sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-05-13 20:44:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

model : fix model type check for granite/llama3 and deepseek2/glm4.7 lite (#22870)

Browse Source
This commit is contained in:
Sigbjørn Skjæret
2026-05-10 08:44:29 +02:00
committed by GitHub
parent 1e5ad35d56
commit 5755a100cd
3 changed files with 4 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama-model.cpp
View File

@@ -1131,10 +1131,6 @@ void llama_model_base::load_hparams(llama_model_loader & ml) {
ml.get_key(LLM_KV_ROPE_DIMENSION_COUNT_SWA, hparams.n_rot_swa, false);
}
// for differentiating model types
uint32_t n_vocab = 0;
ml.get_key(LLM_KV_VOCAB_SIZE, n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, n_vocab, false);
// for classifier models
ml.get_arr(LLM_KV_CLASSIFIER_OUTPUT_LABELS, classifier_labels, false);
if (!classifier_labels.empty()) {

3
src/models/deepseek2.cpp
View File

@@ -1,7 +1,8 @@
#include "models.h"
void llama_model_deepseek2::load_arch_hparams(llama_model_loader & ml) {
const auto n_vocab = vocab.n_tokens();
uint32_t n_vocab = 0;
ml.get_key(LLM_KV_VOCAB_SIZE, n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, n_vocab, false);
// lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B, Kanana-2-30B-A3B
const bool is_lite = (hparams.n_layer == 27 || hparams.n_layer == 26 || (hparams.n_layer == 48 && n_vocab == 128256));

3
src/models/llama.cpp
View File

@@ -1,7 +1,8 @@
#include "models.h"
void llama_model_llama::load_arch_hparams(llama_model_loader & ml) {
const auto n_vocab = vocab.n_tokens();
uint32_t n_vocab = 0;
ml.get_key(LLM_KV_VOCAB_SIZE, n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, n_vocab, false);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);