vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-20 17:54:31 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

model : support NVFP4 tensors for Gemma4 (#21971)

Browse source 
* support nvfp4 tensors for Gemma4

* add wo_s to build_attn

* add wo_s to build_attn

* fix glm4
This commit is contained in:
Sigbjørn Skjæret 2026-04-16 16:51:47 +02:00 • committed by GitHub
parent b572d1ecd6
commit f772f6e434
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
105 changed files with 149 additions and 148 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/models/starcoder.cpp
View file

@ -48,7 +48,7 @@ llm_build_starcoder::llm_build_starcoder(const llama_model & model, const llm_gr
cb(Vcur, "Vcur", il);
cur = build_attn(inp_attn,
model.layers[il].wo, model.layers[il].bo,
model.layers[il].wo, model.layers[il].bo, model.layers[il].wo_s,
Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
}
if (il == n_layer - 1 && inp_out_ids) {