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

bench: add --tensor-read-lazy (#27881)

Browse Source 
* bench: add --tensor-read-lazy

* rm the alias

* rename to LLAMA_LAZY_MODE_*
This commit is contained in:
Xuan-Son Nguyen
2026-08-28 20:51:05 +02:00
committed by GitHub
parent 6fe7498016
commit 50f068ffff
10 changed files with 78 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-model-loader.cpp
+2 -2
View File
@@ -1287,10 +1287,10 @@ struct ggml_tensor * llama_model_loader::create_tensor(
return NULL;
}
if ((flags & TENSOR_READ_LAZY) && use_mmap && tensor_read_lazy != LLAMA_TENSOR_READ_LAZY_OFF) {
if ((flags & TENSOR_READ_LAZY) && use_mmap && lazy_mode != LLAMA_LAZY_MODE_OFF) {
// in auto mode, small tensors are cheap enough to keep resident
constexpr size_t auto_lazy_min_size = 4ull * 1024 * 1024 * 1024;
if (tensor_read_lazy == LLAMA_TENSOR_READ_LAZY_ON || ggml_nbytes(cur) > auto_lazy_min_size) {
if (lazy_mode == LLAMA_LAZY_MODE_ON || ggml_nbytes(cur) > auto_lazy_min_size) {
const auto & w = require_weight(tn.str().c_str());
lazy_tensor_ranges[w.idx].emplace_back(w.offs, w.offs + ggml_nbytes(cur));