EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

misc: update cuda graph capture exception log (#894)

Browse Source
This commit is contained in:
Yineng Zhang
2024-08-02 22:40:52 +08:00
committed by GitHub
parent 7937a886b2
commit 6b8f66efe1
1 changed files with 6 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
python/sglang/srt/model_executor/model_runner.py
View File

@@ -312,10 +312,12 @@ class ModelRunner:
self.cuda_graph_runner.capture(batch_size_list) self.cuda_graph_runner.capture(batch_size_list)
except RuntimeError as e: except RuntimeError as e:
raise Exception( raise Exception(
f"Capture cuda graph failed: {e}. Possible solutions:\n" f"Capture cuda graph failed: {e}\n"
f"1. disable cuda graph by --disable-cuda-graph\n" "Possible solutions:\n"
f"2. set --mem-fraction-static to a smaller value\n" "1. disable torch compile by not using --enable-torch-compile\n"
f"Open an issue on GitHub with reproducible scripts if you need help.\n" "2. disable cuda graph by --disable-cuda-graph\n"
"3. set --mem-fraction-static to a smaller value\n"
"Open an issue on GitHub https://github.com/sgl-project/sglang/issues/new/choose \n"
) )
@torch.inference_mode() @torch.inference_mode()