Skip to content

[PyTorch] Distributed tests set device in NCCL init and manually shut down NCCL #3655

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

Open
timmoon10 wants to merge 4 commits into
base: main
Choose a base branch
from
Open

[PyTorch] Distributed tests set device in NCCL init and manually shut down NCCL #3655

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
379357b
Distributed tests set device in NCCL init and manually shuts down NCCL
timmoon10 Oct 8, 2026
04c9194
Fully destroy NCCL environment when finishing distributed TE ops test
timmoon10 Oct 8, 2026
009bbb2
Merge branch 'main' into tmoon/debug-dist-tests
timmoon10 Oct 9, 2026
1910760
Apply suggestion from @timmoon10
timmoon10 Oct 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
15 changes: 11 additions & 4 deletions tests/pytorch/distributed/test_fusible_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,7 @@ def world_group() -> torch.distributed.ProcessGroup:
init_method=f"file://{os.environ['NVTE_TEST_RDZV_PATH']}",
world_size=world_size,
rank=rank,
device_id=torch.device("cuda", rank),
)
return group

Expand Down Expand Up @@ -1124,10 +1125,16 @@ def main() -> None:
if args.compile and not args.parallel:
parser.error("--compile requires --parallel")
if args.parallel:
if args.compile:
run_compile_parallel_tests()
else:
run_parallel_tests()
try:
if args.compile:
run_compile_parallel_tests()
else:
run_parallel_tests()
torch.distributed.barrier(world_group())
torch.cuda.synchronize()
Comment thread
timmoon10 marked this conversation as resolved.
finally:
if torch.distributed.is_initialized():
torch.distributed.destroy_process_group()
Comment thread
timmoon10 marked this conversation as resolved.


if __name__ == "__main__":
Expand Down
1 change: 1 addition & 0 deletions tests/pytorch/distributed/test_parallel_cross_entropy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,6 +29,7 @@ def _run_tensor_parallel(rank, world_size, init_file, label_smoothing):
init_method=f"file://{init_file}",
rank=rank,
world_size=world_size,
device_id=device,

@vthumbe1503 vthumbe1503 Oct 9, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

Would be good to understand why device_id needs to be set manually.

@timmoon10 and I had offline discussion regarding this. It might be because this particular test doesnt use torchrun like other tests do, and so the device is not auto-set to "cuda:rank" like other tests

@ptrendx ptrendx Oct 9, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

I do actually have a diff already that changes this test to torchrun to avoid this. Can make PR today. @timmoon10

)
try:
generator = torch.Generator().manual_seed(2025)
Expand Down
Loading