Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

llama: model_loader: add TENSOR_READ_LAZY (#27794)

Browse Source 
* llama: model_loader: add TENSOR_GET_ROW_LAZY

* add --tensor-read-lazy

* rename to TENSOR_READ_LAZY

* gen docs

* address comments
This commit is contained in:
Xuan-Son Nguyen
2026-08-27 15:14:34 +02:00
committed by GitHub
parent cae63579b6
commit fac889fb38
15 changed files with 129 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/models/gemma4.cpp
+1 -1
View File
@@ -50,7 +50,7 @@ void llama_model_gemma4::load_arch_tensors(llama_model_loader &) {
tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
if (n_embd_per_layer > 0) {
per_layer_tok_embd = create_tensor(tn(LLM_TENSOR_PER_LAYER_TOKEN_EMBD, "weight"), {n_embd_per_layer * n_layer, n_vocab}, 0);
per_layer_tok_embd = create_tensor(tn(LLM_TENSOR_PER_LAYER_TOKEN_EMBD, "weight"), {n_embd_per_layer * n_layer, n_vocab}, TENSOR_READ_LAZY);
per_layer_model_proj = create_tensor(tn(LLM_TENSOR_PER_LAYER_MODEL_PROJ, "weight", 0), {n_embd, n_embd_per_layer * n_layer}, 0);
per_layer_proj_norm = create_tensor(tn(LLM_TENSOR_PER_LAYER_PROJ_NORM, "weight", 0), {n_embd_per_layer}, 0);
}