ClaudeBot/llama.cpp
0
0
Fork 1
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-17 20:31:47 +02:00
Code Issues Packages Projects Releases Wiki Activity

models : allow Nemotron-H models to only define layer_norm_epsilon (#28989)

Browse Source 
* allows nemotron models to get by with just defining layer_norm_epsilon

* made changes to load_arch_hparams instead
This commit is contained in:
Kartik Gulia
2026-09-16 15:24:47 +02:00
committed by GitHub
parent 7d6f5d02bb
commit 7ceed8737f
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/models/nemotron-h.cpp
+3 -1
View File
@@ -15,8 +15,10 @@ void llama_model_nemotron_h::load_arch_hparams(llama_model_loader & ml) {
hparams.is_recr_impl[i] = i < hparams.n_layer() && hparams.n_head_kv(i) == 0 && hparams.n_ff(i) == 0;
}
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); // MTP head final_layernorm
if (!ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps, false)) {
hparams.f_norm_rms_eps = hparams.f_norm_eps;
}
// Puzzle models set a different expert FFN size per layer
ml.get_key_or_arr(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp_arr, hparams.n_layer_all, false);