Enable all SIM rules except disabled ones (#164645)

`SIM` rules are useful for simplifying boolean expressions and enhances code readability.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/164645
Approved by: https://github.com/ezyang, https://github.com/mlazos

test/distributed/_shard/sharded_tensor/test_sharded_tensor.py

@@ -3074,7 +3074,7 @@ class TestShardedTensorFromLocalShards(ShardedTensorTestBase):
                 wrong_dtype_shards, [10, 10], init_rrefs=True
             )
 
-        tensor_requires_grad = True if self.rank == 0 else False
+        tensor_requires_grad = self.rank == 0
         wrong_requires_grad_shards = [
             sharded_tensor.Shard(
                 torch.randn(
@@ -3121,7 +3121,7 @@ class TestShardedTensorFromLocalShards(ShardedTensorTestBase):
                 wrong_pin_memory_local_shards, [10, 10], init_rrefs=True
             )
 
-        tensor_pin_memory = True if self.rank == 0 else False
+        tensor_pin_memory = self.rank == 0
         wrong_pin_memory_shards_cross_ranks = [
             sharded_tensor.Shard(
                 torch.randn(5, 5, pin_memory=tensor_pin_memory), local_shard_metadata

test/distributed/checkpoint/test_checkpoint.py

@@ -152,7 +152,7 @@ class TestStorageBase:
         self.rank = 0 if not dist.is_initialized() else dist.get_rank()
 
     def _get_ranks(self, name):
-        return self.fail_conf[name] if name in self.fail_conf else None
+        return self.fail_conf.get(name, None)
 
     def _fail_rank(self, name):
         ranks = self._get_ranks(name)

test/distributed/fsdp/test_fsdp_freezing_weights.py

@@ -155,7 +155,7 @@ class TestFreezingWeights(FSDPTest):
 
         ddp_kwargs = {
             "device_ids": [self.rank],
-            "find_unused_parameters": True if disable_autograd else False,
+            "find_unused_parameters": bool(disable_autograd),
         }
 
         model = self._create_model(

test/distributed/pipelining/test_schedule.py

@@ -66,7 +66,7 @@ class MockPipelineStage(_PipelineStageBase):
         self.num_stages = kwargs.get("num_stages", 1)
         self.group_size = kwargs.get("group_size", 1)
         self.group_rank = kwargs.get("group_rank", 0)
-        self.group = kwargs.get("group", None)
+        self.group = kwargs.get("group")
 
     def _create_grad_recv_info(self, *args, **kwargs):
         return None

test/distributed/tensor/test_dtensor.py

@@ -1066,7 +1066,7 @@ class TestDTensorPlacementTypes(DTensorTestBase):
                 assert_array_equal(expected_pad_sizes, pad_sizes)
 
                 is_tensor_empty = [
-                    False if splitted_tensor.numel() > 0 else True
+                    not splitted_tensor.numel() > 0
                     for splitted_tensor in splitted_tensor_list
                 ]
                 expected_is_tensor_empty = [True] * self.world_size
@@ -1089,12 +1089,10 @@ class TestDTensorPlacementTypes(DTensorTestBase):
                     for i, tensor in enumerate(splitted_tensor_list)
                 ]
                 expected_is_tensor_empty = [
-                    False if idx < size else True
-                    for idx, _ in enumerate(range(self.world_size))
+                    not idx < size for idx, _ in enumerate(range(self.world_size))
                 ]
                 is_tensor_empty = [
-                    False if unpadded_tensor.numel() > 0 else True
-                    for unpadded_tensor in unpadded_list
+                    not unpadded_tensor.numel() > 0 for unpadded_tensor in unpadded_list
                 ]
                 assert_array_equal(expected_is_tensor_empty, is_tensor_empty)
 

test/distributed/test_c10d_nccl.py

@@ -2770,11 +2770,7 @@ class WorkHookTest(MultiProcessTestCase):
         # from rank0 to other ranks. However, this is DDP's internal implementation,
         # which is subject to change in future versions.
         self.assertTrue(num_hook_fired[OpType.BROADCAST] > 0)
-        ctor_allreduce = (
-            num_hook_fired[OpType.ALLREDUCE]
-            if OpType.ALLREDUCE in num_hook_fired
-            else 0
-        )
+        ctor_allreduce = num_hook_fired.get(OpType.ALLREDUCE, 0)
 
         x = torch.zeros(2, 1000).cuda(self.rank)
         ddp(x).sum().backward()