thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

graph : add optional scale parameter to build_lora_mm [no ci] (#20427)

Browse source
This commit is contained in:
Richard Davison 2026-03-12 00:22:49 +01:00 • committed by GitHub
parent 4a748b8f15
commit 1eea6a2968
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 23 additions and 59 deletions
Download patch file Download diff file Expand all files Collapse all files

15
src/models/llama.cpp
View file

@ -43,28 +43,19 @@ llm_build_llama<embed>::llm_build_llama(const llama_model & model, const llm_gra
ggml_tensor * rope_factors = model.get_rope_factors(cparams, il);
// compute Q and K and RoPE them
ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
if (model.layers[il].wq_s) {
Qcur = ggml_mul(ctx0, Qcur, model.layers[il].wq_s);
}
ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur, model.layers[il].wq_s);
cb(Qcur, "Qcur", il);
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
if (model.layers[il].wk_s) {
Kcur = ggml_mul(ctx0, Kcur, model.layers[il].wk_s);
}
ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur, model.layers[il].wk_s);
cb(Kcur, "Kcur", il);
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
if (model.layers[il].wv_s) {
Vcur = ggml_mul(ctx0, Vcur, model.layers[il].wv_s);
}
ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur, model.layers[il].wv_s);
cb(Vcur, "Vcur", il);
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);