Skip to content

llama-graph: avoid expand_forward for fusion #17633

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
ggerganov merged 1 commit into from
Dec 1, 2025
Merged

llama-graph: avoid expand_forward for fusion #17633

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
2 changes: 1 addition & 1 deletion ggml/src/ggml-cuda/ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3274,14 +3274,14 @@ static void evaluate_and_capture_cuda_graph(ggml_backend_cuda_context * cuda_ctx
GGML_LOG_DEBUG("Setting stream no to %d for node %s\n", cuda_ctx->curr_stream_no, node->name);
}
}
prev_i = i;

#ifdef GGML_CUDA_DEBUG
const int nodes_fused = i - prev_i - 1;
if (nodes_fused > 0) {
GGML_LOG_INFO("nodes_fused: %d\n", nodes_fused);
}
#endif
prev_i = i;

if (ggml_is_empty(node) || node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) {
continue;
Expand Down
6 changes: 0 additions & 6 deletions src/llama-graph.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -810,9 +810,6 @@ ggml_tensor * llm_graph_context::build_ffn(
GGML_ABORT("fatal error");
}

//expand here so that we can fuse ffn gate
ggml_build_forward_expand(gf, cur);

if (gate && type_gate == LLM_FFN_PAR) {
cur = ggml_mul(ctx0, cur, tmp);
cb(cur, "ffn_gate_par", il);
Expand Down Expand Up @@ -1093,9 +1090,6 @@ ggml_tensor * llm_graph_context::build_moe_ffn(
GGML_ABORT("fatal error");
}

//expand here so that we can fuse ffn gate
ggml_build_forward_expand(gf, cur);

experts = build_lora_mm_id(down_exps, cur, selected_experts); // [n_embd, n_expert_used, n_tokens]
cb(experts, "ffn_moe_down", il);

Expand Down
Loading