vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama/ggml: add LLM training support

Browse source 
more compact progress bar

refactor: llama_prepare_sbatch/ubatch

llama_save_model_to_file

gqa_mode arg for repeat_back

llama_opt_param_filter

ggml_graph_dup force_grads

refactor ggml_opt, fix test-opt
This commit is contained in:
Johannes Gäßler 2024-11-17 14:58:51 +01:00
parent a5203b4465
commit c25557362a
26 changed files with 1294 additions and 339 deletions
Download patch file Download diff file Expand all files Collapse all files

6
common/common.h
View file

@ -697,3 +697,9 @@ const char * const LLM_KV_SPLIT_COUNT = "split.count";
const char * const LLM_KV_SPLIT_TENSORS_COUNT = "split.tensors.count";
}
//
// training utils
//
ggml_opt_dataset_t common_opt_dataset_init(struct llama_context * ctx, const std::vector<llama_token> & tokens, int64_t stride);