Skip to content

Enable CUDA graphs for embed gemma 300m #16844

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

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Enable CUDA graphs for embed gemma 300m #16844

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
6 changes: 5 additions & 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 @@ -2806,6 +2806,8 @@ static bool check_node_graph_compatibility(ggml_cgraph * cgraph,
const std::string ffn_moe_down_bias_prefix = "ffn_moe_down_biased";
const std::string nemotron_h_block_out_prefix = "nemotron_h_block_out";
const std::string mamba2_y_add_d_prefix = "mamba2_y_add_d";
const std::string emGemma_sa_out_prefix = "emGemma_sa_out";
const std::string emGemma_l_out_prefix = "emGemma_l_out";

for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_tensor * node = cgraph->nodes[i];
Expand Down Expand Up @@ -2836,7 +2838,9 @@ static bool check_node_graph_compatibility(ggml_cgraph * cgraph,
strncmp(node->name, ffn_moe_up_bias_prefix.c_str(), ffn_moe_up_bias_prefix.size()) != 0 &&
strncmp(node->name, ffn_moe_down_bias_prefix.c_str(), ffn_moe_down_bias_prefix.size()) != 0 &&
strncmp(node->name, nemotron_h_block_out_prefix.c_str(), nemotron_h_block_out_prefix.size()) != 0 &&
strncmp(node->name, mamba2_y_add_d_prefix.c_str(), mamba2_y_add_d_prefix.size()) != 0) {
strncmp(node->name, mamba2_y_add_d_prefix.c_str(), mamba2_y_add_d_prefix.size()) != 0 &&
strncmp(node->name, emGemma_sa_out_prefix.c_str(), emGemma_sa_out_prefix.size()) != 0 &&
strncmp(node->name, emGemma_l_out_prefix.c_str(), emGemma_l_out_prefix.size()) != 0) {
// disable CUDA graphs for batch size > 1 for now while excluding the matrix-matrix addition as part of Gemma3n's `project_per_layer_input` operation
// by means of matching node names. See
// https://github.com/ggml-org/llama.cpp/blob/f9a31eea06a859e34cecb88b4d020c7f03d86cc4/src/llama-model.cpp#L10199-L10241 and
Expand Down
4 changes: 2 additions & 2 deletions src/llama-model.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11573,7 +11573,7 @@ struct llm_build_gemma_embedding : public llm_graph_context {
cb(cur, "attn_post_norm", il);

ggml_tensor * sa_out = ggml_add(ctx0, cur, inpL);
cb(sa_out, "sa_out", il);
cb(sa_out, "emGemma_sa_out", il);

cur = build_norm(sa_out,
model.layers[il].ffn_norm, NULL,
Expand All @@ -11599,7 +11599,7 @@ struct llm_build_gemma_embedding : public llm_graph_context {
cur = ggml_add(ctx0, cur, sa_out);

cur = build_cvec(cur, il);
cb(cur, "l_out", il);
cb(cur, "emGemma_l_out", il);

// input for next layer
inpL = cur;
Expand Down
Loading