 3b44d30d9b618f0f2eb9abcfe912770a4e7d85d4 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:03:02+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:03:07+03:00 ggml : add ARM_NEON quantize_row_q4_1()
3b44d30d9b618f0f2eb9abcfe912770a4e7d85d4 61cbfff5c95e45236883b1b60e025f8f6fa8c8a3 Georgi Gerganov ggerganov@gmail.com 2023-03-29T21:47:33+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-29T22:03:07+03:00 ggml : add ARM_NEON ggml_vec_dot_q4_1()
61cbfff5c95e45236883b1b60e025f8f6fa8c8a3 d9ad104440d84a0cc0734bff47ef0ba41ba740c4 Pavol Rusnak pavol@rusnak.io 2023-03-29T20:09:25+02:00 GitHub noreply@github.com 2023-03-29T20:09:25+02:00 rename convert_ggml_to_pth.py -> convert-ggml-to-pth.py (#600)
d9ad104440d84a0cc0734bff47ef0ba41ba740c4 b467702b87461543c75013207e9adc6d20dcc01d Thérence 13496987+Royalphax@users.noreply.github.com 2023-03-29T19:21:09+02:00 GitHub noreply@github.com 2023-03-29T20:21:09+03:00 Create chat-13B.bat (#592)
b467702b87461543c75013207e9adc6d20dcc01d 516d88e75c9e768c0001a452dbad212494c586b3 Georgi Gerganov ggerganov@gmail.com 2023-03-29T19:38:31+03:00 GitHub noreply@github.com 2023-03-29T19:38:31+03:00 readme : fix typos
516d88e75c9e768c0001a452dbad212494c586b3 53635c081c49321d523567112f9fddfbba6b787b Georgi Gerganov ggerganov@gmail.com 2023-03-29T19:37:20+03:00 GitHub noreply@github.com 2023-03-29T19:37:20+03:00 readme : add GPT4All instructions (close #588)
53635c081c49321d523567112f9fddfbba6b787b 41318d708ed196ff727dce14d263a64b23c7333d Georgi Gerganov ggerganov@gmail.com 2023-03-29T19:29:26+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-29T19:29:52+03:00 py : add GPT4All conversion script
41318d708ed196ff727dce14d263a64b23c7333d a6956b25a1c783e5e96fe06c9c00438f846ef047 Maël Kerbiriou m431.kerbiriou@gmail.com 2023-03-29T18:10:07+02:00 GitHub noreply@github.com 2023-03-29T19:10:07+03:00 llama : use the same threshold for OpenBLAS and ggml thread limiting (#577)
a6956b25a1c783e5e96fe06c9c00438f846ef047 83df5639eb182ed7c122382907691d8baa3c32df Tobias Lütke tobi@shopify.com 2023-03-29T17:10:24+02:00 GitHub noreply@github.com 2023-03-29T10:10:24-05:00 add example of re-act pattern (#583)
83df5639eb182ed7c122382907691d8baa3c32df a5c42c4b13b3be9e58fe8f9adbb6ee60417674a6 anzz1 anzz1@live.com 2023-03-29T16:20:07+03:00 GitHub noreply@github.com 2023-03-29T13:20:07Z Fix GCC warning about binary literal (#595)
a5c42c4b13b3be9e58fe8f9adbb6ee60417674a6 5a5f8b1501fbb34367225544010ddfc306d6d2fe anzz1 anzz1@live.com 2023-03-29T16:19:29+03:00 GitHub noreply@github.com 2023-03-29T13:19:29Z Fix typo in llama.h (#593)
5a5f8b1501fbb34367225544010ddfc306d6d2fe f1217055eaedfc7214be93d98e529cae89830430 anzz1 anzz1@live.com 2023-03-28T22:44:29+03:00 GitHub noreply@github.com 2023-03-28T22:44:29+03:00 Enable Fused-Multiply-Add (FMA) and F16C/CVT16 vector extensions on MSVC (#375)
f1217055eaedfc7214be93d98e529cae89830430 7f4c5c66514227c3870c2bd189fb0609fdd0de10 anzz1 anzz1@live.com 2023-03-28T22:43:25+03:00 GitHub noreply@github.com 2023-03-28T22:43:25+03:00 CI: fix subdirectory path globbing (#546)
7f4c5c66514227c3870c2bd189fb0609fdd0de10 2a98bc18ea34dbf15f261a0df37080e588a189d1 anzz1 anzz1@live.com 2023-03-28T21:23:09+03:00 GitHub noreply@github.com 2023-03-28T21:23:09+03:00 llama : fix linkage with mingw (#551)
2a98bc18ea34dbf15f261a0df37080e588a189d1 d0aaff571cd5c316b68e3e11d57e274bfd2bd457 slaren 2141330+slaren@users.noreply.github.com 2023-03-28T20:06:03+02:00 GitHub noreply@github.com 2023-03-28T21:06:03+03:00 ggml : add AVX2 implementation of quantize_row_q4_1 (#515)
d0aaff571cd5c316b68e3e11d57e274bfd2bd457 d0330fd783d7c67349cdcce4a56604ef0aeccdb5 thement 40525767+thement@users.noreply.github.com 2023-03-28T19:55:42+02:00 GitHub noreply@github.com 2023-03-28T20:55:42+03:00 py : add temporary script to convert old ggml files to newer version (#539)
d0330fd783d7c67349cdcce4a56604ef0aeccdb5 99c5b2765422232ebb4414f5a63693d734406a7f Tai Duc Nguyen taiducnguyen.drexel@gmail.com 2023-03-28T13:51:29-04:00 GitHub noreply@github.com 2023-03-28T20:51:29+03:00 py : add capabiliy to convert from ggml back to torch or hf format for further consumption/training/finetuning (#403)
99c5b2765422232ebb4414f5a63693d734406a7f 692ce3164ef1201ecb9cfad315cc0a08b965adb8 Stephan Walter stephan@walter.name 2023-03-28T17:13:01Z GitHub noreply@github.com 2023-03-28T20:13:01+03:00 ggml : refactor quantized processing functions (#509)
692ce3164ef1201ecb9cfad315cc0a08b965adb8 96f9c0506fa81cada6f96f45768c34f45406c4bb DooWoong Lee (David) manics99@naver.com 2023-03-29T02:02:34+09:00 GitHub noreply@github.com 2023-03-28T20:02:34+03:00 py : removed unused `model` variable and verified that the code functions correctly with `vocab_only` setting. Also confirmed that the code works as expected after running with reduced memory usage due to deletion of no-longer-needed variable. (#547)
96f9c0506fa81cada6f96f45768c34f45406c4bb d502bc7c9d9d6dfb3a09aea404395b666d7b374d Georgi Gerganov ggerganov@gmail.com 2023-03-28T20:01:09+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-28T20:01:09+03:00 ci : make ctest verbose, hopefully we see what is wrong with the sanitizer
d502bc7c9d9d6dfb3a09aea404395b666d7b374d 436e56193199a1625f8c561069f702e8840a9e08 Georgi Gerganov ggerganov@gmail.com 2023-03-28T19:51:55+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-28T19:51:55+03:00 tests : free llama context at the end of the test
436e56193199a1625f8c561069f702e8840a9e08 20e1e84884376b3fb44ffbfd48d478b2934b0b5e Stephan Walter stephan@walter.name 2023-03-28T16:48:20Z GitHub noreply@github.com 2023-03-28T19:48:20+03:00 all : be more strict about converting float to double (#458)
20e1e84884376b3fb44ffbfd48d478b2934b0b5e c1f885067c61191a07a1aedf684168dda62f3f71 Jed Fox git@jedfox.com 2023-03-28T11:39:01-05:00 GitHub noreply@github.com 2023-03-28T19:39:01+03:00 deploy : add a Package.swift for SwiftPM support (#393)
c1f885067c61191a07a1aedf684168dda62f3f71 e0670260fb50a882b37074112b1881fb0820cf77 Stephan Walter stephan@walter.name 2023-03-28T15:56:03Z GitHub noreply@github.com 2023-03-28T18:56:03+03:00 ggml : introduce structs for the q4 data blocks (#356)
e0670260fb50a882b37074112b1881fb0820cf77 28ba975aea1dcae2f31770516f5d542ff177771e Georgi Gerganov ggerganov@gmail.com 2023-03-28T18:34:35+03:00 Georgi Gerganov ggerganov@gmail.com 2023-03-28T18:34:35+03:00 gitignore : add "embedding"
28ba975aea1dcae2f31770516f5d542ff177771e a6bdc47cba23713a22ade47dd65b6afeb8009ff4 dotpy314 33351922+dotpy314@users.noreply.github.com 2023-03-28T23:06:28+08:00 GitHub noreply@github.com 2023-03-28T18:06:28+03:00 Check the existence of f16_model_path_base in quantize.py (#574)
a6bdc47cba23713a22ade47dd65b6afeb8009ff4 7b8dbcb78b2f65c4676e41da215800d65846edd0 slaren 2141330+slaren@users.noreply.github.com 2023-03-28T16:26:55+02:00 GitHub noreply@github.com 2023-03-28T17:26:55+03:00 Fix usage of F16C intrinsics in AVX code (#563)
7b8dbcb78b2f65c4676e41da215800d65846edd0 4b8efff0e3945090379aa2f897ff125c8f9cdbae anzz1 anzz1@live.com 2023-03-28T17:09:55+03:00 GitHub noreply@github.com 2023-03-28T17:09:55+03:00 main.cpp fixes, refactoring (#571)
4b8efff0e3945090379aa2f897ff125c8f9cdbae 7e5395575a3360598f2565c73c8a2ec0c0abbdb8 RJ Adriaansen adriaansen@eshcc.eur.nl 2023-03-28T08:11:09+02:00 GitHub noreply@github.com 2023-03-28T09:11:09+03:00 Add embedding example to Makefile (#540)
7e5395575a3360598f2565c73c8a2ec0c0abbdb8 34c1072e497eb92d81ee7c0e12aa6741496a41c6 Marco Matthies 71844+marcom@users.noreply.github.com 2023-03-27T06:55:26+02:00 GitHub noreply@github.com 2023-03-27T07:55:26+03:00 Fix missing ggml link in cmake for examples/* on w64-mingw32 (#542)
34c1072e497eb92d81ee7c0e12aa6741496a41c6 939ad2d3a56815f480b6fd5ea432a7ee576a7e6b Erik Scholz Green-Sky@users.noreply.github.com 2023-03-26T17:48:40+02:00 GitHub noreply@github.com 2023-03-26T15:48:40Z ci: add debug build to sanitizer build matrix (#527)
939ad2d3a56815f480b6fd5ea432a7ee576a7e6b 8c2ec5e21d580c99e257c3cfddcf21fa53229aa4 Stephan Walter stephan@walter.name 2023-03-26T15:34:02Z GitHub noreply@github.com 2023-03-26T15:34:02Z Fix undefined variables in debug build, remove unused variables (#531)
8c2ec5e21d580c99e257c3cfddcf21fa53229aa4 b391579db92f095666be1d979899b54ae0981573 Juan Calderon-Perez 835733+gaby@users.noreply.github.com 2023-03-26T10:48:42-04:00 GitHub noreply@github.com 2023-03-26T14:48:42Z Add support for linux/arm64 platform during Docker Builds (#514)
b391579db92f095666be1d979899b54ae0981573 7a87d31f4f0c37bbb2ea695929fa4fe3ad579cda Stephan Walter stephan@walter.name 2023-03-26T13:14:01Z GitHub noreply@github.com 2023-03-26T16:14:01+03:00 Update README and comments for standalone perplexity tool (#525)
7a87d31f4f0c37bbb2ea695929fa4fe3ad579cda 348d6926ee31d4476f9b90e1a627b0925a70f847 anzz1 anzz1@live.com 2023-03-26T16:06:10+03:00 GitHub noreply@github.com 2023-03-26T16:06:10+03:00 [main] fix infinite generation (-n == -1) (#523)
348d6926ee31d4476f9b90e1a627b0925a70f847 33e35b8fe8f09adcac0632e9cece62e1dd629f7d Georgi Gerganov ggerganov@gmail.com 2023-03-26T10:20:49+03:00 GitHub noreply@github.com 2023-03-26T10:20:49+03:00 Add logo to README.md
33e35b8fe8f09adcac0632e9cece62e1dd629f7d 19726169b379bebc96189673a19b89ab1d307659 Harald Fernengel harald.fernengel@here.com 2023-03-26T07:25:46+02:00 GitHub noreply@github.com 2023-03-26T08:25:46+03:00 Exit from interactive mode if input stream is bad (#491)
19726169b379bebc96189673a19b89ab1d307659 f732695cd57fb41e3a1be625cec4edf5be45b40a anzz1 anzz1@live.com 2023-03-26T00:13:28+02:00 GitHub noreply@github.com 2023-03-26T00:13:28+02:00 CI: Run other sanitizer builds even if one fails (#511)
f732695cd57fb41e3a1be625cec4edf5be45b40a 2f7bf7dd7cd7299874d582f7f34834418abf4057 jp-x-g jpxg-dev@protonmail.com 2023-03-25T14:53:55-07:00 GitHub noreply@github.com 2023-03-25T23:53:55+02:00 Clarify console output in convert-pth-to-ggml.py (#512)
2f7bf7dd7cd7299874d582f7f34834418abf4057 34ab5268432fd287caa68d60bdd8aef411def3fa anzz1 anzz1@live.com 2023-03-25T23:38:11+02:00 GitHub noreply@github.com 2023-03-25T23:38:11+02:00 CMake / CI additions (#497)
34ab5268432fd287caa68d60bdd8aef411def3fa c2b25b6912662d2637d9c6e6df3a5de931e0d7ce anzz1 anzz1@live.com 2023-03-25T22:29:22+02:00 GitHub noreply@github.com 2023-03-25T22:29:22+02:00 (Windows) Set console to UTF-8 on init (#420)
c2b25b6912662d2637d9c6e6df3a5de931e0d7ce 79b2b266db6b198b5af450982c3cd61120fac951 Georgi Gerganov ggerganov@gmail.com 2023-03-25T21:53:39+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T21:53:39+02:00 Fix colors enabling on WIN32
79b2b266db6b198b5af450982c3cd61120fac951 e2d490dafd860eaaaf9aa8008ab790527d556daf Georgi Gerganov ggerganov@gmail.com 2023-03-25T21:51:41+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T21:51:41+02:00 If n_predict == -1, generate forever
e2d490dafd860eaaaf9aa8008ab790527d556daf 03f7e335604b3d68f74995aa2ccb4955833ee423 Georgi Gerganov ggerganov@gmail.com 2023-03-25T21:36:22+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T21:36:22+02:00 Inifinite generation via context swapping (#71)
03f7e335604b3d68f74995aa2ccb4955833ee423 55ad42af845127bd0eb0c1f36f327ecec83f4bca Georgi Gerganov ggerganov@gmail.com 2023-03-25T20:51:14+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T20:51:14+02:00 Cleanup STL headers + fix embedding examples + minor stuff
55ad42af845127bd0eb0c1f36f327ecec83f4bca 459e93cce07cab9052c06b5bf360819893442e1e Georgi Gerganov ggerganov@gmail.com 2023-03-25T20:36:52+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T20:37:09+02:00 Move chat scripts into "./examples"
459e93cce07cab9052c06b5bf360819893442e1e a316a425d04027453dc0fd45f003b647c12f66f9 slaren 2141330+slaren@users.noreply.github.com 2023-03-25T19:31:48+01:00 GitHub noreply@github.com 2023-03-25T20:31:48+02:00 Add AVX2 implementation of dequantize_row_q4_1 (#505)
a316a425d04027453dc0fd45f003b647c12f66f9 ecbe466a364876927994e2f1ec14f4d82301d201 Georgi Gerganov ggerganov@gmail.com 2023-03-25T20:26:40+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T20:26:40+02:00 Overhaul the examples structure
ecbe466a364876927994e2f1ec14f4d82301d201 502a400192013d3e95ed87b777e8fa3bec45713c Georgi Gerganov ggerganov@gmail.com 2023-03-25T19:47:21+02:00 GitHub noreply@github.com 2023-03-25T19:47:21+02:00 Retire the ggml_mul_mat() branch for transposed src0 (#500)
502a400192013d3e95ed87b777e8fa3bec45713c 09aecbf6283bbce9449e2d96000073145aaaf5fc Georgi Gerganov ggerganov@gmail.com 2023-03-25T17:16:50+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T17:17:16+02:00 Disable prompt verbosity by default and add option to enable (#480)
09aecbf6283bbce9449e2d96000073145aaaf5fc 4640eff23d341a0273587800e17ff4a378132d60 slaren 2141330+slaren@users.noreply.github.com 2023-03-25T16:06:49+01:00 GitHub noreply@github.com 2023-03-25T17:06:49+02:00 Add AVX2 implementation of dequantize_row_q4_0 (#467)
4640eff23d341a0273587800e17ff4a378132d60 ab77d7631211b299cb734bea6ad1f74324154150 Georgi Gerganov ggerganov@gmail.com 2023-03-25T17:03:10+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T17:03:10+02:00 Don't interefe with BLAS for large prompts by running only 1 thread
ab77d7631211b299cb734bea6ad1f74324154150 29b7baab670ae8b76ac0da21c2ded69ff18971ee Georgi Gerganov ggerganov@gmail.com 2023-03-25T16:47:59+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T16:49:09+02:00 Add longer DAN prompt for testing big batch numbers
29b7baab670ae8b76ac0da21c2ded69ff18971ee 4a7129acd2e939b92d70dd568c746f2fa078232c slaren 2141330+slaren@users.noreply.github.com 2023-03-25T15:34:23+01:00 GitHub noreply@github.com 2023-03-25T16:34:23+02:00 Add timings for the prompt evaluation (#478)
4a7129acd2e939b92d70dd568c746f2fa078232c 6b6dbc8910c6d53f4d96c46c8fcec70e2cd435d8 Georgi Gerganov ggerganov@gmail.com 2023-03-25T16:30:32+02:00 GitHub noreply@github.com 2023-03-25T16:30:32+02:00 Remove obsolete information from README
6b6dbc8910c6d53f4d96c46c8fcec70e2cd435d8 2a2e63ce0503d9bf3e55283e40a052c78c1cc3a8 Georgi Gerganov ggerganov@gmail.com 2023-03-25T16:22:05+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T16:22:05+02:00 Remove obsolete assert and fix compiler warning
2a2e63ce0503d9bf3e55283e40a052c78c1cc3a8 e899bf54b291e8c84173a0e534a2c262f3f63229 Georgi Gerganov ggerganov@gmail.com 2023-03-25T16:09:54+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-25T16:10:14+02:00 Fix nasty bug in ggml_compute_forward_mul_mat_f32() and reenable BLAS
e899bf54b291e8c84173a0e534a2c262f3f63229 fbd4d38c647f82b2598291ea9b8d0c09ac1ffb8c anzz1 anzz1@live.com 2023-03-25T14:42:09+02:00 GitHub noreply@github.com 2023-03-25T14:42:09+02:00 bounds checking for input prefix (#492)
fbd4d38c647f82b2598291ea9b8d0c09ac1ffb8c 58e6c9f36f97d0a3e287b97256dc5f6b0e9fb5ae anzz1 anzz1@live.com 2023-03-25T14:03:19+02:00 GitHub noreply@github.com 2023-03-25T14:03:19+02:00 feat: '--in-prefix STRING' option (#426)
58e6c9f36f97d0a3e287b97256dc5f6b0e9fb5ae 36d07532ef7ccf0bdc12e050472f359a6794957f Jed Fox git@jedfox.com 2023-03-25T01:26:28-04:00 GitHub noreply@github.com 2023-03-25T07:26:28+02:00 Add support for file load progress reporting callbacks (#434)
36d07532ef7ccf0bdc12e050472f359a6794957f 6f1ee4b640912211a4b07965c585d327e32e734d Doomsdayrs 38189170+Doomsdayrs@users.noreply.github.com 2023-03-25T01:21:24-04:00 GitHub noreply@github.com 2023-03-25T07:21:24+02:00 Add missing struct annotation (#483)
6f1ee4b640912211a4b07965c585d327e32e734d 8520fc310eab87f2c4612f2a00d4adbd44a20d0d Chris Kuehl ckuehl@ckuehl.me 2023-03-24T23:38:14-05:00 GitHub noreply@github.com 2023-03-25T06:38:14+02:00 Fix crash for 65B model with pre-allocated memory (#485)
8520fc310eab87f2c4612f2a00d4adbd44a20d0d b3f460e94139cb24b0af81cc8bc10eb86269d704 Georgi Gerganov ggerganov@gmail.com 2023-03-24T23:47:06+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-24T23:47:06+02:00 Disable BLAS altogether - the bug is not just for qunatized mat mul
b3f460e94139cb24b0af81cc8bc10eb86269d704 04c6f5ed6fafd63601fa06757877ed5ccf9d5991 Georgi Gerganov ggerganov@gmail.com 2023-03-24T23:39:17+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-24T23:39:17+02:00 Disable BLAS branch in mul_mat - seems there is a bug
04c6f5ed6fafd63601fa06757877ed5ccf9d5991 7a9b6c3a8bdc1cb75fefc826dfaa7331eb63695d Georgi Gerganov ggerganov@gmail.com 2023-03-24T23:17:58+02:00 GitHub noreply@github.com 2023-03-24T23:17:58+02:00 Immediately start processing the prompt before user input has been provided (#476)
7a9b6c3a8bdc1cb75fefc826dfaa7331eb63695d 31572d966531f7d768eb773322016ab78eb6e835 Georgi Gerganov ggerganov@gmail.com 2023-03-24T23:17:37+02:00 GitHub noreply@github.com 2023-03-24T23:17:37+02:00 Reduce memory usage and allocate enough memory for largest context (#473)
31572d966531f7d768eb773322016ab78eb6e835 f4f5362edb01b05c383b23f36d7b3489c77061b5 Georgi Gerganov ggerganov@gmail.com 2023-03-24T18:23:56+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-24T18:23:56+02:00 Temporary bump the memory buffer size - hopefully fix issues from 483bab2e
f4f5362edb01b05c383b23f36d7b3489c77061b5 863f65e2e32dc1e6d23c96a4811bf382d6b2a548 Gary Mulder gjmulder@gmail.com 2023-03-24T15:23:09Z GitHub noreply@github.com 2023-03-24T15:23:09Z Update README.md (#444)
863f65e2e32dc1e6d23c96a4811bf382d6b2a548 afd220d9c665e4c19107120ace2f0cb742e28aa1 rabidcopy rabidcopy@yahoo.com 2023-03-24T10:22:39-05:00 GitHub noreply@github.com 2023-03-24T17:22:39+02:00 fix instruct mode (#445)
afd220d9c665e4c19107120ace2f0cb742e28aa1 481044d50cfe8eaa6cd0c1a1b445680e4b0b3ebc Georgi Gerganov ggerganov@gmail.com 2023-03-24T17:21:01+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-24T17:21:01+02:00 Properly free llama_context on failure
481044d50cfe8eaa6cd0c1a1b445680e4b0b3ebc 563cdc391dde140f1084d1012234e8e6f57f881f Cameron Kaiser classilla@users.noreply.github.com 2023-03-24T08:19:26-07:00 GitHub noreply@github.com 2023-03-24T17:19:26+02:00 additional optimizations for POWER9 (#454)
563cdc391dde140f1084d1012234e8e6f57f881f 8d4a855c241ecb0f3ddc03447fe56002ebf27a37 comex comexk@gmail.com 2023-03-24T08:19:05-07:00 GitHub noreply@github.com 2023-03-24T17:19:05+02:00 Support calling mlock() on loaded model data on Linux and macOS (#453)
8d4a855c241ecb0f3ddc03447fe56002ebf27a37 b6b268d4415fd3b3e53f22b6619b724d4928f713 Luciano lucianostrika44@gmail.com 2023-03-24T08:05:13-07:00 GitHub noreply@github.com 2023-03-24T17:05:13+02:00 Add embedding mode with arg flag. Currently working (#282)
b6b268d4415fd3b3e53f22b6619b724d4928f713 3cd8dde0d1357b7f11bdd25c45d5bf5e97e284a0 Georgi Gerganov ggerganov@gmail.com 2023-03-24T09:13:35+02:00 GitHub noreply@github.com 2023-03-24T09:13:35+02:00 Add link to Roadmap discussion
3cd8dde0d1357b7f11bdd25c45d5bf5e97e284a0 4870e455b3653f7d7769fa5772b2c90ffad088df Georgi Gerganov ggerganov@gmail.com 2023-03-24T06:22:28+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-24T06:22:28+02:00 Revert "Fix memory allocation issues and seg faults"
4870e455b3653f7d7769fa5772b2c90ffad088df 483bab2e3d4a868fe679d8bb32827d2a4df214dc Georgi Gerganov ggerganov@gmail.com 2023-03-24T00:11:53+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-24T00:11:53+02:00 Fix memory allocation issues and seg faults
483bab2e3d4a868fe679d8bb32827d2a4df214dc 404e1da38ec8025707031a8027da14dc1590f952 Georgi Gerganov ggerganov@gmail.com 2023-03-23T23:22:01+02:00 GitHub noreply@github.com 2023-03-23T23:22:01+02:00 Avoid the transposed X branch in the Z = X * Y matrix multiplication (#439)
404e1da38ec8025707031a8027da14dc1590f952 4cc053b6d5e9df7ac21fa06b7208a70c156d4d7a Jed Fox git@jedfox.com 2023-03-23T16:42:52-04:00 GitHub noreply@github.com 2023-03-23T22:42:52+02:00 Fix quantize script not finding models in parent directory (#428)
4cc053b6d5e9df7ac21fa06b7208a70c156d4d7a 0ba5a3a9a5efedb1aeecbbc70a4e9825542472d5 Georgi Gerganov ggerganov@gmail.com 2023-03-23T22:39:44+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-23T22:39:44+02:00 Remove oboslete command from Docker script
0ba5a3a9a5efedb1aeecbbc70a4e9825542472d5 2e17dfd80a473099dacc0f41c9146d233c6a5972 Georgi Gerganov ggerganov@gmail.com 2023-03-23T22:32:02+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-23T22:32:21+02:00 Obsolete
2e17dfd80a473099dacc0f41c9146d233c6a5972 20a1a4e09c522a80e2a0db51643d25fa38326065 rabidcopy rabidcopy@yahoo.com 2023-03-23T15:22:47-05:00 GitHub noreply@github.com 2023-03-23T22:22:47+02:00 Replace EOS with newline to prevent context/memory being flushed by EOS in interactive mode (#333)
20a1a4e09c522a80e2a0db51643d25fa38326065 ad072fc5ad6f6905a7224ff6ea07c0644aa075b1 Timmy Knight r2d2fish@gmail.com 2023-03-23T10:18:13-10:00 GitHub noreply@github.com 2023-03-23T22:18:13+02:00 Fix GPTQ converter (#423)
ad072fc5ad6f6905a7224ff6ea07c0644aa075b1 ea10d3ded2994106596ddf8e4ed02741b3e053e6 nusu-github 29514220+nusu-github@users.noreply.github.com 2023-03-24T05:16:48+09:00 GitHub noreply@github.com 2023-03-23T21:16:48+01:00 Generate library with CMake (#430)
ea10d3ded2994106596ddf8e4ed02741b3e053e6 a18c19259a3cb9dec332d613e8f15704f678a468 anzz1 anzz1@live.com 2023-03-23T19:54:28+02:00 GitHub noreply@github.com 2023-03-23T19:54:28+02:00 Command line args bounds checking (#424)
a18c19259a3cb9dec332d613e8f15704f678a468 a50e39c6fe36be3de0941b3c05aaf9c37912fd47 Ben Siraphob bensiraphob@gmail.com 2023-03-22T00:37:02-05:00 Pavol Rusnak pavol@rusnak.io 2023-03-23T17:51:26+01:00 Fix Nix build
a50e39c6fe36be3de0941b3c05aaf9c37912fd47 a140219e81cfb80356438112cd2290d701b282bb Stephan Walter stephan@walter.name 2023-03-23T14:15:48Z GitHub noreply@github.com 2023-03-23T15:15:48+01:00 Revert "Delete SHA256SUMS for now" (#429)
a140219e81cfb80356438112cd2290d701b282bb 8a3e5ef801339e57b9b0449220e9ffb11a6648e2 Kerfuffle 44031344+KerfuffleV2@users.noreply.github.com 2023-03-23T05:41:32-06:00 GitHub noreply@github.com 2023-03-23T12:41:32+01:00 Fix Makefile echo escape codes (by removing them). (#418)
8a3e5ef801339e57b9b0449220e9ffb11a6648e2 8eea5ae0e5f31238a97c79ea9103c27647380e37 Gary Mulder gjmulder@gmail.com 2023-03-23T11:30:40Z GitHub noreply@github.com 2023-03-23T11:30:40Z Move model section from issue template to README.md (#421)
8eea5ae0e5f31238a97c79ea9103c27647380e37 93208cfb929c2323e5d2ac6bf354e278040e70ed anzz1 anzz1@live.com 2023-03-23T12:26:19+02:00 GitHub noreply@github.com 2023-03-23T11:26:19+01:00 Delete SHA256SUMS for now (#416)
93208cfb929c2323e5d2ac6bf354e278040e70ed 03ace14cfd68a1289ac3b76563534c8ee72a2e53 Georgi Gerganov ggerganov@gmail.com 2023-03-23T10:46:58+02:00 GitHub noreply@github.com 2023-03-23T10:46:58+02:00 Adjust repetition penalty ..
03ace14cfd68a1289ac3b76563534c8ee72a2e53 e4412b45e395981068d2850d3fa04cc16c77d70d Georgi Gerganov ggerganov@gmail.com 2023-03-23T09:48:51+02:00 GitHub noreply@github.com 2023-03-23T09:48:51+02:00 Add link to recent podcast about whisper.cpp and llama.cpp
e4412b45e395981068d2850d3fa04cc16c77d70d f7dc43bc0d759732815856183246f167111587ad anzz1 anzz1@live.com 2023-03-23T04:20:34+02:00 GitHub noreply@github.com 2023-03-23T04:20:34+02:00 CI: CMake: Separate build and test steps (#376)
f7dc43bc0d759732815856183246f167111587ad ee8a7887865a893be208e0a92d6d94d2cb66a789 tjohnman tjohnman@users.noreply.github.com 2023-03-23T01:30:23+01:00 GitHub noreply@github.com 2023-03-23T01:30:23+01:00 Fix instruct mode broken by PR #354 (#409)
ee8a7887865a893be208e0a92d6d94d2cb66a789 69c92298a9e36dc2363b3bf50452976ce49487b3 Gary Mulder gjmulder@gmail.com 2023-03-22T19:06:18Z GitHub noreply@github.com 2023-03-22T19:06:18Z Update issue template so people will use it (#404)
69c92298a9e36dc2363b3bf50452976ce49487b3 97940520e8fd49c56bb29b71cc350190b723513f Stephan Walter stephan@walter.name 2023-03-22T17:29:06Z GitHub noreply@github.com 2023-03-22T19:29:06+02:00 Deduplicate q4 quantization functions (#383)
97940520e8fd49c56bb29b71cc350190b723513f 305ba6f0e6daa3796aad9dd18053a1945dd4cc58 Valentyn Bezshapkin 61702053+valentynbez@users.noreply.github.com 2023-03-22T18:20:25+01:00 GitHub noreply@github.com 2023-03-22T19:20:25+02:00 fix: add POSIX functionality for Linux compilation (#51)
305ba6f0e6daa3796aad9dd18053a1945dd4cc58 4122dffff958cd137175b58f1f27c0913528d7ba tjohnman tjohnman@users.noreply.github.com 2023-03-22T18:16:35+01:00 GitHub noreply@github.com 2023-03-22T19:16:35+02:00 Don't force immediate interactive without `-i` (#354)
4122dffff958cd137175b58f1f27c0913528d7ba 56e659a0b271436e24813a801640d015e7b05328 Erik Scholz Green-Sky@users.noreply.github.com 2023-03-22T17:37:10+01:00 GitHub noreply@github.com 2023-03-22T18:37:10+02:00 cmake: make llama an actual library (#392)
56e659a0b271436e24813a801640d015e7b05328 40ea807a972ec7b5a426f034ebfa593b5e7a06ed Erik Scholz Green-Sky@users.noreply.github.com 2023-03-22T17:09:38+01:00 GitHub noreply@github.com 2023-03-22T18:09:38+02:00 fix perplexity after c-api refactor (#390)
40ea807a972ec7b5a426f034ebfa593b5e7a06ed d5850c53ca179b9674b98f35d359763416a3cc11 Gary Linscott glinscott@gmail.com 2023-03-22T08:53:54-07:00 GitHub noreply@github.com 2023-03-22T08:53:54-07:00 Add details on perplexity to README.md (#395)
d5850c53ca179b9674b98f35d359763416a3cc11 ae44e23ee36c02da0e37ab508a4b473ace724f8e Yusuf Kağan Hanoğlu hanoglu@yahoo.com 2023-03-22T11:55:45+03:00 GitHub noreply@github.com 2023-03-22T10:55:45+02:00 Add missing header for memcpy (#386)
ae44e23ee36c02da0e37ab508a4b473ace724f8e 928480ef5b7b03d7a07e98286aebe3d8b24457d9 Georgi Gerganov ggerganov@gmail.com 2023-03-22T07:47:15+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-22T07:47:15+02:00 When seed <= 0 - use the clock to generate one
928480ef5b7b03d7a07e98286aebe3d8b24457d9 56817b1f882b1894daa4051d0de0bf9a0926d315 Georgi Gerganov ggerganov@gmail.com 2023-03-22T07:45:00+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-22T07:45:14+02:00 Init llama_context_params properly from CLI (#370)
56817b1f882b1894daa4051d0de0bf9a0926d315 f5a77a629bd0f37ae1696747633ab42a5530ec15 Georgi Gerganov ggerganov@gmail.com 2023-03-22T07:34:02+02:00 GitHub noreply@github.com 2023-03-22T07:34:02+02:00 Remove temporary notice and update hot topics
f5a77a629bd0f37ae1696747633ab42a5530ec15 da0e9fe90ccf6e73597eb19dd0cfc0a28363fb3b Georgi Gerganov ggerganov@gmail.com 2023-03-22T07:32:36+02:00 GitHub noreply@github.com 2023-03-22T07:32:36+02:00 Introduce C-style API (#370)
da0e9fe90ccf6e73597eb19dd0cfc0a28363fb3b e6c9e0986c79ba1cc8848879b2fcce979f9b4672 Gary Mulder gjmulder@gmail.com 2023-03-20T20:14:06Z Pavol Rusnak pavol@rusnak.io 2023-03-21T23:19:11+01:00 Add SHA256SUMS file and instructions to README how to obtain and verify the downloads
e6c9e0986c79ba1cc8848879b2fcce979f9b4672 01a297b09932e29f3319d6588977c32a926c7907 anzz1 anzz1@live.com 2023-03-21T23:49:24+02:00 anzz1 anzz1@live.com 2023-03-22T00:01:08+02:00 Fix bin dir for win ci
01a297b09932e29f3319d6588977c32a926c7907 3366853e41fcc818222a0271c76b6106179106fb Erik Scholz Green-Sky@users.noreply.github.com 2023-03-21T22:34:25+01:00 GitHub noreply@github.com 2023-03-21T23:34:25+02:00 specify build type for ctest on windows (#371)
3366853e41fcc818222a0271c76b6106179106fb 3f9c6135e45ae3f520b1e17197004cc60c9ca45b Georgi Gerganov ggerganov@gmail.com 2023-03-21T22:57:35+02:00 GitHub noreply@github.com 2023-03-21T22:57:35+02:00 Add notice about pending change
3f9c6135e45ae3f520b1e17197004cc60c9ca45b 0f6135270839f0715843c4d480c63ae150def419 Mathieu Nayrolles MathieuNls@users.noreply.github.com 2023-03-21T16:52:27-04:00 GitHub noreply@github.com 2023-03-21T22:52:27+02:00 fix typo in chatLLaMa (#368)
0f6135270839f0715843c4d480c63ae150def419 353ec251a42491f5192c48561da4b444ef67f23c Georgi Gerganov ggerganov@gmail.com 2023-03-21T19:47:27+02:00 GitHub noreply@github.com 2023-03-21T19:47:27+02:00 Update issue templates
353ec251a42491f5192c48561da4b444ef67f23c 89d5d90f3b6d25f134da7a8e252c3432bffcf674 Fabio R. Sluzala Fabio3rs@users.noreply.github.com 2023-03-21T14:21:50-03:00 GitHub noreply@github.com 2023-03-21T19:21:50+02:00 We could use std::unordered_map over std::map (#305)
89d5d90f3b6d25f134da7a8e252c3432bffcf674 16ffc013c62f22bdaa3cdc022d7a13fd952d73fc Matvey Soloviev blackhole89@gmail.com 2023-03-21T18:11:01+01:00 GitHub noreply@github.com 2023-03-21T19:11:01+02:00 Fix color codes emitting mid-UTF8 code. (#312)
16ffc013c62f22bdaa3cdc022d7a13fd952d73fc 486ae645fd3eda8b9d7413d5ff34fb65a3e337fb comex comexk@gmail.com 2023-03-21T09:42:25-07:00 GitHub noreply@github.com 2023-03-21T18:42:25+02:00 Importer for GPTQ quantized LLaMA models (#301)
486ae645fd3eda8b9d7413d5ff34fb65a3e337fb 3ab3e6582f7320c2b6568c892fdfc8215caf7e6c Gary Linscott glinscott@gmail.com 2023-03-21T09:27:42-07:00 GitHub noreply@github.com 2023-03-21T18:27:42+02:00 Compute perplexity over prompt (#270)
3ab3e6582f7320c2b6568c892fdfc8215caf7e6c f157088cb75f23208abc92b473a132ef3f7a7f15 Jean-Christophe Hoelt hoelt@fovea.cc 2023-03-21T18:23:15+02:00 GitHub noreply@github.com 2023-03-21T18:23:15+02:00 Add chatLLaMa script (#198)
f157088cb75f23208abc92b473a132ef3f7a7f15 c86ba036e613d46815501a4c6775117c9fc7afce Alex von Gluck IV kallisti5@unixzen.com 2023-03-21T11:21:06-05:00 GitHub noreply@github.com 2023-03-21T18:21:06+02:00 makefile: Fix CPU feature detection on Haiku (#218)
c86ba036e613d46815501a4c6775117c9fc7afce 1daf4dd71235dbbf537738e7ad53daad8d97586f anzz1 anzz1@live.com 2023-03-21T18:14:46+02:00 GitHub noreply@github.com 2023-03-21T18:14:46+02:00 Enable ANSI colors on Windows 10+ (#311)
1daf4dd71235dbbf537738e7ad53daad8d97586f dc6a845b8573cd7d06c6b295241d26f311602a1f Georgi Gerganov ggerganov@gmail.com 2023-03-21T18:10:32+02:00 GitHub noreply@github.com 2023-03-21T18:10:32+02:00 Minor style changes
dc6a845b8573cd7d06c6b295241d26f311602a1f 6a612959e1b6c37b68b6b141329751a2902b1030 Georgi Gerganov ggerganov@gmail.com 2023-03-21T18:09:37+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-21T18:09:46+02:00 Add chat.sh script
6a612959e1b6c37b68b6b141329751a2902b1030 d5f56a5e5a0069329a81f96460221e7afb1daddc tjohnman tjohnman@users.noreply.github.com 2023-03-21T17:05:06+01:00 GitHub noreply@github.com 2023-03-21T18:05:06+02:00 Check for reverse prompt by characters instead of tokens (#292) (#330)
d5f56a5e5a0069329a81f96460221e7afb1daddc 3bfa3b43b7319b71853bfc7d3cf4e9767c24bbc8 tjohnman tjohnman@users.noreply.github.com 2023-03-21T17:04:43+01:00 GitHub noreply@github.com 2023-03-21T18:04:43+02:00 Check for reverse prompt by characters instead of tokens (#292) (#330)
3bfa3b43b7319b71853bfc7d3cf4e9767c24bbc8 715d292ee0e34d27f27af43d7feaad1f1344981d Georgi Gerganov ggerganov@gmail.com 2023-03-21T17:59:16+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-21T17:59:16+02:00 Fix convert script, warnings alpaca instructions, default params
715d292ee0e34d27f27af43d7feaad1f1344981d c98ae02668a25916954b1653e25a5a35ca048d63 Kevin Lo kevlo@kevlo.org 2023-03-21T09:50:09-06:00 GitHub noreply@github.com 2023-03-21T17:50:09+02:00 Add OpenBSD support (#314)
c98ae02668a25916954b1653e25a5a35ca048d63 c3b2306b18a087799acc431e485b8a2e3162cd52 Mack Straight eiz@users.noreply.github.com 2023-03-21T08:49:43-07:00 GitHub noreply@github.com 2023-03-21T17:49:43+02:00 fix typo in comment (#318)
c3b2306b18a087799acc431e485b8a2e3162cd52 975d2cebf97ce888fa0aeee6f5ac774d7135891f Qingyou Meng meng.qingyou@gmail.com 2023-03-21T23:44:11+08:00 GitHub noreply@github.com 2023-03-21T17:44:11+02:00 Makefile: slightly cleanup for Mac Intel; echo instead of run ./main -h (#335)
975d2cebf97ce888fa0aeee6f5ac774d7135891f e0ffc861fae5ac8b40ce973f822d03db02929d36 anzz1 anzz1@live.com 2023-03-21T17:42:43+02:00 GitHub noreply@github.com 2023-03-21T17:42:43+02:00 cmdline option for custom amount of model parts (--n_parts N) (#348)
e0ffc861fae5ac8b40ce973f822d03db02929d36 8f644a0a859938c787d329d27f98e03c58d7df27 Kevin Kwok antimatter15@gmail.com 2023-03-21T08:34:49-07:00 GitHub noreply@github.com 2023-03-21T17:34:49+02:00 Update IPFS links to quantized alpaca with new tokenizer format (#352)
8f644a0a859938c787d329d27f98e03c58d7df27 eb34620aeceaf9d9df7fcb19acc17ad41b9f60f8 Georgi Gerganov ggerganov@gmail.com 2023-03-21T17:32:14+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-21T17:32:14+02:00 Change default repeat_penalty to 1.0
eb34620aeceaf9d9df7fcb19acc17ad41b9f60f8 2e664f1ff413995506c9a54f3a8d5b8c64e37a91 Georgi Gerganov ggerganov@gmail.com 2023-03-21T17:29:41+02:00 GitHub noreply@github.com 2023-03-21T17:29:41+02:00 Add tokenizer test + revert to C++11 (#355)
2e664f1ff413995506c9a54f3a8d5b8c64e37a91 8cf9f34eddc124d4ab28f4d2fe8e99d574510bde Casey Primozic casey@cprimozic.net 2023-03-21T07:35:42-07:00 GitHub noreply@github.com 2023-03-21T15:35:42+01:00 Add initial AVX512 support for dot product on Linux (#320)
8cf9f34eddc124d4ab28f4d2fe8e99d574510bde bd4b46d6ba504b99c936f43fc014529adffb6048 nusu-github 29514220+nusu-github@users.noreply.github.com 2023-03-21T09:37:16+09:00 GitHub noreply@github.com 2023-03-21T01:37:16+01:00 Adding missing features of CMakeLists.txt & Refactoring (#131)
bd4b46d6ba504b99c936f43fc014529adffb6048 6b6d5b5024faaf82019d08cde5e8a9d69c6ca316 Ben Siraphob bensiraphob@gmail.com 2023-03-20T16:44:30-05:00 Pavol Rusnak pavol@rusnak.io 2023-03-20T22:50:22+01:00 Nix flake: set meta.mainProgram to llama
6b6d5b5024faaf82019d08cde5e8a9d69c6ca316 a791a68b613b162c88a83f5f0225223bc167c762 Qingyou Meng meng.qingyou@gmail.com 2023-03-21T03:33:10+08:00 GitHub noreply@github.com 2023-03-20T19:33:10Z Fixed tokenizer.model not found error when model dir is symlink (#325)
a791a68b613b162c88a83f5f0225223bc167c762 0f1b21cb90ac6b84a9af70cafb8e13b5389e3b32 Mack Straight eiz@users.noreply.github.com 2023-03-20T12:26:01-07:00 GitHub noreply@github.com 2023-03-20T19:26:01Z move file magic/version to header, print expected version (#319)
0f1b21cb90ac6b84a9af70cafb8e13b5389e3b32 074bea2eb1f1349a0118239c4152914aecaa1be4 Bernat Vadell hounter.caza@gmail.com 2023-03-20T18:05:20+01:00 GitHub noreply@github.com 2023-03-20T18:05:20+01:00 Docker - Fix publish docker image in GitHub Registry (#235)
074bea2eb1f1349a0118239c4152914aecaa1be4 5cb63e2493c49bc2c3b9b355696e8dc26cdd0380 Mack Straight eiz@users.noreply.github.com 2023-03-20T03:17:23-07:00 GitHub noreply@github.com 2023-03-20T03:17:23-07:00 sentencepiece bpe compatible tokenizer (#252)
5cb63e2493c49bc2c3b9b355696e8dc26cdd0380 da5303c1ea68aa19db829c634f1e10d08d409680 Stephan Walter stephan@walter.name 2023-03-20T08:24:11Z GitHub noreply@github.com 2023-03-20T09:24:11+01:00 Add tqdm to Python requirements (#293)
da5303c1ea68aa19db829c634f1e10d08d409680 4545539d718cf88f4c3a76669b8ac2e26cd8a1e5 cocktailpeanut 121128867+cocktailpeanut@users.noreply.github.com 2023-03-19T17:44:20-04:00 GitHub noreply@github.com 2023-03-19T23:44:20+02:00 bugfix: default should not be interactive (#304)
4545539d718cf88f4c3a76669b8ac2e26cd8a1e5 edeba283665591f2f726024a92efe4b0b40434b3 Georgi Gerganov ggerganov@gmail.com 2023-03-19T21:58:51+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-19T21:58:51+02:00 Rename script
edeba283665591f2f726024a92efe4b0b40434b3 5c19c70ba631a8f5d54feb6634e0eea178911a84 Georgi Gerganov ggerganov@gmail.com 2023-03-19T21:57:28+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-19T21:57:48+02:00 Add temporary helper script for Alpaca chat
5c19c70ba631a8f5d54feb6634e0eea178911a84 24568371ae0d7caf85164abe4753f36a7dba0288 Rickey Bowers Jr bitRAKE@gmail.com 2023-03-19T13:44:30-06:00 GitHub noreply@github.com 2023-03-19T19:44:30Z fix coloring of last `n_batch` of prompt, and refactor line input (#221)
24568371ae0d7caf85164abe4753f36a7dba0288 7392f1cd2cef4dfed41f4db7c4160ab86c0dfcd9 tjohnman tjohnman@users.noreply.github.com 2023-03-19T20:33:06+01:00 GitHub noreply@github.com 2023-03-19T21:33:06+02:00 Support for multiple reverse prompts. (#299)
7392f1cd2cef4dfed41f4db7c4160ab86c0dfcd9 ad5fd5b60cfdfbfb22b0f2bc9e9f6c9692768f8d Suaj Carrot 72162667+SuajCarrot@users.noreply.github.com 2023-03-19T12:38:44-06:00 GitHub noreply@github.com 2023-03-19T20:38:44+02:00 Improved quantize script (#222)
ad5fd5b60cfdfbfb22b0f2bc9e9f6c9692768f8d 368d0c8a9ebae16a20e1c8971b21ee888bdefad5 tjohnman tjohnman@users.noreply.github.com 2023-03-19T19:36:19+01:00 GitHub noreply@github.com 2023-03-19T20:36:19+02:00 Make prompt randomization optional. (#300)
368d0c8a9ebae16a20e1c8971b21ee888bdefad5 50fae10d0339f2bd639f69dd679c0201d939a265 tjohnman tjohnman@users.noreply.github.com 2023-03-19T19:31:17+01:00 GitHub noreply@github.com 2023-03-19T20:31:17+02:00 Respect the maximum number of tokens in interactive. (#298)
50fae10d0339f2bd639f69dd679c0201d939a265 084e2f0ec081c929343d44b09df07ae87cd1ed32 slaren 2141330+slaren@users.noreply.github.com 2023-03-19T19:22:48+01:00 GitHub noreply@github.com 2023-03-19T20:22:48+02:00 Add --ignore-eos parameter (#181)
084e2f0ec081c929343d44b09df07ae87cd1ed32 0b366e735729327476ec31da02de3c9c9771ddfb Qingyou Meng meng.qingyou@gmail.com 2023-03-20T02:10:00+08:00 GitHub noreply@github.com 2023-03-19T20:10:00+02:00 interactive mode: print '\n' in sigint_handler, this flush stdout thus ensure color reset. (#283)
0b366e735729327476ec31da02de3c9c9771ddfb 160bfb217da5038ccbd74438f9f16a16012d7866 Erik Scholz Green-Sky@users.noreply.github.com 2023-03-19T18:57:00+01:00 GitHub noreply@github.com 2023-03-19T19:57:00+02:00 Command line switch to use F16 for memory_k and memory_v (refactor of #154) (#294)
160bfb217da5038ccbd74438f9f16a16012d7866 c494ed5b94b429d3d73721235e78c9f5fa6e5652 Georgi Gerganov ggerganov@gmail.com 2023-03-19T19:51:55+02:00 GitHub noreply@github.com 2023-03-19T19:51:55+02:00 Update hot topics to mention Alpaca support
c494ed5b94b429d3d73721235e78c9f5fa6e5652 c1c7026b470ced0b8a6c67e968c04bb47864def1 Georgi Gerganov ggerganov@gmail.com 2023-03-19T19:46:32+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-19T19:46:32+02:00 Fix off-by-one bug (#115)
c1c7026b470ced0b8a6c67e968c04bb47864def1 467b149761cd63248b00d6ffb204d50a4cbb451a Georgi Gerganov ggerganov@gmail.com 2023-03-19T19:33:18+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-19T19:33:18+02:00 Fix python stuff (#109)
467b149761cd63248b00d6ffb204d50a4cbb451a 70f01cb8632f73b5cf70428608b89cd3c0775d23 qunash anzoria@gmail.com 2023-03-19T20:17:39+03:00 GitHub noreply@github.com 2023-03-19T19:17:39+02:00 Refactoring `convert-pth-to-ggml.py`: more concise and readable (#109)
70f01cb8632f73b5cf70428608b89cd3c0775d23 a4e63b73dfa1894387926cc8072b5f36deebf0a5 Georgi Gerganov ggerganov@gmail.com 2023-03-19T19:04:44+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-19T19:05:04+02:00 Drop trailing new line from file prompts (#80)
a4e63b73dfa1894387926cc8072b5f36deebf0a5 9e1707218a24ff758c7b623594f8c0ce5e12eb6c Georgi Gerganov ggerganov@gmail.com 2023-03-19T18:49:50+02:00 GitHub noreply@github.com 2023-03-19T18:49:50+02:00 Add instruction for using Alpaca (#240)
9e1707218a24ff758c7b623594f8c0ce5e12eb6c 22213a17b56336bbea384a572a9484ce208c0333 Georgi Gerganov ggerganov@gmail.com 2023-03-19T18:37:02+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-19T18:37:02+02:00 Add "--instruct" argument for usage with Alpaca (#240)
22213a17b56336bbea384a572a9484ce208c0333 d7def1a7524f712e5ebb7cd02bab0f13aa56a7f9 Georgi Gerganov ggerganov@gmail.com 2023-03-19T17:30:00+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-19T17:30:00+02:00 Change RMSNorm eps to 1e-6 (#173)
d7def1a7524f712e5ebb7cd02bab0f13aa56a7f9 6f61c18ec9a30416e21ed5abfb1321bdb14979be Ronsor ronsor@ronsor.pw 2023-03-18T17:10:47-07:00 GitHub noreply@github.com 2023-03-18T20:10:47-04:00 Warn user if a context size greater than 2048 tokens is specified (#274)
6f61c18ec9a30416e21ed5abfb1321bdb14979be 1e5a6d088d0f3a967c6e86298a756daec9e8df12 Pavol Rusnak pavol@rusnak.io 2023-03-18T22:39:46+01:00 Pavol Rusnak pavol@rusnak.io 2023-03-18T23:18:04+01:00 Fix typo in readme
1e5a6d088d0f3a967c6e86298a756daec9e8df12 554b54152145c30618bac171efb712cf4a7d1e96 Pavol Rusnak pavol@rusnak.io 2023-03-18T22:20:04+01:00 Pavol Rusnak pavol@rusnak.io 2023-03-18T22:25:35+01:00 Add note about Python 3.11 to readme
554b54152145c30618bac171efb712cf4a7d1e96 d3f202d57b694376cef6f381a6b6901825c3f6d9 Pavol Rusnak pavol@rusnak.io 2023-03-18T21:58:46+01:00 Pavol Rusnak pavol@rusnak.io 2023-03-18T22:25:35+01:00 Add memory/disk requirements to readme
d3f202d57b694376cef6f381a6b6901825c3f6d9 e03e359730c127f888fcf00e93375771bc0a3500 Alex Nguyen tiendung@users.noreply.github.com 2023-03-18T20:51:49+07:00 GitHub noreply@github.com 2023-03-18T13:51:49Z Remove unused code since n_vocab is model.hparams.n_vocab (#262)
e03e359730c127f888fcf00e93375771bc0a3500 a81d0c2a171a4446e6a21a3ec74a0c0768d71184 Justin Suess justin.suess@westpoint.edu 2023-03-18T07:44:09-04:00 GitHub noreply@github.com 2023-03-18T11:44:09Z fixed warning with std::ignore about unused function result (#151)
a81d0c2a171a4446e6a21a3ec74a0c0768d71184 b2de7f18dfbb93463eeb5b4392117bbe82d5bd1b Gary Linscott glinscott@gmail.com 2023-03-18T04:17:19-07:00 GitHub noreply@github.com 2023-03-18T11:17:19Z Fix n^2 loop in tokenization (#254)
b2de7f18dfbb93463eeb5b4392117bbe82d5bd1b a29274789309029fd88a9465e6d0832d4632272b anzz1 anzz1@live.com 2023-03-18T09:27:12+02:00 GitHub noreply@github.com 2023-03-18T09:27:12+02:00 CI Improvements (#230)
a29274789309029fd88a9465e6d0832d4632272b c9f670a17755311aa28c411f5c7f3c8c05434770 Niklas Korz niklas@niklaskorz.de 2023-03-17T23:03:48+01:00 GitHub noreply@github.com 2023-03-17T23:03:48+01:00 Nix flake (#40)
c9f670a17755311aa28c411f5c7f3c8c05434770 4f546091102a418ffdc6230f872ac56e5cedb835 thement 40525767+thement@users.noreply.github.com 2023-03-17T21:05:58+01:00 GitHub noreply@github.com 2023-03-17T21:05:58+01:00 Implement non-greedy tokenizer that tries to maximize token lengths (#242)
4f546091102a418ffdc6230f872ac56e5cedb835 e81b9c81c101f64531ef0fa1ee6b77d562635652 Georgi Gerganov ggerganov@gmail.com 2023-03-17T21:46:46+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-17T21:46:46+02:00 Default to 4 threads (#243)
e81b9c81c101f64531ef0fa1ee6b77d562635652 367946c668757532deed929e1d78673c6ac6bcb8 Georgi Gerganov ggerganov@gmail.com 2023-03-17T20:30:04+02:00 GitHub noreply@github.com 2023-03-17T20:30:04+02:00 Update Contributing section
367946c668757532deed929e1d78673c6ac6bcb8 6b0df5ccf360fe5c015f6607f0375bfc6849005e Stephan Walter stephan@walter.name 2023-03-17T17:47:35Z GitHub noreply@github.com 2023-03-17T19:47:35+02:00 Don't tell users to use a bad number of threads (#243)
6b0df5ccf360fe5c015f6607f0375bfc6849005e 2af23d30434a677c6416812eea52ccc0af65119c mmyjona jonathan.gonse@gmail.com 2023-03-18T00:38:24+08:00 GitHub noreply@github.com 2023-03-17T13:38:24-03:00 add ptread link to fix cmake build under linux (#114)
2af23d30434a677c6416812eea52ccc0af65119c 904d2a8d6acd667c9633138d45a361d40fbf76d0 Bernat Vadell hounter.caza@gmail.com 2023-03-17T10:47:06+01:00 GitHub noreply@github.com 2023-03-17T10:47:06+01:00 🚀 Dockerize llamacpp (#132)
904d2a8d6acd667c9633138d45a361d40fbf76d0 721311070e31464ac12bef9a4444093eb3eaebf7 Matvey Soloviev blackhole89@gmail.com 2023-03-17T05:48:39+01:00 GitHub noreply@github.com 2023-03-17T06:48:39+02:00 Q4_1 quantization (#193)
721311070e31464ac12bef9a4444093eb3eaebf7 ac15de789547e5a6e93df552e787379b3a23ef26 Georgi Gerganov ggerganov@gmail.com 2023-03-16T15:00:09+02:00 GitHub noreply@github.com 2023-03-16T15:00:09+02:00 Update README.md
ac15de789547e5a6e93df552e787379b3a23ef26 273abc47ff9dd899b3c4f58acd19d4649e90d6b4 Georgi Gerganov ggerganov@gmail.com 2023-03-16T08:55:13+02:00 GitHub noreply@github.com 2023-03-16T08:55:13+02:00 Expand "Contributing" section
273abc47ff9dd899b3c4f58acd19d4649e90d6b4 9b4a15b17d8395eb075379b140fcd0b0283f4ef6 Georgi Gerganov ggerganov@gmail.com 2023-03-16T07:12:12+02:00 GitHub noreply@github.com 2023-03-16T07:12:12+02:00 Update hot topics - RMSnorm
9b4a15b17d8395eb075379b140fcd0b0283f4ef6 6eac39ba953acaeec396cea2969dbf413907e2ec Nebula infinitewormhole@gmail.com 2023-03-15T19:29:25-04:00 GitHub noreply@github.com 2023-03-15T19:29:25-04:00 Fix RMS norm in GGML (#191)
6eac39ba953acaeec396cea2969dbf413907e2ec 27944c4206a49bbe003021a2610bacaa3044e619 hoangmit hoangmit@users.noreply.github.com 2023-03-15T18:41:38-04:00 GitHub noreply@github.com 2023-03-16T00:41:38+02:00 Add RMS norm and use it (#187)
27944c4206a49bbe003021a2610bacaa3044e619 2d15d6c9a959749f954d4fbbf44d711e19c5bdff moritzbrantner 31051084+moritzbrantner@users.noreply.github.com 2023-03-15T21:35:25+01:00 GitHub noreply@github.com 2023-03-15T22:35:25+02:00 fixed typo (#178)
2d15d6c9a959749f954d4fbbf44d711e19c5bdff 2d64715ad475f192a4004a52d134c67ccb6f44ad Rickey Bowers Jr bitRAKE@gmail.com 2023-03-15T13:56:24-06:00 GitHub noreply@github.com 2023-03-15T21:56:24+02:00 add SIGINT support for _WIN32 environments (#120)
2d64715ad475f192a4004a52d134c67ccb6f44ad 16b2c61a22f828ea77d9f084ca871c63bc5cc283 Justin Suess justin.suess@westpoint.edu 2023-03-15T15:42:40-04:00 GitHub noreply@github.com 2023-03-15T21:42:40+02:00 added ctx_size parameter (#148)
16b2c61a22f828ea77d9f084ca871c63bc5cc283 977295c700a2952c18400026d57467077dcd1a20 Justin Suess justin.suess@westpoint.edu 2023-03-15T15:39:38-04:00 GitHub noreply@github.com 2023-03-15T21:39:38+02:00 fixed color reset on exit (#149)
977295c700a2952c18400026d57467077dcd1a20 956dfda8ad8cea7961e22e0384bbc315bf79aed2 Musab Gultekin musabgultekin@users.noreply.github.com 2023-03-15T22:39:06+03:00 GitHub noreply@github.com 2023-03-15T21:39:06+02:00 Fix potential licensing issue (#126)
956dfda8ad8cea7961e22e0384bbc315bf79aed2 113e685d18ac4edb20f647fd34b000941556f6a6 Ronsor ronsor@ronsor.pw 2023-03-15T12:37:50-07:00 GitHub noreply@github.com 2023-03-15T21:37:50+02:00 Use `tokenizer.vocab_size()` instead of hardcoding 32000 in convert-pth-to-ggml.py (#142)
113e685d18ac4edb20f647fd34b000941556f6a6 47857e564c218a2c38346d0cdd94314632878fcb hoangmit hoangmit@users.noreply.github.com 2023-03-15T15:05:14-04:00 GitHub noreply@github.com 2023-03-15T21:05:14+02:00 inline -> static inline for "bytesFromNibbles" (#161)
47857e564c218a2c38346d0cdd94314632878fcb 60f819a2b10475055a36415bc489e5b55df2d052 Ronsor ronsor@ronsor.pw 2023-03-14T12:34:37-07:00 GitHub noreply@github.com 2023-03-14T21:34:37+02:00 Don't use vdotq_s32 if it's not available (#139)
60f819a2b10475055a36415bc489e5b55df2d052 97ab2b257897bfe7e2ae72876a3e50ed41b8c7ce Radoslav Gerganov rgerganov@gmail.com 2023-03-14T15:30:08+02:00 GitHub noreply@github.com 2023-03-14T15:30:08+02:00 Add section to README on how to run the project on Android (#130)
97ab2b257897bfe7e2ae72876a3e50ed41b8c7ce 2f700a27381e558a4eb5a3f8fd56757f4c7a417c Georgi Gerganov ggerganov@gmail.com 2023-03-14T09:43:52+02:00 GitHub noreply@github.com 2023-03-14T09:43:52+02:00 Add Misc section + update hot topics + minor fixes
2f700a27381e558a4eb5a3f8fd56757f4c7a417c c09a9cfb06c87d114615c105adda91b0e6273b69 Sebastián A sebastian.aedo29@gmail.com 2023-03-13T17:29:10-03:00 GitHub noreply@github.com 2023-03-13T22:29:10+02:00 Add windows to the CI (#98)
c09a9cfb06c87d114615c105adda91b0e6273b69 7ec903d3c162417c11463f14ad5b773a918fb7f1 Georgi Gerganov ggerganov@gmail.com 2023-03-13T21:22:15+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-13T21:22:15+02:00 CMake build in Release by default (#75)
7ec903d3c162417c11463f14ad5b773a918fb7f1 4497ad819c0010a8b19ffeaf8c0428eb7558d3e0 Georgi Gerganov ggerganov@gmail.com 2023-03-13T19:21:51+02:00 GitHub noreply@github.com 2023-03-13T19:21:51+02:00 Update contribution section, hot topics, limitations, etc.
4497ad819c0010a8b19ffeaf8c0428eb7558d3e0 ed6849cc07a8973e5d31947b9df2df2da975ac96 Georgi Gerganov ggerganov@gmail.com 2023-03-13T19:15:08+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-13T19:15:08+02:00 Print system information
ed6849cc07a8973e5d31947b9df2df2da975ac96 41be0a3b3d76ee4f254dc81b42bd8ed26ee324e7 Sebastián A sebastian.aedo29@gmail.com 2023-03-13T14:12:33-03:00 GitHub noreply@github.com 2023-03-13T19:12:33+02:00 Initial support for CMake (#75)
41be0a3b3d76ee4f254dc81b42bd8ed26ee324e7 671d5cac15241b495006f56482bf2d6967dca91f Thomas Klausner wiz@gatalith.at 2023-03-13T17:40:54+01:00 GitHub noreply@github.com 2023-03-13T18:40:54+02:00 Add NetBSD support. (#90)
671d5cac15241b495006f56482bf2d6967dca91f 84d9015c4a91ab586ba65d5bd31a8482baf46ba1 Pavol Rusnak pavol@rusnak.io 2023-03-13T17:39:56+01:00 GitHub noreply@github.com 2023-03-13T18:39:56+02:00 Use fprintf for diagnostic output (#48)
84d9015c4a91ab586ba65d5bd31a8482baf46ba1 63fd76fbb06f9b723ca11505352387a3148b1814 Georgi Gerganov ggerganov@gmail.com 2023-03-13T18:36:44+02:00 GitHub noreply@github.com 2023-03-13T18:36:44+02:00 Use vdotq_s32 to improve performance (#67)
63fd76fbb06f9b723ca11505352387a3148b1814 2a20f48efad692a8c2744f10c673bbdbe0c751b7 uint256_t konndennsa@gmail.com 2023-03-14T01:33:43+09:00 GitHub noreply@github.com 2023-03-13T18:33:43+02:00 Reduce model loading time (#43)
2a20f48efad692a8c2744f10c673bbdbe0c751b7 d1f224712d78ab2cbb78777acfeb6739f660eb96 Val Kharitonov mail@kharvd.com 2023-03-13T12:24:18-04:00 GitHub noreply@github.com 2023-03-13T18:24:18+02:00 Fix UTF-8 handling (including colors) (#79)
d1f224712d78ab2cbb78777acfeb6739f660eb96 1808ee0500ea674b4bc2911acd0489ee5cbcef87 Pavol Rusnak pavol@rusnak.io 2023-03-13T17:15:20+01:00 GitHub noreply@github.com 2023-03-13T18:15:20+02:00 Add quantize script for batch quantization (#92)
1808ee0500ea674b4bc2911acd0489ee5cbcef87 a169bb889cfe7b77a798f04fbc573e67ccb4316a Georgi Gerganov ggerganov@gmail.com 2023-03-13T09:42:26+02:00 GitHub noreply@github.com 2023-03-13T09:42:26+02:00 Add initial contribution guidelines
a169bb889cfe7b77a798f04fbc573e67ccb4316a 460c48254098b28d422382a2bbff6a0b3d7f7e17 Matvey Soloviev blackhole89@gmail.com 2023-03-13T04:08:01+01:00 Matvey Soloviev blackhole89@gmail.com 2023-03-13T04:08:01+01:00 Gate signal support on being on a unixoid system. (#74)
460c48254098b28d422382a2bbff6a0b3d7f7e17 c80e2a8f2adeda202cbffe76ef800f134e51f03f Matvey Soloviev blackhole89@gmail.com 2023-03-13T00:35:51+01:00 Matvey Soloviev blackhole89@gmail.com 2023-03-13T01:04:41+01:00 Fix token count accounting
c80e2a8f2adeda202cbffe76ef800f134e51f03f 54a0e66ea0ed3248e6c95a070a2da0bf5c6d4817 Georgi Gerganov ggerganov@gmail.com 2023-03-13T01:28:08+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-13T01:28:08+02:00 Revert "10% performance boost on ARM"
54a0e66ea0ed3248e6c95a070a2da0bf5c6d4817 543c57e991a23121c666561c2837faa09c4a78ca Georgi Gerganov ggerganov@gmail.com 2023-03-13T01:21:03+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-13T01:21:03+02:00 Check for vdotq_s32 availability
543c57e991a23121c666561c2837faa09c4a78ca 113a9e83ebc0f788f861394437087bf3ca0e019b Georgi Gerganov ggerganov@gmail.com 2023-03-13T01:05:24+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-13T01:05:24+02:00 Ammend to previous commit - forgot to update non-QRDMX branch
113a9e83ebc0f788f861394437087bf3ca0e019b 404fac0d623c9eea74ad7a9347da69e33f10984e Georgi Gerganov ggerganov@gmail.com 2023-03-13T00:56:10+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-13T00:56:10+02:00 10% performance boost on ARM
404fac0d623c9eea74ad7a9347da69e33f10984e 1a0a74300f35ad4868715d684d0bc0effdaa9d31 Matvey Soloviev blackhole89@gmail.com 2023-03-12T23:07:34+01:00 GitHub noreply@github.com 2023-03-13T00:07:34+02:00 Fix color getting reset before prompt output done (#65)
1a0a74300f35ad4868715d684d0bc0effdaa9d31 96ea727f4780620b60c1897b538654931411a3fd Georgi Gerganov ggerganov@gmail.com 2023-03-12T23:39:01+02:00 GitHub noreply@github.com 2023-03-12T23:39:01+02:00 Update README.md
96ea727f4780620b60c1897b538654931411a3fd 966195483549e201cff062096a848d9e9833a1a6 Matvey Soloviev blackhole89@gmail.com 2023-03-12T22:13:28+01:00 GitHub noreply@github.com 2023-03-12T23:13:28+02:00 Add interactive mode (#61)
966195483549e201cff062096a848d9e9833a1a6 f385f8dee83d1baf59896b2eb09f1524dc9cde45 Marc Köhlbrugge subscriptions@marckohlbrugge.com 2023-03-13T03:30:08+07:00 GitHub noreply@github.com 2023-03-12T22:30:08+02:00 Fix typo in README (#45)
f385f8dee83d1baf59896b2eb09f1524dc9cde45 02f0c6fe7f9b7be24c7d339aed016e54a92388ea Ben Garney bengarney@users.noreply.github.com 2023-03-12T13:28:36-07:00 GitHub noreply@github.com 2023-03-12T22:28:36+02:00 Allow using prompt files (#59)
02f0c6fe7f9b7be24c7d339aed016e54a92388ea eb062bb012c4e131818dd757a6d3a757fdee3961 beiller beiller@gmail.com 2023-03-12T16:23:15-04:00 GitHub noreply@github.com 2023-03-12T22:23:15+02:00 Add back top_k (#56)
eb062bb012c4e131818dd757a6d3a757fdee3961 7027a97837c351e0a7bc48db2027af368de382db Sebastián A sebastian.aedo29@gmail.com 2023-03-12T17:15:00-03:00 GitHub noreply@github.com 2023-03-12T22:15:00+02:00 Windows fixes (#31)
7027a97837c351e0a7bc48db2027af368de382db 2d555e5b42922cda6dfc0c3ff54df7b1ee4d0ff4 Georgi Gerganov ggerganov@gmail.com 2023-03-12T22:09:26+02:00 GitHub noreply@github.com 2023-03-12T22:09:26+02:00 Update README.md
2d555e5b42922cda6dfc0c3ff54df7b1ee4d0ff4 7c9e54e55e4106f84688245fb15207f6df917e12 Georgi Gerganov ggerganov@gmail.com 2023-03-12T22:08:24+02:00 GitHub noreply@github.com 2023-03-12T22:08:24+02:00 Add CI (#60)
7c9e54e55e4106f84688245fb15207f6df917e12 b9bd1d014113b7498f04ad4d28e6021d5f4cddad Georgi Gerganov ggerganov@gmail.com 2023-03-12T20:59:01+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-12T20:59:01+02:00 Revert "weights_only" arg - this causing more trouble than help
b9bd1d014113b7498f04ad4d28e6021d5f4cddad 129c7d1ea886e52ac1b87ff6184310bab3158806 Oleksandr Nikitin oleksandr@tvori.info 2023-03-12T14:16:33+02:00 GitHub noreply@github.com 2023-03-12T14:16:33+02:00 python/pytorch compat notes (#44)
129c7d1ea886e52ac1b87ff6184310bab3158806 702fddf5c5c3c1377e169ba9ecdfed4cb16c268b beiller beiller@gmail.com 2023-03-12T05:27:42-04:00 GitHub noreply@github.com 2023-03-12T11:27:42+02:00 Add repetition penalty (#20)
702fddf5c5c3c1377e169ba9ecdfed4cb16c268b 7d86e25bf648eb369a3a8388bf239b6b19f7a789 Georgi Gerganov ggerganov@gmail.com 2023-03-12T09:03:25+02:00 GitHub noreply@github.com 2023-03-12T09:03:25+02:00 Clarify meaning of hacking
7d86e25bf648eb369a3a8388bf239b6b19f7a789 a93120236f99e13d77e4b278e47ffcaad4a899e4 Georgi Gerganov ggerganov@gmail.com 2023-03-12T08:41:54+02:00 GitHub noreply@github.com 2023-03-12T08:41:54+02:00 README: add "Supported platforms" + update hot topics
a93120236f99e13d77e4b278e47ffcaad4a899e4 6a9a67f0bee5eed67cf8bf03f74f77619da40d3f deepdiffuser 112834445+deepdiffuser@users.noreply.github.com 2023-03-11T22:36:35-08:00 GitHub noreply@github.com 2023-03-12T08:36:35+02:00 use weights_only in conversion script (#32)
6a9a67f0bee5eed67cf8bf03f74f77619da40d3f da1a4ff01f42d058cfa59806dd5679c0fe5a8604 Pavol Rusnak pavol@rusnak.io 2023-03-12T07:36:03+01:00 GitHub noreply@github.com 2023-03-12T08:36:03+02:00 Add LICENSE (#21)
da1a4ff01f42d058cfa59806dd5679c0fe5a8604 6b2cb6302ffaf8264e33af1dc52e3ea54003e690 Georgi Gerganov ggerganov@gmail.com 2023-03-12T01:26:32+02:00 GitHub noreply@github.com 2023-03-12T01:26:32+02:00 Update README.md
6b2cb6302ffaf8264e33af1dc52e3ea54003e690 4235e3d5b3f4a0e6844f6291322ebb42181345c9 Juraj Bednar juraj@bednar.io 2023-03-11T18:32:20+01:00 GitHub noreply@github.com 2023-03-11T19:32:20+02:00 Fix a typo in model name (#16)
4235e3d5b3f4a0e6844f6291322ebb42181345c9 f1eaff4721153a5a5094fd1bd8cbdae7a3c079cc Georgi Gerganov ggerganov@gmail.com 2023-03-11T18:10:18+02:00 GitHub noreply@github.com 2023-03-11T18:10:18+02:00 Update README.md
f1eaff4721153a5a5094fd1bd8cbdae7a3c079cc a9e58529ea507ac15cd2df4c39d1b9613d6acb6e Georgi Gerganov ggerganov@gmail.com 2023-03-11T17:58:18+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-11T18:04:25+02:00 Add AVX2 support for x86 architectures thanks to @Const-me !
a9e58529ea507ac15cd2df4c39d1b9613d6acb6e 7d9ed7b25fe17db3fc8848b5116d14682864ce8e Georgi Gerganov ggerganov@gmail.com 2023-03-11T17:40:14+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-11T17:40:14+02:00 Fix un-initialized FP16 tables on x86 (#15, #2)
7d9ed7b25fe17db3fc8848b5116d14682864ce8e 0c6803321c818f3f2da4a0693d20128b0f79ad28 Georgi Gerganov ggerganov@gmail.com 2023-03-11T12:44:21+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-11T12:45:01+02:00 Bump memory buffer
0c6803321c818f3f2da4a0693d20128b0f79ad28 f60fa9e50afce35e7ebe1fedf34d4a9327353927 Georgi Gerganov ggerganov@gmail.com 2023-03-11T12:31:21+02:00 GitHub noreply@github.com 2023-03-11T12:31:21+02:00 Update README.md
f60fa9e50afce35e7ebe1fedf34d4a9327353927 7211862c943273fc8ce4b7fdf4c04f9821b7b591 Georgi Gerganov ggerganov@gmail.com 2023-03-11T12:26:46+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-11T12:27:02+02:00 .gitignore models/
7211862c943273fc8ce4b7fdf4c04f9821b7b591 a5c5ae2f545d0d3338f5b2a7840457e35d5bccc1 Georgi Gerganov ggerganov@gmail.com 2023-03-11T12:26:16+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-11T12:27:02+02:00 Update Makefile var + add comment
a5c5ae2f545d0d3338f5b2a7840457e35d5bccc1 ea977e85ecda7b983f0e7b1db20b509998ddc889 Georgi Gerganov ggerganov@gmail.com 2023-03-11T11:34:25+02:00 GitHub noreply@github.com 2023-03-11T11:34:25+02:00 Update README.md
ea977e85ecda7b983f0e7b1db20b509998ddc889 007a8f6f459c6eb56678fdee4c09219ddb85b640 Georgi Gerganov ggerganov@gmail.com 2023-03-11T11:34:11+02:00 GitHub noreply@github.com 2023-03-11T11:34:11+02:00 Update README.md
007a8f6f459c6eb56678fdee4c09219ddb85b640 5f2f970d51a04b783799bc92fd1d006408269f26 Georgi Gerganov ggerganov@gmail.com 2023-03-11T10:47:09+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-11T11:28:30+02:00 Support all LLaMA models + change Q4_0 quantization storage
5f2f970d51a04b783799bc92fd1d006408269f26 73c6ed5e8784a20f89d51b1703a09bc690c68227 Simon Willison swillison@gmail.com 2023-03-10T21:47:26-08:00 GitHub noreply@github.com 2023-03-11T07:47:26+02:00 Include Python dependencies in README (#6)
73c6ed5e8784a20f89d51b1703a09bc690c68227 01eeed8fb1437978603a8523c0b8ea2f6280f5d7 Georgi Gerganov ggerganov@gmail.com 2023-03-11T01:30:47+02:00 GitHub noreply@github.com 2023-03-11T01:30:47+02:00 Update README.md
01eeed8fb1437978603a8523c0b8ea2f6280f5d7 6da2df34ee40301d9ecb126968ec4c0c6195f26d Georgi Gerganov ggerganov@gmail.com 2023-03-11T01:22:58+02:00 GitHub noreply@github.com 2023-03-11T01:22:58+02:00 Update README.md
6da2df34ee40301d9ecb126968ec4c0c6195f26d 9dcf4dba459ba6482a7a5aced22645a387ec6991 Georgi Gerganov ggerganov@gmail.com 2023-03-11T01:18:10+02:00 GitHub noreply@github.com 2023-03-11T01:18:10+02:00 Update README.md
9dcf4dba459ba6482a7a5aced22645a387ec6991 920a7fe2d94cc7e4fed0e88db830b674c91865c5 Jean-Michaël Celerier jeanmichael.celerier+github@gmail.com 2023-03-10T18:04:06-05:00 GitHub noreply@github.com 2023-03-11T01:04:06+02:00 Add missing headers for memcpy and assert (#3)
920a7fe2d94cc7e4fed0e88db830b674c91865c5 3a57ee59de53c2a9d2c3a2c643b609ce07a58a16 Georgi Gerganov ggerganov@gmail.com 2023-03-11T00:55:22+02:00 GitHub noreply@github.com 2023-03-11T00:55:22+02:00 Update README.md
3a57ee59de53c2a9d2c3a2c643b609ce07a58a16 b85028522d6e924473159ba0da3543fc174d2ded Georgi Gerganov ggerganov@gmail.com 2023-03-11T00:51:46+02:00 GitHub noreply@github.com 2023-03-11T00:51:46+02:00 Update README.md
b85028522d6e924473159ba0da3543fc174d2ded 8a01f565ff78cc6c0c5a9fa402787a2f179f2d78 Georgi Gerganov ggerganov@gmail.com 2023-03-11T00:09:19+02:00 GitHub noreply@github.com 2023-03-11T00:09:19+02:00 Update README.md
8a01f565ff78cc6c0c5a9fa402787a2f179f2d78 70bc0b8b15b98dca23b28f0c8f5e34b27e424cda Georgi Gerganov ggerganov@gmail.com 2023-03-10T23:53:11+02:00 GitHub noreply@github.com 2023-03-10T23:53:11+02:00 Update README.md
70bc0b8b15b98dca23b28f0c8f5e34b27e424cda 18ebda34d67c05f4f5584a9209e7efb949f5fd56 Georgi Gerganov ggerganov@gmail.com 2023-03-10T23:46:39+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-10T23:46:57+02:00 Fix a bug in the rope calculation
18ebda34d67c05f4f5584a9209e7efb949f5fd56 319cdb3e1ffe263cf5b08249c9559e011396c1de Georgi Gerganov ggerganov@gmail.com 2023-03-10T21:52:27+02:00 GitHub noreply@github.com 2023-03-10T21:52:27+02:00 Update README.md
319cdb3e1ffe263cf5b08249c9559e011396c1de 775328064e69db1ebd7e19ccb59d2a7fa6142470 Georgi Gerganov ggerganov@gmail.com 2023-03-10T21:50:46+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-10T21:50:46+02:00 Final touches
775328064e69db1ebd7e19ccb59d2a7fa6142470 26c084662903ddaca19bef982831bfb0856e8257 Georgi Gerganov ggerganov@gmail.com 2023-03-10T21:47:46+02:00 GitHub noreply@github.com 2023-03-10T21:47:46+02:00 Create README.md
26c084662903ddaca19bef982831bfb0856e8257 Georgi Gerganov ggerganov@gmail.com 2023-03-10T20:40:58+02:00 Georgi Gerganov ggerganov@gmail.com 2023-03-10T20:56:40+02:00 Initial release
Reference in New Issue View Git Blame Copy Permalink