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

llama-quant: add support for mmproj (#16592)

Browse Source 
* llama-quant: add support for mmproj

* Update src/llama.cpp

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* check prefix instead

* small fix

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Xuan-Son Nguyen
2025-10-15 14:48:08 +02:00
committed by GitHub
parent 5acd455460
commit 3e3cb19f64
5 changed files with 19 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama-model.cpp
View File

@@ -478,7 +478,8 @@ void llama_model::load_hparams(llama_model_loader & ml) {
ml.get_key(LLM_KV_GENERAL_NAME, name, false);
// everything past this point is not vocab-related
if (hparams.vocab_only) {
// for CLIP models, we only need to load tensors, no hparams
if (hparams.vocab_only || ml.get_arch() == LLM_ARCH_CLIP) {
return;
}
@@ -20013,6 +20014,7 @@ int32_t llama_n_head(const llama_model * model) {
llama_rope_type llama_model_rope_type(const llama_model * model) {
switch (model->arch) {
// these models do not use RoPE
case LLM_ARCH_CLIP:
case LLM_ARCH_GPT2:
case LLM_ARCH_GPTJ:
case LLM_ARCH_MPT: