bugfixed: stuck when training with dist_train.sh (#728)

open-mmlab · Dec 27, 2021 · 1f5b787 · 1f5b787
1 parent ec98288
commit 1f5b787
Showing 1 changed file with 3 additions and 3 deletions.
diff --git a/pcdet/utils/common_utils.py b/pcdet/utils/common_utils.py
@@ -166,9 +166,9 @@ def init_dist_pytorch(tcp_port, local_rank, backend='nccl'):
     torch.cuda.set_device(local_rank % num_gpus)
     dist.init_process_group(
         backend=backend,
-        init_method='tcp://127.0.0.1:%d' % tcp_port,
-        rank=local_rank,
-        world_size=num_gpus
+        # init_method='tcp://127.0.0.1:%d' % tcp_port,
+        # rank=local_rank,
+        # world_size=num_gpus
     )
     rank = dist.get_rank()
     return num_gpus, rank