Revert "Enforce explicit ProcessGroup passed into DefaultState (pytorch#84105)"

This reverts commit adc9a1e.

Reverted pytorch#84105 on behalf of https://github.com/facebook-github-bot due to Diff reverted internally

Author: pytorchmergebot

test/distributed/fsdp/test_fsdp_comm_hooks.py

@@ -7,7 +7,6 @@
 import torch.nn as nn
 import torch.nn.functional as F
 from torch import distributed as dist
-from torch.distributed.distributed_c10d import _get_default_group
 from torch.distributed.algorithms._comm_hooks import default_hooks
 from torch.distributed.fsdp import FullyShardedDataParallel as FSDP
 from torch.distributed.fsdp import MixedPrecision
@@ -424,7 +423,7 @@ def test_fp16_hook(
         sharding_strategy: Optional[ShardingStrategy]
     ):
 
-        state = default_hooks.LowPrecisionState(process_group=_get_default_group())
+        state = default_hooks.LowPrecisionState(process_group=None)
         hook = default_hooks.fp16_compress_hook
 
         self._check_low_precision_hook(state, hook, sharding_strategy, torch.float16, has_wrapping)
@@ -451,7 +450,7 @@ def test_bf16_hook(
         sharding_strategy: Optional[ShardingStrategy]
     ):
 
-        state = default_hooks.LowPrecisionState(process_group=_get_default_group())
+        state = default_hooks.LowPrecisionState(process_group=None)
         hook = default_hooks.bf16_compress_hook
 
         self._check_low_precision_hook(state, hook, sharding_strategy, torch.bfloat16, has_wrapping)

torch/distributed/algorithms/_comm_hooks/default_hooks.py

@@ -1,6 +1,7 @@
 import functools
 import torch
 import torch.distributed as dist
+from torch.distributed import distributed_c10d
 
 
 class DefaultState(object):
@@ -21,11 +22,9 @@ class DefaultState(object):
 
     def __init__(
         self,
-        process_group: dist.ProcessGroup
+        process_group
     ):
-        if process_group is None:
-            raise ValueError(f"Expected to pass in an explicit ProcessGroup to {self}.")
-        self.process_group = process_group
+        self.process_group = process_group if process_group is not None else distributed_c10d._get_default_group()
         self.world_size = dist.get_world_size(process_group)
         # Setting two factors `self.gradient_predivide_factor`
         # and `self.gradient_postdivide_factor` to avoid underflow and overflow