happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

model : detect GigaChat3-10-A1.8B as deepseek lite (#17420) 

* Detect GigaChat3-10-A1.8B as deepseek lite

Hardcodes checking number of layers to detect if lite version of deepseek.

* Add commnent identifying deepseek lite variants

deepseek lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B
This commit is contained in:
ubergarm 2025-11-21 08:51:38 -05:00 committed by GitHub
parent 28175f857d
commit 23bc779a6e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/llama-model.cpp
View File

@ -1593,7 +1593,8 @@ void llama_model::load_hparams(llama_model_loader & ml) {
} break; } break;
case LLM_ARCH_DEEPSEEK2: case LLM_ARCH_DEEPSEEK2:
{ {
bool is_lite = (hparams.n_layer == 27); // lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B
bool is_lite = (hparams.n_layer == 27 || hparams.n_layer == 26);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead); ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead);
if (!is_lite) { if (!is_lite) {
@ -4581,7 +4582,8 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
} break; } break;
case LLM_ARCH_DEEPSEEK2: case LLM_ARCH_DEEPSEEK2:
{ {
const bool is_lite = (hparams.n_layer == 27); // lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B
const bool is_lite = (hparams.n_layer == 27 || hparams.n_layer == 26);
const bool is_mla = (hparams.n_embd_head_k_mla != 0 && hparams.n_embd_head_v_mla != 0); const bool is_mla = (hparams.n_embd_head_k_mla != 0 && hparams.n_embd_head_v_mla != 0);

3
src/models/deepseek2.cpp
View File

@ -4,7 +4,8 @@
llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_graph_params & params) : llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_graph_params & params) :
llm_graph_context(params) { llm_graph_context(params) {
bool is_lite = (hparams.n_layer == 27); // lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B
bool is_lite = (hparams.n_layer == 27 || hparams.n_layer == 26);
const bool is_mla = (hparams.n_embd_head_k_mla != 0 && hparams.n_embd_head_v_mla != 0); const bool is_mla = (hparams.n_embd_head_k_mla != 0 && hparams.n_embd_head_v_mla != 0);