Increase NCCL timeout from 1 hour to 3 hours (#3343)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
This commit is contained in:
Jieun Park
2023-06-24 15:56:18 +02:00
committed by GitHub
co-authored by pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Glenn Jocher
parent 137552996a
commit 2ebd808b69
+5 -4
View File
@@ -197,10 +197,11 @@ class BaseTrainer:
self.device = torch.device('cuda', RANK)
LOGGER.info(f'DDP info: RANK {RANK}, WORLD_SIZE {world_size}, DEVICE {self.device}')
os.environ['NCCL_BLOCKING_WAIT'] = '1' # set to enforce timeout
dist.init_process_group('nccl' if dist.is_nccl_available() else 'gloo',
timeout=timedelta(seconds=3600),
rank=RANK,
world_size=world_size)
dist.init_process_group(
'nccl' if dist.is_nccl_available() else 'gloo',
timeout=timedelta(seconds=10800), # 3 hours
rank=RANK,
world_size=world_size)
def _setup_train(self, world_size):
"""