nicolas
/
ik_llama_opt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix compiler warning

This commit is contained in:
Kawrakow 2026-04-17 06:04:43 +00:00
parent 7b6507ddac
commit 64234e3c4e
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/llama-load-tensors.cpp
View File

@ -312,7 +312,7 @@ create_tensors_helper::create_tensors_helper(llama_model_loader & _ml, llama_mod
// Split MTP layer's to graph // Split MTP layer's to graph
if ((model.split_mode == LLAMA_SPLIT_MODE_GRAPH || model.split_mode == LLAMA_SPLIT_MODE_ATTN) && if ((model.split_mode == LLAMA_SPLIT_MODE_GRAPH || model.split_mode == LLAMA_SPLIT_MODE_ATTN) &&
model.hparams.nextn_predict_layers > 0 && model.splits.size() > 1) { model.hparams.nextn_predict_layers > 0 && model.splits.size() > 1) {
int mtp_first = n_layer - model.hparams.nextn_predict_layers; [[maybe_unused]] int mtp_first = n_layer - model.hparams.nextn_predict_layers;
LLAMA_LOG_DEBUG("%s: MTP layer(s) %d-%d: split attention+FFN, nextn on per-device CUDA\n", LLAMA_LOG_DEBUG("%s: MTP layer(s) %d-%d: split attention+FFN, nextn on per-device CUDA\n",
__func__, mtp_first, n_layer - 1); __func__, mtp_first, n_layer - 1);
} }