Skip to content

graph : add optional scale parameter to build_lora_mm #20427

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
CISC merged 1 commit into from
Mar 11, 2026
Merged

graph : add optional scale parameter to build_lora_mm #20427

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion src/llama-graph.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -900,7 +900,8 @@ ggml_tensor * llm_graph_context::build_cvec(

ggml_tensor * llm_graph_context::build_lora_mm(
ggml_tensor * w,
ggml_tensor * cur) const {
ggml_tensor * cur,
ggml_tensor * w_s) const {
ggml_tensor * res = ggml_mul_mat(ctx0, w, cur);

for (const auto & lora : *loras) {
Expand All @@ -921,6 +922,10 @@ ggml_tensor * llm_graph_context::build_lora_mm(
res = ggml_add(ctx0, res, ab_cur);
}

if (w_s) {
res = ggml_mul(ctx0, res, w_s);
}

return res;
}

Expand Down
5 changes: 3 additions & 2 deletions src/llama-graph.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -764,10 +764,11 @@ struct llm_graph_context {
ggml_tensor * cur,
int il) const;

// do mat_mul, while optionally apply lora
// do mat_mul, while optionally apply lora and per-tensor scale
ggml_tensor * build_lora_mm(
ggml_tensor * w,
ggml_tensor * cur) const;
ggml_tensor * cur,
ggml_tensor * w_s = nullptr) const;
Comment on lines 768 to +771
Copy link
Member

@ggerganov ggerganov Mar 11, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For a follow-up PR, we should move the cur at the front, to increase consistency of the interfaces:

    ggml_tensor * build_lora_mm(
              ggml_tensor * cur,
              ggml_tensor * w,
              ggml_tensor * w_s = nullptr) const;

It will touch a lot of lines though.


// do mat_mul_id, while optionally apply lora
ggml_tensor * build_lora_mm_id(
Expand Down
25 changes: 5 additions & 20 deletions src/models/bitnet.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,32 +29,23 @@ llm_build_bitnet::llm_build_bitnet(const llama_model & model, const llm_graph_pa
// self-attention
{
// 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);
}

// B1.K
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);
}

// B1.V
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);
Expand Down Expand Up @@ -90,10 +81,7 @@ llm_build_bitnet::llm_build_bitnet(const llama_model & model, const llm_graph_pa
LLM_NORM_RMS, il);
cb(cur, "attn_sub_norm", il);

cur = build_lora_mm(model.layers[il].wo, cur);
if (model.layers[il].wo_s) {
cur = ggml_mul(ctx0, cur, model.layers[il].wo_s);
}
cur = build_lora_mm(model.layers[il].wo, cur, model.layers[il].wo_s);
if (model.layers[il].bo) {
cur = ggml_add(ctx0, cur, model.layers[il].bo);
}
Expand Down Expand Up @@ -127,10 +115,7 @@ llm_build_bitnet::llm_build_bitnet(const llama_model & model, const llm_graph_pa
LLM_NORM_RMS, il);
cb(cur, "ffn_sub_norm", il);

cur = build_lora_mm(model.layers[il].ffn_down, cur);
if (model.layers[il].ffn_down_s) {
cur = ggml_mul(ctx0, cur, model.layers[il].ffn_down_s);
}
cur = build_lora_mm(model.layers[il].ffn_down, cur, model.layers[il].ffn_down_s);
cb(cur, "ffn_down", il);

cur = ggml_add(ctx0, cur, ffn_inp);
Expand Down
15 changes: 3 additions & 12 deletions src/models/llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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);
Expand Down
15 changes: 3 additions & 12 deletions src/models/qwen3.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,22 +30,13 @@ llm_build_qwen3::llm_build_qwen3(const llama_model & model, const llm_graph_para
// self-attention
{
// 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);

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);

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);

Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Expand Down
15 changes: 3 additions & 12 deletions src/models/qwen3moe.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,22 +30,13 @@ llm_build_qwen3moe::llm_build_qwen3moe(const llama_model & model, const llm_grap
// self_attention
{
// 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);

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);

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);

Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Expand Down
Loading