Skip to content

CUDA: Enable cuda graphs for qwen3 next-style architectures #19521

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

Closed
ORippler wants to merge 3 commits into from
Closed

CUDA: Enable cuda graphs for qwen3 next-style architectures #19521

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
7f09585
Add exceptions for qwen 3 next to batch-size-heuristic
ORippler Feb 11, 2026
89decb7
Apply code-formater
ORippler Feb 11, 2026
d2c2b40
Do strncmp instead of regex, as asked for by @ngxson
ORippler Feb 13, 2026
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
19 changes: 11 additions & 8 deletions ggml/src/ggml-cuda/ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2867,11 +2867,13 @@ static bool ggml_cuda_graph_check_compability(ggml_cgraph * cgraph) {

const std::string gemma3n_per_layer_proj_src0_name = "inp_per_layer_selected";
const std::string gemma3n_per_layer_proj_src1_name = "per_layer_proj";
const std::string ffn_moe_gate_bias_prefix = "ffn_moe_gate_biased";
const std::string ffn_moe_up_bias_prefix = "ffn_moe_up_biased";
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 ffn_moe_gate_bias_prefix = "ffn_moe_gate_biased";
const std::string ffn_moe_up_bias_prefix = "ffn_moe_up_biased";
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 qwen3next_diag_mask = "diag_mask";
const std::string delta_net_linear_attn_prefix = "new_state-";

for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_tensor * node = cgraph->nodes[i];
Expand All @@ -2894,15 +2896,16 @@ static bool ggml_cuda_graph_check_compability(ggml_cgraph * cgraph) {
#endif
}

if (node->op == GGML_OP_ADD &&
node->src[1] && node->src[1]->ne[1] > 1 &&
if (node->op == GGML_OP_ADD && node->src[1] && node->src[1]->ne[1] > 1 &&
(node->src[0] ? node->src[0]->name != gemma3n_per_layer_proj_src0_name : true) &&
(node->src[1] ? node->src[1]->name != gemma3n_per_layer_proj_src1_name : true) &&
strncmp(node->name, ffn_moe_gate_bias_prefix.c_str(), ffn_moe_gate_bias_prefix.size()) != 0 &&
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, qwen3next_diag_mask.c_str(), qwen3next_diag_mask.size()) != 0 &&
strncmp(node->name, delta_net_linear_attn_prefix.c_str(), delta_net_linear_attn_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
1 change: 1 addition & 0 deletions src/models/qwen3next.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@ llm_build_qwen3next::llm_build_qwen3next(const llama_model & model, const llm_gr

ggml_tensor * identity = ggml_diag(ctx0, ggml_fill_inplace(ctx0, ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, CHUNK_SIZE), 1.0f));
ggml_tensor * diag_mask = ggml_add(ctx0, causal_mask, identity);
ggml_set_name(diag_mask, "diag_mask");

ggml_build_forward_expand(gf, causal_mask);
ggml_build_forward_expand(gf, identity);
Expand Down
Loading