vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-10-03 11:35:46 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

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 18:54:47 +05:30 • committed by GitHub
parent 7d6f5d02bb
commit 7ceed8737f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 4 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/models/nemotron-h.cpp
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
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);