Skip to content

[ROCm] Use unreg path for custom all-reduce during CUDA graph capture #19162

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

Merged
HaiShaw merged 1 commit into from
Feb 23, 2026
Merged

[ROCm] Use unreg path for custom all-reduce during CUDA graph capture #19162

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 2 additions & 0 deletions python/sglang/srt/distributed/device_communicators/custom_all_reduce.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -411,6 +411,8 @@ def custom_all_reduce(self, input: torch.Tensor) -> Optional[torch.Tensor]:
if self._IS_CAPTURING:
if torch.cuda.is_current_stream_capturing():
if _is_hip:
if self.tms_cudagraph:
return self.all_reduce_unreg(input)
return self.all_reduce_reg(input)
Comment on lines +414 to 416
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Feb 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

For better readability and conciseness, you can use a conditional expression (ternary operator) to simplify this logic.

                    return self.all_reduce_unreg(input) if self.tms_cudagraph else self.all_reduce_reg(input)

else:
return self.all_reduce(input, registered=not self.tms_cudagraph)
Expand Down
Loading