update dags
This commit is contained in:
@@ -83,7 +83,7 @@ def run_training_node_func(rank, world_size):
|
|||||||
os.environ['MASTER_PORT'] = MASTER_PORT
|
os.environ['MASTER_PORT'] = MASTER_PORT
|
||||||
os.environ['WORLD_SIZE'] = str(world_size)
|
os.environ['WORLD_SIZE'] = str(world_size)
|
||||||
os.environ['RANK'] = str(rank)
|
os.environ['RANK'] = str(rank)
|
||||||
# os.environ['NCCL_SOCKET_IFNAME'] = 'eth0'
|
os.environ['NCCL_SOCKET_IFNAME'] = 'eth0'
|
||||||
os.environ['NCCL_DEBUG'] = 'INFO'
|
os.environ['NCCL_DEBUG'] = 'INFO'
|
||||||
os.environ['NCCL_TIMEOUT'] = str(NCCL_TIMEOUT)
|
os.environ['NCCL_TIMEOUT'] = str(NCCL_TIMEOUT)
|
||||||
os.environ['NCCL_BLOCKING_WAIT'] = '1'
|
os.environ['NCCL_BLOCKING_WAIT'] = '1'
|
||||||
|
|||||||
Reference in New Issue
Block a user