CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-12 10:47:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

cuda : Enable CUDA Graph usage for Nemotron Nano v2 (NemotronH) (#16328)

Browse Source 
* Fix Nemotron Nano v2 9B not executing as CUDA Graph on NVIDIA GPUs

* fix to ensure test-backend-ops check passes
This commit is contained in:
anavp-nvidia
2025-09-30 08:13:22 +00:00
committed by GitHub
parent 35fb82497e
commit a014310374
3 changed files with 20 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -2641,6 +2641,8 @@ static bool check_node_graph_compatibility_and_refresh_copy_ops(ggml_backend_cud
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";
for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_tensor * node = cgraph->nodes[i];
@@ -2669,7 +2671,9 @@ static bool check_node_graph_compatibility_and_refresh_copy_ops(ggml_backend_cud
(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, 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) {
// 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