Просмотр исходного кода

model : remove modern-bert iswa template (#18529)

* remove modern-bert iswa template

* forgotten
Sigbjørn Skjæret 3 недель назад
Родитель
Сommit
169ee68ffb
3 измененных файлов с 3 добавлено и 15 удалено
  1. 1 1
      src/llama-model.cpp
  2. 0 1
      src/models/models.h
  3. 2 13
      src/models/modern-bert.cpp

+ 1 - 1
src/llama-model.cpp

@@ -7457,7 +7457,7 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
             } break;
             } break;
         case LLM_ARCH_MODERN_BERT:
         case LLM_ARCH_MODERN_BERT:
             {
             {
-                llm = std::make_unique<llm_build_modern_bert<true>>(*this, params);
+                llm = std::make_unique<llm_build_modern_bert>(*this, params);
             } break;
             } break;
         case LLM_ARCH_NEO_BERT:
         case LLM_ARCH_NEO_BERT:
             {
             {

+ 0 - 1
src/models/models.h

@@ -332,7 +332,6 @@ struct llm_build_mistral3 : public llm_graph_context {
     llm_build_mistral3(const llama_model & model, const llm_graph_params & params);
     llm_build_mistral3(const llama_model & model, const llm_graph_params & params);
 };
 };
 
 
-template <bool iswa>
 struct llm_build_modern_bert : public llm_graph_context {
 struct llm_build_modern_bert : public llm_graph_context {
     llm_build_modern_bert(const llama_model & model, const llm_graph_params & params);
     llm_build_modern_bert(const llama_model & model, const llm_graph_params & params);
 };
 };

+ 2 - 13
src/models/modern-bert.cpp

@@ -1,7 +1,6 @@
 #include "models.h"
 #include "models.h"
 
 
-template <bool iswa>
-llm_build_modern_bert<iswa>::llm_build_modern_bert(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+llm_build_modern_bert::llm_build_modern_bert(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
     const int64_t n_embd_head = hparams.n_embd_head_v;
     const int64_t n_embd_head = hparams.n_embd_head_v;
     const int64_t n_embd_gqa  = hparams.n_embd_v_gqa();
     const int64_t n_embd_gqa  = hparams.n_embd_v_gqa();
 
 
@@ -24,13 +23,7 @@ llm_build_modern_bert<iswa>::llm_build_modern_bert(const llama_model & model, co
     auto * inp_attn = build_attn_inp_no_cache();
     auto * inp_attn = build_attn_inp_no_cache();
 
 
     for (int il = 0; il < n_layer; ++il) {
     for (int il = 0; il < n_layer; ++il) {
-        float freq_base_l  = 0.0f;
-
-        if constexpr (iswa) {
-            freq_base_l = model.get_rope_freq_base(cparams, il);
-        } else {
-            freq_base_l = freq_base;
-        }
+        float freq_base_l = model.get_rope_freq_base(cparams, il);
 
 
         cur = inpL;
         cur = inpL;
 
 
@@ -120,7 +113,3 @@ llm_build_modern_bert<iswa>::llm_build_modern_bert(const llama_model & model, co
     res->t_embd = cur;
     res->t_embd = cur;
     ggml_build_forward_expand(gf, cur);
     ggml_build_forward_expand(gf, cur);
 }
 }
-
-// Explicit template instantiations
-template struct llm_build_modern_bert<false>;
-template struct llm_build_modern_bert<true>;