Skip to content

use global rank for flight recorder #269

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

Closed
wants to merge 1 commit into from
Closed

use global rank for flight recorder #269

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
6 changes: 2 additions & 4 deletions torchft/process_group.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -738,7 +738,6 @@ def __init__(self, timeout: timedelta = timedelta(seconds=60.0)) -> None:
self._use_abort: bool = torch.cuda.nccl.version() >= (2, 25)

self._errored: Optional[Exception] = None
self._rank: int = 0

NONBLOCKING_TIMEOUT_ENV = "TORCH_NCCL_NONBLOCKING_TIMEOUT"
if NONBLOCKING_TIMEOUT_ENV not in os.environ:
Expand Down Expand Up @@ -788,7 +787,6 @@ def _create_pg(self, store: Store, rank: int, world_size: int) -> BaseProcessGro
from torch.distributed import ProcessGroupNCCL as BaseProcessGroupNCCL

self._errored = None
self._rank = rank

# pyre-fixme[16]: no attribute ProcessGroupNCCL
opts = BaseProcessGroupNCCL.Options()
Expand All @@ -811,8 +809,8 @@ def abort(self) -> None:
# We need to set the error before aborting to ensure that errored()
# returns the error correctly when NCCL abort fires and unblocks the
# stream.
if os.environ.get("TORCHFT_TRIGGER_FR_ON_ABORT", "true") == "true":
trigger_nccl_fr_trace_through_pipe(self._rank)
if os.environ.get("TORCHFT_TRIGGER_FR_ON_ABORT", "false") == "true":
trigger_nccl_fr_trace_through_pipe(dist.get_rank())
self._errored = RuntimeError("aborted")

super().abort()
Expand Down
Loading