[4/N] Remove unused loop variables in tests (#166690)

This PR removes unused loop variables in tests.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/166690
Approved by: https://github.com/justinchuby, https://github.com/mlazos

test/dynamo/test_dicts.py

@@ -341,7 +341,7 @@ class DictTests(torch._dynamo.test_case.TestCase):
 
         def fn(x, d):
             y = 0
-            for idx, (key, value) in enumerate(d.items()):
+            for idx, value in enumerate(d.values()):
                 if idx == 0:
                     y += torch.sin(x * value)
                 else:
@@ -366,7 +366,7 @@ class DictTests(torch._dynamo.test_case.TestCase):
 
         def fn(x, d):
             y = 0
-            for idx, (key, value) in enumerate(d.items()):
+            for idx, value in enumerate(d.values()):
                 if idx == 0:
                     y += torch.sin(x * value)
                 else:
@@ -847,7 +847,7 @@ class DictTests(torch._dynamo.test_case.TestCase):
             d = {"a": 2, "b": 3, "c": 5 * x}
             mp = types.MappingProxyType(d)
             y = torch.sin(x * mp["a"])
-            for k, v in mp.items():  # noqa: PERF102
+            for v in mp.values():
                 y += torch.cos(x * v)
             return mp
 
@@ -864,7 +864,7 @@ class DictTests(torch._dynamo.test_case.TestCase):
         def fn(x):
             mp = types.MappingProxyType(d)
             y = torch.sin(x * mp["a"])
-            for k, v in mp.items():  # noqa: PERF102
+            for v in mp.values():
                 y += torch.cos(x * v)
             d["d"] = 4
             return mp
@@ -885,7 +885,7 @@ class DictTests(torch._dynamo.test_case.TestCase):
 
         def fn(x, mp):
             y = torch.sin(x * mp["a"])
-            for k, v in mp.items():  # noqa: PERF102
+            for v in mp.values():
                 y += torch.cos(x * v)
             if isinstance(mp, types.MappingProxyType):
                 y *= 2

test/dynamo/test_higher_order_ops.py

@@ -2858,7 +2858,7 @@ class GraphModule(torch.nn.Module):
         def fn(x):
             return wrap(lambda x: model(x), x)
 
-        for i in range(2):
+        for _ in range(2):
             # second iteration is key, hooks would have fired during aot trace
             # on first iter
             activations.clear()

test/dynamo/test_hooks.py

@@ -807,7 +807,7 @@ class HooksTests(torch._dynamo.test_case.TestCase):
             def __init__(self) -> None:
                 super().__init__()
                 self.layers = torch.nn.ModuleList()
-                for i in range(10):
+                for _ in range(10):
                     layer = torch.nn.Linear(16, 16)
                     layer.register_forward_pre_hook(lambda _, inp: fw_hook(inp))
                     layer = torch.compile(layer, backend=cnts)

test/dynamo/test_unspec.py

@@ -697,7 +697,7 @@ class UnspecTests(torch._dynamo.test_case.TestCase):
     @torch._dynamo.config.patch(specialize_float=False, capture_scalar_outputs=True)
     def test_unspecialized_float_multiply_precision(self):
         dtypes = [torch.bfloat16, torch.float16, torch.float32, torch.float64]
-        for i, dtype in enumerate(dtypes):
+        for dtype in dtypes:
 
             def fn(x, y):
                 return x * y
@@ -722,7 +722,7 @@ class UnspecTests(torch._dynamo.test_case.TestCase):
             return x + y.item()
 
         dtypes = [torch.bfloat16, torch.float16, torch.float32, torch.float64]
-        for i, dtype in enumerate(dtypes):
+        for dtype in dtypes:
             x = torch.ones(3, 3, dtype=dtype)
             self.assertEqual(f(x), x + x.sum().item())
 

test/functorch/test_aot_joint_with_descriptors.py

@@ -675,7 +675,7 @@ class inner_f(torch.nn.Module):
 
             # Verify buffer handling
             buffer_count = 0
-            for desc, (node, grad_node) in input_grad_nodes.items():
+            for desc, (node, _grad_node) in input_grad_nodes.items():
                 if isinstance(desc, BufferAOTInput):
                     buffer_count += 1
                     self.assertIsNotNone(node)
@@ -764,13 +764,13 @@ class inner_f(torch.nn.Module):
                 self.assertIn(node, named_params.values())
 
             # Check that param_grads contains the same parameter nodes
-            for desc, (param_node, grad_node) in param_grads.items():
+            for desc, (param_node, _grad_node) in param_grads.items():
                 self.assertIn(param_node, param_nodes)
                 self.assertEqual(param_node, named_params[desc.target])
 
             # Check that all_input_grads contains the parameter nodes
             param_count = 0
-            for desc, (input_node, grad_node) in all_input_grads.items():
+            for desc, (input_node, _grad_node) in all_input_grads.items():
                 if isinstance(desc, ParamAOTInput):
                     param_count += 1
                     self.assertIn(input_node, param_nodes)

test/functorch/test_control_flow.py

@@ -3088,9 +3088,7 @@ class GraphModule(torch.nn.Module):
             )
 
             # Compare gradients for each layer
-            for i, (uncompiled_grad, compiled_grad) in enumerate(
-                zip(uncompiled_grads, compiled_grads)
-            ):
+            for uncompiled_grad, compiled_grad in zip(uncompiled_grads, compiled_grads):
                 self.assertEqual(
                     uncompiled_grad,
                     compiled_grad,

test/functorch/test_dims.py

@@ -282,7 +282,7 @@ class TestMin(TestCase):
 
         # python 3.11 adapts bytecode after a number of iterations
         # check that we still match names correctly
-        for i in range(10):
+        for _ in range(10):
             f()
 
     @skipIf(not TEST_CUDA, "no CUDA")

test/inductor/test_aot_inductor.py

@@ -4869,7 +4869,7 @@ class AOTInductorTestsTemplate:
                 return result
 
         inputs = []
-        for i in range(1000):
+        for _ in range(1000):
             inputs.append(torch.ones(8, 8, 8, dtype=torch.float16, device=self.device))
         inputs = tuple(inputs)
         model = Model()

test/inductor/test_compile_subprocess.py

@@ -182,7 +182,7 @@ class TestSubprocess(TestCase):
         @torch.compile(fullgraph=True, backend="inductor")
         def model_add(x, y):
             out = x
-            for i in range(500):
+            for _ in range(500):
                 out = torch.add(out, y)
             return out
 

test/inductor/test_compiled_autograd.py

@@ -405,7 +405,7 @@ main()
                 self.grad_acc_hooks = []
                 self.grad_acc = []
                 self.params = [self.fc1.weight, self.fc2.weight]
-                for i, param in enumerate(self.params):
+                for param in self.params:
 
                     def wrapper(param):
                         param_tmp = param.expand_as(param)
@@ -1558,7 +1558,7 @@ main()
             dtype=input_tensor.dtype, device=DEVICE
         )
 
-        for iteration in range(10):
+        for _ in range(10):
             for param in model_parameters:
                 param.grad = None
             output_tensor = model(
@@ -1599,7 +1599,7 @@ main()
 
         eager_check()
 
-        for i in range(5):
+        for _ in range(5):
             with compiled_autograd._enable(compiler_fn):
                 eager_check()
 

test/inductor/test_cuda_repro.py

@@ -544,7 +544,7 @@ class CudaReproTests(TestCase):
 
         input = torch.randn(10, 10, device="cuda", requires_grad=True)
 
-        for i in range(2):
+        for _ in range(2):
             output_ref = model_ref(input)
             output_res = model_opt(input)
             output_ref.sum().backward()

test/inductor/test_cudagraph_trees.py

@@ -383,7 +383,7 @@ if HAS_CUDA_AND_TRITON:
             foo = get_compile_fn(backend)(foo)
 
             with capture_stderr() as captured_output:
-                for i in range(3):
+                for _ in range(3):
                     torch.compiler.cudagraph_mark_step_begin()
                     inp = torch.rand([4], device="cuda")
 
@@ -415,7 +415,7 @@ if HAS_CUDA_AND_TRITON:
             foo = get_compile_fn(backend)(foo)
 
             with capture_stderr() as captured_output:
-                for i in range(3):
+                for _ in range(3):
                     torch.compiler.cudagraph_mark_step_begin()
                     inp = torch.rand([4], device="cuda")
 
@@ -493,7 +493,7 @@ if HAS_CUDA_AND_TRITON:
                 # Should warn for current_node=None
                 mut(inp())
 
-                for i in range(3):
+                for _ in range(3):
                     torch.compiler.cudagraph_mark_step_begin()
                     tmp = foo(inp())
                     mut(tmp)  # should not warn
@@ -2180,7 +2180,7 @@ if HAS_CUDA_AND_TRITON:
 
             model = torch.nn.Linear(10, 10, bias=False, device="cuda")
             x = torch.randn(10, 10, device="cuda")
-            for i in range(5):
+            for _ in range(5):
                 out = model(x)
                 bwd(out.sum())
                 model.weight.grad = None
@@ -4505,7 +4505,7 @@ if HAS_CUDA_AND_TRITON:
             ]
             for i, compile_fn in enumerate(compile_fns):
                 torch.manual_seed(0)
-                for index in range(3):
+                for _ in range(3):
                     x = torch.randn(4, 4, device=device, requires_grad=True)
                     y = torch.randn(4, 4, device=device, requires_grad=True)
 

test/inductor/test_custom_op_autotune.py

@@ -485,7 +485,7 @@ class TestCustomOpAutoTune(TestCase):
             (3, 16),  # einsum, chunk_size ignored
         ]
 
-        for i, (scale_mode, chunk_size) in enumerate(configs):
+        for scale_mode, chunk_size in configs:
             result = multi_param_scaling(
                 test_x, test_factor, scale_mode=scale_mode, chunk_size=chunk_size
             )

test/inductor/test_inductor_freezing.py

@@ -970,9 +970,7 @@ class OptimizeForInferenceTemplate(TestCase):
 
         self.assertEqual(len(actual_outputs), len(expected_outputs))
         self.assertEqual(2, len(actual_outputs))
-        for i, actual, expected in zip(
-            itertools.count(), actual_outputs, expected_outputs
-        ):
+        for actual, expected in zip(actual_outputs, expected_outputs):
             self.assertEqual(expected, actual)
 
         if self.device == "cpu":

test/inductor/test_max_autotune.py

@@ -2095,7 +2095,7 @@ class TestMaxAutotune(TestCase):
 
         # Test loop.
         def test_func2(x):
-            for i in range(10):
+            for _ in range(10):
                 x = torch.matmul(x, x)
             return x
 

test/inductor/test_memory.py

@@ -343,7 +343,7 @@ class TestOperatorReorderForPeakMemory(TestCase):
     def test_fusion_acc_large_reads(self):
         def f(x, y, z):
             res = torch.zeros_like(x[0])
-            for i in range(4):
+            for _ in range(4):
                 temp = torch.matmul(x, y) + z
                 res = res + temp
             return res

test/inductor/test_ordered_set.py

@@ -539,7 +539,7 @@ class TestSet(TestJointOps, TestCase):
         # s.discard(self.thetype(self.word))
 
     def test_pop(self):
-        for i in range(len(self.s)):
+        for _ in range(len(self.s)):
             elem = self.s.pop()
             self.assertNotIn(elem, self.s)
         self.assertRaises(KeyError, self.s.pop)
@@ -990,7 +990,7 @@ class TestExceptionPropagation(TestCase):
     def test_changingSizeWhileIterating(self):
         s = OrderedSet([1, 2, 3])
         try:
-            for i in s:
+            for _ in s:
                 s.update([4])  # noqa: B909
         except RuntimeError:
             pass

test/inductor/test_profiler.py

@@ -245,7 +245,7 @@ class DynamoProfilerTests(torch._inductor.test_case.TestCase):
                 skip_first=3, wait=1, warmup=1, active=2, repeat=1
             ),
         ) as prof:
-            for idx in range(10):
+            for _ in range(10):
                 fn(*inputs)
                 prof.step()
 

test/inductor/test_torchinductor.py

@@ -2090,7 +2090,7 @@ class CommonTemplate:
         from torch._inductor.runtime.triton_heuristics import triton_config_reduction
 
         size_hints = {"x": 67108864, "r0_": 8192}
-        for i in range(4):
+        for _ in range(4):
             size_hints["x"] = next_power_of_2(size_hints["x"])
             triton_config_reduction(size_hints, 1, 2048, 1, 8)
 
@@ -5033,13 +5033,13 @@ class CommonTemplate:
 
         def run_weights_sharing_model(m, inp):
             with torch.no_grad():
-                for i in range(num_run):
+                for _ in range(num_run):
                     y = m(inp)
 
         numb_instance = 2
         threads = []
         compiled_m = torch.compile(model)
-        for i in range(1, numb_instance + 1):
+        for _ in range(1, numb_instance + 1):
             thread = threading.Thread(
                 target=run_weights_sharing_model, args=(compiled_m, inp)
             )

test/inductor/test_triton_kernels.py

@@ -497,7 +497,7 @@ def forward(self, x_1, output_1):
             x: torch.Tensor,
             y: torch.Tensor,
         ):
-            for i in range(4):
+            for _ in range(4):
                 x = add_in_loop(x, y)
             return x
 
@@ -2971,7 +2971,7 @@ class MutationTests(torch._inductor.test_case.TestCase):
             x = tl.load(in_ptr0 + offsets, mask=mask)
             y = tl.load(in_ptr1 + offsets, mask=mask)
             output = tl.zeros((n_elements,), dtype=tl.float32)
-            for i in range(4):
+            for _ in range(4):
                 output += x + y
             tl.store(out_ptr + offsets, output, mask=mask)
 
@@ -3041,8 +3041,8 @@ class MutationTests(torch._inductor.test_case.TestCase):
             x = tl.load(in_ptr0 + offsets, mask=mask)
             y = tl.load(in_ptr1 + offsets, mask=mask)
             output = tl.zeros((n_elements,), dtype=tl.float32)
-            for i in range(2):
-                for j in range(2):
+            for _ in range(2):
+                for _ in range(2):
                     output += x + y
             tl.store(out_ptr + offsets, output, mask=mask)
 
@@ -3078,8 +3078,8 @@ class MutationTests(torch._inductor.test_case.TestCase):
             y = tl.load(in_ptr1 + offsets, mask=mask)
             output1 = tl.zeros((n_elements,), dtype=tl.float32)
             output2 = tl.zeros((n_elements,), dtype=tl.float32)
-            for i in range(2):
-                for j in range(2):
+            for _ in range(2):
+                for _ in range(2):
                     output1 += y
                     output2 += x
             output = output1 + output2

test/nn/test_module_hooks.py

@@ -873,7 +873,7 @@ class TestStateDictHooks(TestCase):
         )
 
         def linear_state_dict_post_hook(module, state_dict, prefix, local_metadata):
-            for name, param in module.named_parameters(recurse=False):
+            for name, _param in module.named_parameters(recurse=False):
                 state_dict[prefix + name] = torch.nn.Parameter(
                     state_dict[prefix + name]
                 )

test/onnx/test_pytorch_onnx_onnxruntime.py

@@ -6106,7 +6106,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
         class NestedLoopsModel(torch.jit.ScriptModule):
             @torch.jit.script_method
             def forward(self, x):
-                for i in range(5):
+                for _ in range(5):
                     a = 0
                     while a < 4:
                         a += 1
@@ -6145,7 +6145,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
         class LoopModel(torch.nn.Module):
             def forward(self, x):
                 res = torch.zeros_like(x[0])
-                for i in range(x.size(0)):
+                for _ in range(x.size(0)):
                     res += x[0].transpose(0, 1)
                 return res
 
@@ -6780,7 +6780,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                 a = torch.ones(
                     12,
                 )
-                for i in range(10):
+                for _ in range(10):
                     a.add_(
                         torch.ones(
                             12,
@@ -6809,7 +6809,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                 b_ref = b  # not used in loop, should not be altered.
                 for i in range(10):
                     if i == 3:
-                        for j in range(5):
+                        for _ in range(5):
                             a += _bias
                             _bias.add_(
                                 torch.ones(
@@ -6854,7 +6854,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                 )
                 for i in range(10):
                     if i == 3:
-                        for j in range(5):
+                        for _ in range(5):
                             self._bias += torch.arange(
                                 12,
                             )
@@ -6881,7 +6881,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                 )
                 for i in range(10):
                     if i == 3:
-                        for j in range(5):
+                        for _ in range(5):
                             self._bias.copy_(
                                 torch.arange(
                                     12,
@@ -8567,7 +8567,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
         class SequanceLoopModel(torch.nn.Module):
             def forward(self, x):
                 outputs = []
-                for i in range(3):
+                for _ in range(3):
                     outputs += [x]
                 return torch.stack(outputs).transpose(0, 1)
 
@@ -9768,9 +9768,9 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                 a = (input1, input2)
                 b = a
                 c = (input1, input2, input3)
-                for i in range(5):
+                for _ in range(5):
                     d = a[0]
-                    for j in range(2):
+                    for _ in range(2):
                         e, f = a
                         a = (d, f)
                         f = c[2]
@@ -9794,7 +9794,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
         class TupleModule(torch.nn.Module):
             def forward(self, input1: Tensor, input2: Tensor) -> tuple[Tensor, Tensor]:
                 a = (input1, input2)
-                for x in range(5):
+                for _ in range(5):
                     c, d = a
                     a = (c, d)
                 return a
@@ -9812,7 +9812,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
             ) -> tuple[tuple[Tensor, Tensor], tuple[Tensor, Tensor]]:
                 a = input1
                 b = input2
-                for x in range(5):
+                for _ in range(5):
                     c, d = a
                     e, f = b
                     if c.shape[0] == e.shape[0]:
@@ -11418,7 +11418,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                 self.conv.weight = torch.arange(10)
                 for i in range(10):
                     if i == 3:
-                        for j in range(10):
+                        for _ in range(10):
                             w = self.conv.weight
                             self.conv.weight = torch.arange(10) + w
 
@@ -11480,7 +11480,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
             def set_cell_anchors(self, anchors):
                 self.conv.weight = torch.randn(3, 10)
                 for i in range(self.conv.weight.size(0)):
-                    for j in range(10):
+                    for _ in range(10):
                         self.conv.bias = torch.randn(3, 10, 3)
                         self.conv.weight = anchors * i
                         self.boxes.append(torch.ones(3, 3))
@@ -12452,7 +12452,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                 self.loop_count = loop_count
 
             def forward(self, x):
-                for i in range(self.loop_count):
+                for _ in range(self.loop_count):
                     x.index_add_(self.dim, self.index, self.updates)
                 return x
 

test/optim/test_lrscheduler.py

@@ -192,7 +192,7 @@ class TestLRScheduler(TestCase):
 
     def test_old_pattern_warning_resuming(self):
         epochs = 35
-        for i, group in enumerate(self.opt.param_groups):
+        for group in self.opt.param_groups:
             group["initial_lr"] = 0.01
 
         with warnings.catch_warnings(record=True) as ws:
@@ -209,7 +209,7 @@ class TestLRScheduler(TestCase):
 
     def test_old_pattern_warning_resuming_with_arg(self):
         epochs = 35
-        for i, group in enumerate(self.opt.param_groups):
+        for group in self.opt.param_groups:
             group["initial_lr"] = 0.01
 
         with warnings.catch_warnings(record=True) as ws:
@@ -226,7 +226,7 @@ class TestLRScheduler(TestCase):
 
     def test_old_pattern_warning_with_overridden_optim_step(self):
         epochs = 35
-        for i, group in enumerate(self.opt.param_groups):
+        for group in self.opt.param_groups:
             group["initial_lr"] = 0.01
 
         with warnings.catch_warnings(record=True) as ws:
@@ -299,7 +299,7 @@ class TestLRScheduler(TestCase):
         self.opt.step = types.MethodType(new_step, self.opt)
 
         def new_pattern():
-            for e in range(epochs):
+            for _ in range(epochs):
                 self.opt.step()
                 scheduler.step()
 
@@ -2617,7 +2617,7 @@ class TestLRScheduler(TestCase):
         sch = SWALR(opt, swa_lr=swa_lr)
         ori_param_groups = copy.deepcopy(opt.param_groups)
 
-        for i in range(2):
+        for _ in range(2):
             lr.multiply_(0.5)
             swa_lr.multiply_(0.5)
             opt.step()

test/quantization/core/test_workflow_ops.py

@@ -344,7 +344,7 @@ class TestFakeQuantizeOps(TestCase):
         maxi = 255
         mini = 0
 
-        for i in range(20):
+        for _ in range(20):
             X1 = torch.randn(5, 5).to(torch.float16)
             Y1 = torch.fake_quantize_per_tensor_affine(X1, scale, zero, mini, maxi)
             Y1r = _fake_quantize_per_tensor_affine_reference(X1, scale, zero, mini, maxi)
@@ -770,7 +770,7 @@ class TestFakeQuantizeOps(TestCase):
         mini = 0
         maxi = 255
 
-        for i in range(20):
+        for _ in range(20):
             X1 = torch.randn(4, 5).to(torch.float16)
             Y1 = torch.fake_quantize_per_channel_affine(X1, scale, zero, axis, mini, maxi)
             Y1r = _fake_quantize_per_channel_affine_reference(X1, scale, zero, axis, mini, maxi)
@@ -1028,7 +1028,7 @@ class TestFakeQuantizeOps(TestCase):
             zero_types = [torch.int]
         devices = [torch.device('cpu'), torch.device('cuda')] if torch.cuda.is_available() else [torch.device('cpu')]
         axis = 1
-        for i in range(20):
+        for _ in range(20):
             for torch_type, float_type, device, zero_type in itertools.product(torch_types, float_types, devices, zero_types):
                 X = torch.randn(3, 3, device=device).to(float_type)
                 scales = (10 * torch.randn(3, device=device)).abs()

test/quantization/fx/test_quantize_fx.py

@@ -4672,7 +4672,7 @@ class TestQuantizeFx(QuantizationTestCase):
             m = prepare(m, {"": qconfig}, example_inputs=example_inputs)
             # check that there is a duplicated observer instance
             actpp_module_count = 0
-            for name, module in m.named_modules(remove_duplicate=False):
+            for module in m.modules(remove_duplicate=False):
                 if isinstance(module, actpp_module_class):
                     actpp_module_count += 1
             self.assertEqual(actpp_module_count, 2)

test/quantization/jit/test_quantize_jit.py

@@ -331,7 +331,7 @@ class TestQuantizeJitPasses(QuantizationTestCase):
             def __init__(self, dim, num_blocks, enable_bias, enable_affine):
                 super().__init__()
                 layers = []
-                for i in range(num_blocks):
+                for _ in range(num_blocks):
                     layers.append(conv_module[dim](20, 20, 5, 1, bias=enable_bias))
                     bn_obj = bn_module[dim](num_features=20, affine=enable_affine)
                     if enable_affine:

test/test_datapipe.py

@@ -658,7 +658,7 @@ class TestDataFramesPipes(TestCase):
         ]
 
         actual_i = []
-        for i, j in df_numbers:
+        for i, _ in df_numbers:
             actual_i.append(i)
         self.assertEqual(expected_i, actual_i)
 
@@ -2632,7 +2632,7 @@ class TestTyping(TestCase):
                 self.dp = dp
 
             def __iter__(self) -> Iterator[int]:
-                for a, b in self.dp:
+                for a, _ in self.dp:
                     yield a
 
         # Non-DataPipe input with DataPipe hint

test/test_fx.py

@@ -2365,7 +2365,7 @@ class TestFX(JitTestCase):
 
         g = torch.fx.Graph()
         x = g.placeholder("x")
-        for i in range(depth):
+        for _ in range(depth):
             x = g.call_function(torch.relu, (x,))
         g.output(x)
 

test/test_serialization.py

@@ -295,7 +295,7 @@ class SerializationMixin:
             5,
             6
         ]
-        for i in range(100):
+        for _ in range(100):
             data.append(0)
         t = torch.tensor(data, dtype=torch.uint8)
 

test/test_spectral_ops.py

@@ -1315,7 +1315,7 @@ class TestFFT(TestCase):
             istft_kwargs = stft_kwargs.copy()
             del istft_kwargs['pad_mode']
             for sizes in data_sizes:
-                for i in range(num_trials):
+                for _ in range(num_trials):
                     original = torch.randn(*sizes, dtype=dtype, device=device)
                     stft = torch.stft(original, return_complex=True, **stft_kwargs)
                     inversed = torch.istft(stft, length=original.size(1), **istft_kwargs)
@@ -1386,7 +1386,7 @@ class TestFFT(TestCase):
             del stft_kwargs['size']
             istft_kwargs = stft_kwargs.copy()
             del istft_kwargs['pad_mode']
-            for i in range(num_trials):
+            for _ in range(num_trials):
                 original = torch.randn(*sizes, dtype=dtype, device=device)
                 stft = torch.stft(original, return_complex=True, **stft_kwargs)
                 with self.assertWarnsOnceRegex(UserWarning, "The length of signal is shorter than the length parameter."):
@@ -1501,7 +1501,7 @@ class TestFFT(TestCase):
         complex_dtype = corresponding_complex_dtype(dtype)
 
         def _test(data_size, kwargs):
-            for i in range(num_trials):
+            for _ in range(num_trials):
                 tensor1 = torch.randn(data_size, device=device, dtype=complex_dtype)
                 tensor2 = torch.randn(data_size, device=device, dtype=complex_dtype)
                 a, b = torch.rand(2, dtype=dtype, device=device)

test/test_torch.py

@@ -138,7 +138,7 @@ class TestTorchDeviceType(TestCase):
     # TODO: move all tensor creation to common ops
     def _rand_shape(self, dim, min_size, max_size):
         shape = []
-        for i in range(dim):
+        for _ in range(dim):
             shape.append(random.randint(min_size, max_size))
         return tuple(shape)
 
@@ -172,7 +172,7 @@ class TestTorchDeviceType(TestCase):
 
         element_size = torch._utils._element_size(dtype)
 
-        for i in range(10):
+        for _ in range(10):
             bytes_list = [rand_byte() for _ in range(element_size)]
             scalar = bytes_to_scalar(bytes_list, dtype, device)
             self.assertEqual(scalar.storage().untyped().tolist(), bytes_list)
@@ -2263,7 +2263,7 @@ class TestTorchDeviceType(TestCase):
             if num_observations > 0:
                 fweights = torch.randint(1, 10, (num_observations,), device=device)
                 aweights = make_tensor((num_observations,), dtype=torch.float, device=device, low=1)
-                for correction, fw, aw in product([0, 1, 2], [None, fweights], [None, aweights]):
+                for correction, _fw, _aw in product([0, 1, 2], [None, fweights], [None, aweights]):
                     check(x, correction, fweights, aweights)
 
     @skipIfNoSciPy
@@ -5151,7 +5151,7 @@ class TestTorchDeviceType(TestCase):
         prob_dist = torch.rand(10000, 1000, device=device, dtype=dtype)
         n_sample = 1
 
-        for i in range(trials):
+        for _ in range(trials):
             gen.manual_seed(seed)
             samples_1 = torch.multinomial(prob_dist, n_sample, True, generator=gen)
 
@@ -5229,7 +5229,7 @@ class TestTorchDeviceType(TestCase):
         # TODO copy _like constructors to stride permutation instead of just layout
         if not TEST_WITH_TORCHINDUCTOR:
             x = torch.randn((3, 4, 5, 6, 7, 8, 9), device=device)
-            for i in range(10):
+            for _ in range(10):
                 permutation = list(range(len(x.shape)))
                 random.shuffle(permutation)
                 x = x.permute(permutation)