Replace list(map(...)) constructs by list comprehensions (#46461)

Flamefire · facebook-github-bot · commit 5b0f40048899 · 2020-10-19T18:42:49.000-07:00
Summary: As discussed in #46392 this makes the code more readable and possibly more performant. It also fixes a bug detected by this where the argument order of `map` was confused: Flamefire@030a249#diff-5bb26bd3a23ee3bb540aeadcc0385df2a4e48de39f87ed9ea76b21990738fe98L1537-R1537 Fixes #46392 Pull Request resolved: #46461 Reviewed By: ailzhang Differential Revision: D24367015 Pulled By: ezyang fbshipit-source-id: d55a67933cc22346b00544c9671f09982ad920e7
diff --git a/.circleci/cimodel/data/simple/util/versions.py b/.circleci/cimodel/data/simple/util/versions.py
@@ -9,7 +9,7 @@ def prefixed_parts(self):
         with the prefix string.
         """
         if self.parts:
-            return [self.prefix + str(self.parts[0])] + list(map(str, self.parts[1:]))
+            return [self.prefix + str(self.parts[0])] + [str(part) for part in self.parts[1:]]
         else:
             return [self.prefix]
 
diff --git a/caffe2/python/fused_8bit_rowwise_conversion_ops_test.py b/caffe2/python/fused_8bit_rowwise_conversion_ops_test.py
@@ -30,7 +30,7 @@ def floats_to_bytes(floats):
         if isinstance(as_bytes[0], int):
             byte_matrix[i] = list(as_bytes)
         else:
-            byte_matrix[i] = list(map(ord, as_bytes))
+            byte_matrix[i] = [ord(i) for i in as_bytes]
     return byte_matrix
 
 
diff --git a/caffe2/python/operator_test/fused_nbit_rowwise_conversion_ops_test.py b/caffe2/python/operator_test/fused_nbit_rowwise_conversion_ops_test.py
@@ -46,7 +46,7 @@ def int8_to_bytes(int8s):
         if isinstance(as_bytes[0], int):
             byte_matrix[i] = list(as_bytes)
         else:
-            byte_matrix[i] = list(map(ord, as_bytes))
+            byte_matrix[i] = [ord(i) for i in as_bytes]
     return byte_matrix
 
 
diff --git a/caffe2/python/operator_test/torch_integration_test.py b/caffe2/python/operator_test/torch_integration_test.py
@@ -90,7 +90,7 @@ def floats_to_bytes(floats):
         if isinstance(as_bytes[0], int):
             byte_matrix[i] = list(as_bytes)
         else:
-            byte_matrix[i] = list(map(ord, as_bytes))
+            byte_matrix[i] = [ord(i) for i in as_bytes]
     return byte_matrix
 
 
diff --git a/caffe2/python/rnn_cell.py b/caffe2/python/rnn_cell.py
@@ -42,7 +42,7 @@ def _RectifyName(blob_reference_or_name):
 def _RectifyNames(blob_references_or_names):
     if blob_references_or_names is None:
         return None
-    return list(map(_RectifyName, blob_references_or_names))
+    return [_RectifyName(i) for i in blob_references_or_names]
 
 
 class RNNCell(object):
@@ -236,7 +236,7 @@ def get_state_names(self):
         '''
         Returns recurrent state names with self.name scoping applied
         '''
-        return list(map(self.scope, self.get_state_names_override()))
+        return [self.scope(name) for name in self.get_state_names_override()]
 
     def get_state_names_override(self):
         '''
diff --git a/test/jit/test_list_dict.py b/test/jit/test_list_dict.py
@@ -948,11 +948,11 @@ def check_list(fn, li):
             check_list(min_intlist, int_list)
             check_list(max_intlist, int_list)
 
-            bool_li = list(map(lambda x: bool(x), int_list))
+            bool_li = [bool(x) for x in int_list]
             check_list(min_boollist, bool_li)
             check_list(max_boollist, bool_li)
 
-            float_li = list(map(lambda x: float(x), int_list))
+            float_li = [float(x) for x in int_list]
             check_list(min_floatlist, float_li)
             check_list(max_floatlist, float_li)
 
diff --git a/test/onnx/test_pytorch_onnx_onnxruntime.py b/test/onnx/test_pytorch_onnx_onnxruntime.py
@@ -53,7 +53,7 @@ def convert_to_onnx(model, input=None, opset_version=9, example_outputs=None,
 def run_ort(ort_sess, input):
     input_copy = copy.deepcopy(input)
     input, _ = torch.jit._flatten(input_copy)
-    inputs = list(map(to_numpy, input))
+    inputs = [to_numpy(inp) for inp in input]
 
     ort_inputs = dict((ort_sess.get_inputs()[i].name, input) for i, input in enumerate(inputs))
     ort_outs = ort_sess.run(None, ort_inputs)
@@ -62,7 +62,7 @@ def run_ort(ort_sess, input):
 
 def ort_compare_with_pytorch(ort_outs, output, rtol, atol):
     output, _ = torch.jit._flatten(output)
-    outputs = list(map(to_numpy, output))
+    outputs = [to_numpy(outp) for outp in output]
 
     # compare onnxruntime and PyTorch results
     assert len(outputs) == len(ort_outs), "number of outputs differ"
diff --git a/test/onnx/verify.py b/test/onnx/verify.py
@@ -386,8 +386,8 @@ def run(args):
                                               "it had a different set of parameters.  Are you assigning Parameters\n"
                                               "in the forward() of your model definition?")
                     with errs.addErrCtxt(initializer_order_hint):
-                        errs.requireEqual(list(map(lambda x: x.name, proto.graph.initializer)),
-                                          list(map(lambda x: x.name, alt_proto.graph.initializer)),
+                        errs.requireEqual([x.name for x in proto.graph.initializer],
+                                          [x.name for x in alt_proto.graph.initializer],
                                           msg="Parameters list differs")
 
                     # Now check if the embedded parameters are actually the same
diff --git a/test/test_cuda.py b/test/test_cuda.py
@@ -2959,9 +2959,9 @@ def test_reduce_add(self):
         self.assertEqual(result.cpu(), x + y)
 
     def _test_reduce_add_coalesced(self, tensors, buffer_size):
-        dup_tensors = [tensors, list(map(lambda t: t.cuda(1), tensors))]
+        dup_tensors = [tensors, [t.cuda(1) for t in tensors]]
 
-        r_tensors = list(map(comm.reduce_add, zip(*dup_tensors)))
+        r_tensors = [comm.reduce_add(t) for t in zip(*dup_tensors)]
         for r, t in zip(r_tensors, tensors):
             self.assertEqualTypeString(r, t)
             self.assertEqual(r, t * 2)
diff --git a/test/test_jit.py b/test/test_jit.py
@@ -8030,7 +8030,7 @@ def contained_blocks(node):
                 return len(node.findAllNodes("prim::If")) * 2 + len(node.findAllNodes("prim::Loop"))
             for node in ifs + loops:
                 outs = list(node.outputs())
-                out_name = list(map(lambda x: x.debugName(), outs))
+                out_name = [x.debugName() for x in outs]
                 if len(out_name) == 0:
                     continue
                 fc = FileCheck()
diff --git a/test/test_nn.py b/test/test_nn.py
@@ -112,7 +112,7 @@ def _ordered_sequence(self, tensor_type):
     def _padded_sequence(self, tensor_type):
         """Create Tensor of random padded sequences"""
         ordered = self._ordered_sequence(tensor_type)
-        lengths = list(map(len, ordered))
+        lengths = [len(i) for i in ordered]
         padded_tensor = rnn_utils.pad_sequence(ordered)
         return padded_tensor, lengths
 
@@ -11335,7 +11335,7 @@ def _ordered_sequence(self, device, dtype):
     def _padded_sequence(self, device, dtype):
         """Create Tensor of random padded sequences"""
         ordered = self._ordered_sequence(device, dtype)
-        lengths = list(map(len, ordered))
+        lengths = [len(i) for i in ordered]
         padded_tensor = rnn_utils.pad_sequence(ordered)
         return padded_tensor, lengths
 
diff --git a/test/test_optim.py b/test/test_optim.py
@@ -878,7 +878,7 @@ def test_step_lr(self):
         # lr = 0.0005   if epoch >= 9
         epochs = 10
         single_targets = [0.05] * 3 + [0.005] * 3 + [0.0005] * 3 + [0.00005] * 3
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         scheduler = StepLR(self.opt, gamma=0.1, step_size=3)
         self._test(scheduler, targets, epochs)
 
@@ -897,7 +897,7 @@ def test_get_last_lr_multi_step_lr(self):
         # lr = 0.00005   if 9 <= epoch
         epochs = 10
         single_targets = [0.05] * 2 + [0.005] * 3 + [0.0005] * 4 + [0.00005] * 1
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         scheduler = MultiStepLR(self.opt, gamma=0.1, milestones=[2, 5, 9])
         self._test_get_last_lr(scheduler, targets, epochs)
 
@@ -908,7 +908,7 @@ def test_multi_step_lr(self):
         # lr = 0.00005   if epoch >= 9
         epochs = 10
         single_targets = [0.05] * 2 + [0.005] * 3 + [0.0005] * 4 + [0.00005] * 3
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         scheduler = MultiStepLR(self.opt, gamma=0.1, milestones=[2, 5, 9])
         self._test(scheduler, targets, epochs)
 
@@ -919,14 +919,14 @@ def test_multi_step_lr_with_epoch(self):
         # lr = 0.00005   if epoch >= 9
         epochs = 10
         single_targets = [0.05] * 2 + [0.005] * 3 + [0.0005] * 4 + [0.00005] * 3
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         scheduler = MultiStepLR(self.opt, gamma=0.1, milestones=[2, 5, 9])
         self._test_with_epoch(scheduler, targets, epochs)
 
     def test_exp_lr(self):
         epochs = 10
         single_targets = [0.05 * (0.9 ** x) for x in range(epochs)]
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         scheduler = ExponentialLR(self.opt, gamma=0.9)
         self._test(scheduler, targets, epochs)
 
@@ -936,7 +936,7 @@ def test_cos_anneal_lr(self):
         single_targets = [eta_min + (0.05 - eta_min) *
                           (1 + math.cos(math.pi * x / epochs)) / 2
                           for x in range(epochs)]
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         scheduler = CosineAnnealingLR(self.opt, T_max=epochs, eta_min=eta_min)
         self._test(scheduler, targets, epochs)
 
@@ -1058,7 +1058,7 @@ def test_compound_step_and_exp_lr(self):
         single_targets += [0.005 * (0.9 ** x) for x in range(3, 6)]
         single_targets += [0.0005 * (0.9 ** x) for x in range(6, 9)]
         single_targets += [0.00005 * (0.9 ** x) for x in range(9, 12)]
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         schedulers[0] = StepLR(self.opt, gamma=0.1, step_size=3)
         schedulers[1] = ExponentialLR(self.opt, gamma=0.9)
         self._test(schedulers, targets, epochs)
@@ -1070,7 +1070,7 @@ def test_compound_exp_and_multistep_lr(self):
         single_targets += [0.005 * (0.9 ** x) for x in range(2, 5)]
         single_targets += [0.0005 * (0.9 ** x) for x in range(5, 9)]
         single_targets += [0.00005 * (0.9 ** x) for x in range(9, 11)]
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         schedulers[0] = MultiStepLR(self.opt, gamma=0.1, milestones=[2, 5, 9])
         schedulers[1] = ExponentialLR(self.opt, gamma=0.9)
         self._test(schedulers, targets, epochs)
@@ -1082,7 +1082,7 @@ def test_compound_cosanneal_and_step_lr(self):
                           (1 + math.cos(math.pi * x / epochs)) / 2
                           for x in range(epochs)]
         single_targets = [x * 0.1 ** (i // 3) for i, x in enumerate(single_targets)]
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         schedulers = [None] * 2
         schedulers[0] = CosineAnnealingLR(self.opt, T_max=epochs, eta_min=eta_min)
         schedulers[1] = StepLR(self.opt, gamma=0.1, step_size=3)
@@ -1096,7 +1096,7 @@ def test_compound_cosanneal_and_multistep_lr(self):
                           for x in range(epochs)]
         multipliers = [1] * 2 + [0.1] * 3 + [0.01] * 4 + [0.001]
         single_targets = [x * y for x, y in zip(single_targets, multipliers)]
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         schedulers = [None] * 2
         schedulers[0] = CosineAnnealingLR(self.opt, T_max=epochs, eta_min=eta_min)
         schedulers[1] = MultiStepLR(self.opt, gamma=0.1, milestones=[2, 5, 9])
@@ -1110,7 +1110,7 @@ def test_compound_cosanneal_and_exp_lr(self):
                           for x in range(epochs)]
         multipliers = [0.1 ** i for i in range(epochs)]
         single_targets = [x * y for x, y in zip(single_targets, multipliers)]
-        targets = [single_targets, list(map(lambda x: x * epochs, single_targets))]
+        targets = [single_targets, [x * epochs for x in single_targets]]
         schedulers = [None] * 2
         schedulers[0] = CosineAnnealingLR(self.opt, T_max=epochs, eta_min=eta_min)
         schedulers[1] = ExponentialLR(self.opt, gamma=0.1)
@@ -1222,8 +1222,8 @@ def test_cycle_lr_exp_range_mode_one_lr(self):
         diff_lr = max_lr - base_lr
         gamma = 0.9
         xs = [0, 0.25, 0.5, 0.75, 1, 0.75, 0.50, 0.25, 0, 0.25, 0.5, 0.75, 1]
-        lr_target = list(map(lambda x: base_lr + x[1] * diff_lr * gamma**x[0], enumerate(xs)))
-        momentum_target = list(map(lambda x: max_lr - x[1] * diff_lr * gamma**x[0], enumerate(xs)))
+        lr_target = [base_lr + x * diff_lr * gamma**i for i, x in enumerate(xs)]
+        momentum_target = [max_lr - x * diff_lr * gamma**i for i, x in enumerate(xs)]
         lr_targets = [lr_target, lr_target]
         momentum_targets = [momentum_target, momentum_target]
         scheduler = CyclicLR(self.opt, base_lr=base_lr,
@@ -1234,10 +1234,10 @@ def test_cycle_lr_exp_range_mode_one_lr(self):
 
     def test_cycle_lr_triangular_mode(self):
         lr_target_1 = [1, 2, 3, 4, 5, 4, 3, 2, 1, 2, 3]
-        lr_target_2 = list(map(lambda x: x + 1, lr_target_1))
+        lr_target_2 = [x + 1 for x in lr_target_1]
         lr_targets = [lr_target_1, lr_target_2]
         momentum_target_1 = [5, 4, 3, 2, 1, 2, 3, 4, 5, 4, 3]
-        momentum_target_2 = list(map(lambda x: x + 1, momentum_target_1))
+        momentum_target_2 = [x + 1 for x in momentum_target_1]
         momentum_targets = [momentum_target_1, momentum_target_2]
         scheduler = CyclicLR(self.opt, base_lr=[1, 2], max_lr=[5, 6], step_size_up=4,
                              cycle_momentum=True, base_momentum=[1, 2], max_momentum=[5, 6],
@@ -1247,11 +1247,11 @@ def test_cycle_lr_triangular_mode(self):
     def test_cycle_lr_triangular2_mode(self):
         lr_target_1 = [1, 2, 3, 4, 5, 4, 3, 2, 1, 1.5, 2.0, 2.5, 3.0, 2.5, 2.0, 1.5, 1,
                        1.25, 1.50, 1.75, 2.00, 1.75]
-        lr_target_2 = list(map(lambda x: x + 2, lr_target_1))
+        lr_target_2 = [x + 2 for x in lr_target_1]
         lr_targets = [lr_target_1, lr_target_2]
         momentum_target_1 = [5.0, 4.0, 3.0, 2.0, 1.0, 2.0, 3.0, 4.0, 5.0, 4.5, 4.0, 3.5,
                              3.0, 3.5, 4.0, 4.5, 5.0, 4.75, 4.5, 4.25, 4.0, 4.25]
-        momentum_target_2 = list(map(lambda x: x + 2, momentum_target_1))
+        momentum_target_2 = [x + 2 for x in momentum_target_1]
         momentum_targets = [momentum_target_1, momentum_target_2]
         scheduler = CyclicLR(self.opt, base_lr=[1, 3], max_lr=[5, 7], step_size_up=4,
                              cycle_momentum=True, base_momentum=[1, 3], max_momentum=[5, 7],
@@ -1267,11 +1267,11 @@ def test_cycle_lr_exp_range_mode(self):
 
         gamma = 0.9
         xs = [0, 0.25, 0.5, 0.75, 1, 0.75, 0.50, 0.25, 0, 0.25, 0.5, 0.75, 1]
-        lr_target_1 = list(map(lambda x: base_lr_1 + x[1] * diff_lr_1 * gamma**x[0], enumerate(xs)))
-        lr_target_2 = list(map(lambda x: base_lr_2 + x[1] * diff_lr_2 * gamma**x[0], enumerate(xs)))
+        lr_target_1 = [base_lr_1 + x * diff_lr_1 * gamma**i for i, x in enumerate(xs)]
+        lr_target_2 = [base_lr_2 + x * diff_lr_2 * gamma**i for i, x in enumerate(xs)]
         lr_targets = [lr_target_1, lr_target_2]
-        momentum_target_1 = list(map(lambda x: max_lr_1 - x[1] * diff_lr_1 * gamma**x[0], enumerate(xs)))
-        momentum_target_2 = list(map(lambda x: max_lr_2 - x[1] * diff_lr_2 * gamma**x[0], enumerate(xs)))
+        momentum_target_1 = [max_lr_1 - x * diff_lr_1 * gamma**i for i, x in enumerate(xs)]
+        momentum_target_2 = [max_lr_2 - x * diff_lr_2 * gamma**i for i, x in enumerate(xs)]
         momentum_targets = [momentum_target_1, momentum_target_2]
         scheduler = CyclicLR(self.opt, base_lr=[base_lr_1, base_lr_2],
                              max_lr=[max_lr_1, max_lr_2], step_size_up=4,
diff --git a/test/test_torch.py b/test/test_torch.py
@@ -2676,7 +2676,7 @@ def test_permute(self):
             orig = [1, 2, 3, 4, 5, 6, 7]
             perm = torch.randperm(7).tolist()
             x = torch.Tensor(*orig).fill_(0)
-            new = list(map(lambda x: x - 1, x.permute(*perm).size()))
+            new = [i - 1 for i in x.permute(*perm).size()]
             self.assertEqual(perm, new)
             self.assertEqual(x.size(), orig)
 
@@ -8553,7 +8553,7 @@ def consec(size, start=1):
                 idx2_end = idx2_start + random.randrange(1, 10 - idx2_start + 1)
                 idx2_step = random.randrange(1, 8)
                 idx2 = slice(idx2_start, idx2_end, idx2_step)
-                lst_indexed = list(map(lambda l: l[idx2], lst[idx1]))
+                lst_indexed = [l[idx2] for l in lst[idx1]]
                 tensor_indexed = tensor[idx1, idx2]
             else:
                 lst_indexed = lst[idx1]
diff --git a/tools/codegen/gen.py b/tools/codegen/gen.py
@@ -974,7 +974,7 @@ def main() -> None:
         d = pre_grouped_native_functions[f.func.signature()]
         assert f.func.kind() not in d
         d[f.func.kind()] = f
-    grouped_native_functions = list(map(NativeFunctionGroup.from_dict, pre_grouped_native_functions.values()))
+    grouped_native_functions = [NativeFunctionGroup.from_dict(v) for v in pre_grouped_native_functions.values()]
     # NB: At the moment, grouped_native_functions isn't used by anything,
     # this code lives here to help potential future consumers; for a live
     # example see https://github.com/pytorch/pytorch/pull/45277
@@ -1130,9 +1130,9 @@ def computeSchemaRegister() -> Dict[str, object]:
             }
         cpu_fm.write('SchemaRegister.cpp', computeSchemaRegister)
 
-    cpu_fm.write('Declarations.yaml', lambda: format_yaml(list(map(compute_declaration_yaml, native_functions))))
+    cpu_fm.write('Declarations.yaml', lambda: format_yaml([compute_declaration_yaml(f) for f in native_functions]))
     cpu_fm.write('RegistrationDeclarations.h', lambda: {
-        'registration_declarations': list(map(compute_registration_declarations, native_functions)),
+        'registration_declarations': [compute_registration_declarations(f) for f in native_functions],
     })
 
     if options.output_dependencies:
diff --git a/torch/jit/frontend.py b/torch/jit/frontend.py
@@ -394,7 +394,7 @@ def build_Expr(ctx, stmt):
     @staticmethod
     def build_Assign(ctx, stmt):
         rhs = build_expr(ctx, stmt.value)
-        lhs = list(map(lambda x: build_expr(ctx, x), stmt.targets))
+        lhs = [build_expr(ctx, x) for x in stmt.targets]
         return Assign(lhs, rhs)
 
     @staticmethod
diff --git a/torch/nn/parallel/_functions.py b/torch/nn/parallel/_functions.py
@@ -13,7 +13,7 @@ def forward(ctx, target_gpus, *inputs):
         assert all(map(lambda i: i.device.type != 'cpu', inputs)), (
             'Broadcast function not implemented for CPU tensors'
         )
-        target_gpus = list(map(lambda x: _get_device_index(x, True), target_gpus))
+        target_gpus = [_get_device_index(x, True) for x in target_gpus]
         ctx.target_gpus = target_gpus
         if len(inputs) == 0:
             return tuple()
@@ -82,7 +82,7 @@ class Scatter(Function):
 
     @staticmethod
     def forward(ctx, target_gpus, chunk_sizes, dim, input):
-        target_gpus = list(map(lambda x: _get_device_index(x, True), target_gpus))
+        target_gpus = [_get_device_index(x, True) for x in target_gpus]
         ctx.dim = dim
         ctx.input_device = input.get_device() if input.device.type != "cpu" else -1
         streams = None
diff --git a/torch/nn/parallel/data_parallel.py b/torch/nn/parallel/data_parallel.py
@@ -19,7 +19,7 @@ def _check_balance(device_ids):
     has less than 75% of the memory or cores of GPU {}. You can do so by setting
     the device_ids argument to DataParallel, or by setting the CUDA_VISIBLE_DEVICES
     environment variable."""
-    device_ids = list(map(lambda x: _get_device_index(x, True), device_ids))
+    device_ids = [_get_device_index(x, True) for x in device_ids]
     dev_props = _get_devices_properties(device_ids)
 
     def warn_imbalance(get_prop):
@@ -135,7 +135,7 @@ def __init__(self, module, device_ids=None, output_device=None, dim=0):
 
         self.dim = dim
         self.module = module
-        self.device_ids = list(map(lambda x: _get_device_index(x, True), device_ids))
+        self.device_ids = [_get_device_index(x, True) for x in device_ids]
         self.output_device = _get_device_index(output_device, True)
         self.src_device_obj = torch.device(device_type, self.device_ids[0])
 
@@ -200,7 +200,7 @@ def data_parallel(module, inputs, device_ids=None, output_device=None, dim=0, mo
     if output_device is None:
         output_device = device_ids[0]
 
-    device_ids = list(map(lambda x: _get_device_index(x, True), device_ids))
+    device_ids = [_get_device_index(x, True) for x in device_ids]
     output_device = _get_device_index(output_device, True)
     src_device_obj = torch.device(device_type, device_ids[0])
 
diff --git a/torch/nn/parallel/distributed.py b/torch/nn/parallel/distributed.py
diff --git a/torch/nn/parallel/parallel_apply.py b/torch/nn/parallel/parallel_apply.py
diff --git a/torch/nn/parallel/replicate.py b/torch/nn/parallel/replicate.py
diff --git a/torch/nn/parallel/scatter_gather.py b/torch/nn/parallel/scatter_gather.py
diff --git a/torch/optim/lr_scheduler.py b/torch/optim/lr_scheduler.py
diff --git a/torch/overrides.py b/torch/overrides.py
diff --git a/torch/testing/_internal/distributed/distributed_test.py b/torch/testing/_internal/distributed/distributed_test.py
diff --git a/torch/utils/cpp_extension.py b/torch/utils/cpp_extension.py
