CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-06 09:46:50 +00:00
Code Issues Packages Projects Releases Wiki Activity

bugfix: fix the arch check for qwen3vl-moe.

Browse Source
This commit is contained in:
JJJYmmm
2025-10-26 20:00:33 +08:00
parent 1e4fd19446
commit f84bd67c80
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/llama-model.cpp
View File

@@ -3350,7 +3350,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
{ {
// for deepstack features, we consider the embd to be [main_embd, deepstack_embd_1, deepstack_embd_2, ...] // for deepstack features, we consider the embd to be [main_embd, deepstack_embd_1, deepstack_embd_2, ...]
int64_t n_embd = hparams.n_embd; int64_t n_embd = hparams.n_embd;
if (arch == LLM_ARCH_QWEN3VL) { if (arch == LLM_ARCH_QWEN3VLMOE) {
n_embd = hparams.n_embd / (hparams.n_deepstack_layers + 1); n_embd = hparams.n_embd / (hparams.n_deepstack_layers + 1);
} }
tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0); tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);