EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

llama/ggml: add LLM training support (#10544)

Browse Source 
* llama/ggml: add LLM training support

more compact progress bar

llama_save_model_to_file

llama_opt_param_filter

ggml_graph_dup force_grads

refactor ggml_opt, fix test-opt

* remove logits_all

* refactor CUDA implementation for ACC

* reset graph at beginning of opt period
This commit is contained in:
Johannes Gäßler
2025-05-12 14:44:49 +02:00
committed by GitHub
parent 064cc596ac
commit 10d2af0eaa
31 changed files with 1415 additions and 359 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/llama.cpp
View File

@@ -4,6 +4,7 @@
#include "llama-mmap.h"
#include "llama-vocab.h"
#include "llama-model-loader.h"
#include "llama-model-saver.h"
#include "llama-model.h"
#include "ggml.h"
@@ -253,6 +254,13 @@ struct llama_model * llama_model_load_from_splits(
return llama_model_load_from_file_impl(splits.front(), splits, params);
}
void llama_model_save_to_file(const struct llama_model * model, const char * path_model) {
llama_model_saver ms(*model);
ms.add_kv_from_model();
ms.add_tensors_from_model();
ms.save(path_model);
}
//
// chat templates
//
@@ -338,3 +346,4 @@ const char * llama_print_system_info(void) {
return s.c_str();
}