vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update CUDA graph on scale change plus clear nodes/params (#9550)

Browse source 
* Avoid using saved CUDA graph if scale changes and reset nodes/params on update

Fixes https://github.com/ggerganov/llama.cpp/issues/9451

* clear before resize
This commit is contained in:
agray3 2024-09-21 01:41:07 +01:00 • committed by GitHub
parent e948a7da7a
commit 41f477879f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 10 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ggml/src/ggml-cuda/common.cuh
View file

@ -569,6 +569,7 @@ struct ggml_graph_node_properties {
int64_t ne[GGML_MAX_DIMS];
size_t nb[GGML_MAX_DIMS];
void * src_address[GGML_MAX_SRC];
int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(int32_t)];
};
struct ggml_cuda_graph {