141 lines
5.5 KiB
Diff
141 lines
5.5 KiB
Diff
diff --git a/common/common.cpp b/common/common.cpp
|
|
index 625b369b26a5981cc19eda5181226d916acc28ad..898d445293a51a4d1e0037f05bc83b5f773cc430 100644
|
|
--- a/common/common.cpp
|
|
+++ b/common/common.cpp
|
|
@@ -1058,51 +1058,51 @@ struct common_init_result common_init_from_params(common_params & params) {
|
|
params.sampling.dry_penalty_last_n = llama_n_ctx(lctx);
|
|
}
|
|
|
|
- if (params.warmup) {
|
|
- LOG_WRN("%s: warming up the model with an empty run - please wait ... (--no-warmup to disable)\n", __func__);
|
|
-
|
|
- llama_set_warmup(lctx, true);
|
|
-
|
|
- std::vector<llama_token> tmp;
|
|
- llama_token bos = llama_vocab_bos(vocab);
|
|
- llama_token eos = llama_vocab_eos(vocab);
|
|
-
|
|
- // some models (e.g. T5) don't have a BOS token
|
|
- if (bos != LLAMA_TOKEN_NULL) {
|
|
- tmp.push_back(bos);
|
|
- }
|
|
- if (eos != LLAMA_TOKEN_NULL) {
|
|
- tmp.push_back(eos);
|
|
- }
|
|
- if (tmp.empty()) {
|
|
- tmp.push_back(0);
|
|
- }
|
|
-
|
|
- if (llama_model_has_encoder(model)) {
|
|
- #ifdef USE_CALRT
|
|
- //calrt_encode(cal_ctx, llama_batch_get_one(tmp.data(), tmp.size()));
|
|
- #else
|
|
- llama_encode(lctx, llama_batch_get_one(tmp.data(), tmp.size()));
|
|
- #endif
|
|
- llama_token decoder_start_token_id = llama_model_decoder_start_token(model);
|
|
- if (decoder_start_token_id == LLAMA_TOKEN_NULL) {
|
|
- decoder_start_token_id = bos;
|
|
- }
|
|
- tmp.clear();
|
|
- tmp.push_back(decoder_start_token_id);
|
|
- }
|
|
- if (llama_model_has_decoder(model)) {
|
|
- #ifdef USE_CALRT
|
|
- calrt_decode(cal_ctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch)));
|
|
- #else
|
|
- llama_decode(lctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch)));
|
|
- #endif
|
|
- }
|
|
- llama_memory_clear(llama_get_memory(lctx), true);
|
|
- llama_synchronize(lctx);
|
|
- llama_perf_context_reset(lctx);
|
|
- llama_set_warmup(lctx, false);
|
|
- }
|
|
+ // if (params.warmup) {
|
|
+ // LOG_WRN("%s: warming up the model with an empty run - please wait ... (--no-warmup to disable)\n", __func__);
|
|
+
|
|
+ // llama_set_warmup(lctx, true);
|
|
+
|
|
+ // std::vector<llama_token> tmp;
|
|
+ // llama_token bos = llama_vocab_bos(vocab);
|
|
+ // llama_token eos = llama_vocab_eos(vocab);
|
|
+
|
|
+ // // some models (e.g. T5) don't have a BOS token
|
|
+ // if (bos != LLAMA_TOKEN_NULL) {
|
|
+ // tmp.push_back(bos);
|
|
+ // }
|
|
+ // if (eos != LLAMA_TOKEN_NULL) {
|
|
+ // tmp.push_back(eos);
|
|
+ // }
|
|
+ // if (tmp.empty()) {
|
|
+ // tmp.push_back(0);
|
|
+ // }
|
|
+
|
|
+ // if (llama_model_has_encoder(model)) {
|
|
+ // #ifdef USE_CALRT
|
|
+ // //calrt_encode(cal_ctx, llama_batch_get_one(tmp.data(), tmp.size()));
|
|
+ // #else
|
|
+ // llama_encode(lctx, llama_batch_get_one(tmp.data(), tmp.size()));
|
|
+ // #endif
|
|
+ // llama_token decoder_start_token_id = llama_model_decoder_start_token(model);
|
|
+ // if (decoder_start_token_id == LLAMA_TOKEN_NULL) {
|
|
+ // decoder_start_token_id = bos;
|
|
+ // }
|
|
+ // tmp.clear();
|
|
+ // tmp.push_back(decoder_start_token_id);
|
|
+ // }
|
|
+ // if (llama_model_has_decoder(model)) {
|
|
+ // #ifdef USE_CALRT
|
|
+ // calrt_decode(cal_ctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch)));
|
|
+ // #else
|
|
+ // llama_decode(lctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch)));
|
|
+ // #endif
|
|
+ // }
|
|
+ // llama_memory_clear(llama_get_memory(lctx), true);
|
|
+ // llama_synchronize(lctx);
|
|
+ // llama_perf_context_reset(lctx);
|
|
+ // llama_set_warmup(lctx, false);
|
|
+ // }
|
|
|
|
iparams.model.reset(model);
|
|
iparams.context.reset(lctx);
|
|
diff --git a/src/llama-calrt.cpp b/src/llama-calrt.cpp
|
|
index 9f41c955b7bb36cdc4dfc0d50c5fdd76667a2373..e8dc07c6fe6282ff3d7d4e43210b3491982d10f0 100644
|
|
--- a/src/llama-calrt.cpp
|
|
+++ b/src/llama-calrt.cpp
|
|
@@ -43,6 +43,16 @@ std::unique_ptr<calrt::VirtualDevice> init_calrt_vdev() {
|
|
throw std::runtime_error("failed to create vdevice");
|
|
}
|
|
|
|
+ uint64_t commitId = 0;
|
|
+ uint32_t rdata = 0;
|
|
+ vDev->GetDevice()->ReadReg(0x0a238254, rdata);
|
|
+ commitId |=(((uint64_t)rdata) << 32);
|
|
+ vDev->GetDevice()->ReadReg(0x0a238258, rdata);
|
|
+ commitId |=((uint64_t)rdata);
|
|
+
|
|
+ LLAMA_LOG_INFO("CalcoreRT: =================== CalcoreRT Version: %016llx ====================\n", commitId);
|
|
+ LLAMA_LOG_INFO("CalRT: =================== CalRT Version %s ====================\n", calrt::calrt_version());
|
|
+
|
|
return vDev;
|
|
}
|
|
|
|
@@ -664,7 +674,7 @@ calrt::CalbinModel * calrt_context::get_model_by_name(std::string & name) {
|
|
}
|
|
|
|
if (smodelName.empty()) {
|
|
- LLAMA_LOG_ERROR("submodel who's name contains %s not found", name.c_str());
|
|
+ //LLAMA_LOG_ERROR("submodel who's name contains %s not found", name.c_str());
|
|
return nullptr;
|
|
}
|
|
|
|
@@ -681,7 +691,7 @@ calrt::CalbinModel * calrt_context::get_model_by_names(std::string & name1, std:
|
|
}
|
|
|
|
if (smodelName.empty()) {
|
|
- LLAMA_LOG_ERROR("submodel who's name contains %s not found", name1.c_str(), name2.c_str());
|
|
+ //LLAMA_LOG_ERROR("submodel who's name contains %s not found", name1.c_str(), name2.c_str());
|
|
return nullptr;
|
|
}
|
|
|