ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-14 15:57:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

Guards against using merge_qkv with split mode "graph"

Browse Source
This commit is contained in:
Kawrakow
2025-12-01 07:11:54 +00:00
parent ee0f02dcb0
commit a8cb1860b3
1 changed files with 7 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/llama-load-tensors.cpp
View File

@@ -2789,6 +2789,13 @@ static void prepare_split_tensors(int split_dim, ggml_context * ctx, ggml_tensor
bool create_tensors_helper::create_tensors() {
const auto tn = LLM_TN(model.arch);
bool use_mmap_buffer = true;
if (ml.merge_qkv && model.split_mode == LLAMA_SPLIT_MODE_GRAPH) {
LLAMA_LOG_WARN("\n========================================================\n");
LLAMA_LOG_WARN("merge_qkv is not compatible with split model 'graph'\n");
LLAMA_LOG_WARN(" => turning off merge_qkv\n");
LLAMA_LOG_WARN("========================================================\n\n");
ml.merge_qkv = false;
}
switch (model.arch) {
case LLM_ARCH_LLAMA:
case LLM_ARCH_REFACT: