tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-08 18:04:54 -04:00
Code Issues Packages Projects Releases Wiki Activity

ggml: Re-enable CUDA graphs in presence of CONT and DUP nodes (#12970)

Browse Source
This commit is contained in:
Alan Gray
2025-04-17 14:19:42 +01:00
committed by GitHub
parent 7a395f67a7
commit 207c22ec2d
3 changed files with 7 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -2489,7 +2489,7 @@ static bool check_node_graph_compatibility_and_refresh_copy_ops(ggml_backend_cud
#endif
}
if (node->op == GGML_OP_MUL_MAT_ID || node->op == GGML_OP_CONT || node->op == GGML_OP_DUP) {
if (node->op == GGML_OP_MUL_MAT_ID) {
use_cuda_graph = false; // This node type is not supported by CUDA graph capture
#ifndef NDEBUG
GGML_LOG_DEBUG("%s: disabling CUDA graphs due to unsupported node type\n", __func__);