Skip to content

[ROCm] Use unreg path for aiter custom all-reduce during CUDA graph capture #20155

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 2 commits into from
Mar 9, 2026
Merged

[ROCm] Use unreg path for aiter custom all-reduce during CUDA graph capture #20155

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
dd77542
1
zyzshishui Mar 9, 2026
9561ddc
Merge branch 'main' into 1
zyzshishui Mar 9, 2026
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
7 changes: 6 additions & 1 deletion 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 @@ -4,6 +4,7 @@
import logging
import os
from contextlib import contextmanager
from functools import partial
from typing import Any, List, Optional, Union

import torch
Expand Down Expand Up @@ -495,7 +496,11 @@ def dispatch_custom_allreduce():
)

logger.info("[AR] Using AiterCustomAllreduce (AMD default)")
return AiterCustomAllreduce
tms_cudagraph = envs.SGLANG_MEMORY_SAVER_CUDA_GRAPH.get()
return partial(
AiterCustomAllreduce,
enable_register_for_capturing=not tms_cudagraph,
)
except ImportError as e:
logger.warning(
"[AR] Aiter custom all-reduce not available; "
Expand Down
Loading