LLM-Inference/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-08-12 22:29:39 +04:00
Code Issues Packages Projects Releases Wiki Activity

Let's tell the user what we did

Browse Source
This commit is contained in:
Kawrakow
2026-08-04 16:24:07 +00:00
parent 6254979861
commit 2015e6262c
1 changed files with 5 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-hparams.cpp
+5
View File
@@ -1665,6 +1665,11 @@ void llm_load_hparams(
{
ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false);
if (model.arch == LLM_ARCH_DEEPSEEK4 && hparams.n_layer == 43 && hparams.nextn_predict_layers > 0) {
LLAMA_LOG_WARN("===============================================================================================\n");
LLAMA_LOG_WARN("Unexpected number of layers (%d) and nextn_predict_layers (%d) for DeepSeek4-Flash\n",
hparams.n_layer, hparams.nextn_predict_layers);
LLAMA_LOG_WARN(" -> setting nextn_predict_layers to zero\n");
LLAMA_LOG_WARN("===============================================================================================\n");
hparams.nextn_predict_layers = 0;
}
// Probe the first appended predictor block, or n_layer - 1 for base GGUFs.