diff --git a/.flake8 b/.flake8
index dfc371b386a4..5da4f8e575ea 100644
--- a/.flake8
+++ b/.flake8
@@ -14,7 +14,7 @@ ignore =
     # to line this up with executable bit
     EXE001,
     # these ignores are from flake8-bugbear; please fix!
-    B007,B008,B017,B019,B020,B023,B026,B028,B903,B904,B905,B906,B907
+    B007,B008,B017,B019,B023,B026,B028,B903,B904,B905,B906,B907
     # these ignores are from flake8-comprehensions; please fix!
     C407,
     # these ignores are from flake8-logging-format; please fix!
diff --git a/caffe2/python/net_builder_test.py b/caffe2/python/net_builder_test.py
index 5320c2b04588..ed0d0fb01d04 100644
--- a/caffe2/python/net_builder_test.py
+++ b/caffe2/python/net_builder_test.py
@@ -90,7 +90,7 @@ class TestNetBuilder(unittest.TestCase):
         plan.AddStep(to_execution_step(nb))
         ws = workspace.C.Workspace()
         ws.run(plan)
-        expected = [
+        expected_results = [
             (y, 5),
             (z, False),
             (w, True),
@@ -99,7 +99,7 @@ class TestNetBuilder(unittest.TestCase):
             (p, 2),
             (q, 3),
         ]
-        for b, expected in expected:
+        for b, expected in expected_results:
             actual = ws.blobs[str(b)].fetch()
             self.assertEqual(actual, expected)
 
diff --git a/caffe2/python/schema.py b/caffe2/python/schema.py
index 6e401bb89532..924afed41bd5 100644
--- a/caffe2/python/schema.py
+++ b/caffe2/python/schema.py
@@ -1081,7 +1081,7 @@ def from_column_list(
         'col_names and col_blobs must have the same length.'
     )
     root = _SchemaNode('root', 'Struct')
-    for col_name, col_type, col_blob, col_metadata in zip(
+    for col_name, col_type, col_blob, col_md in zip(
         col_names, col_types, col_blobs, col_metadata
     ):
         columns = col_name.split(FIELD_SEPARATOR)
@@ -1095,7 +1095,7 @@ def from_column_list(
                 field = Scalar(
                     dtype=col_type,
                     blob=col_blob,
-                    metadata=col_metadata
+                    metadata=col_md
                 )
             next = current.add_child(name, type_str)
             if field is not None:
diff --git a/pyproject.toml b/pyproject.toml
index d0ad34eb3790..77663a9d68d8 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -28,7 +28,7 @@ ignore = [
     # these ignores are from flake8-bugbear; please fix!
     "B007", "B008", "B017",
     "B018", # Useless expression
-    "B019", "B020",
+    "B019",
     "B023", "B026",
     "B028", # No explicit `stacklevel` keyword argument found
     "B904",
diff --git a/scripts/release_notes/test_release_notes.py b/scripts/release_notes/test_release_notes.py
index 97df4561157a..adef14e20f97 100644
--- a/scripts/release_notes/test_release_notes.py
+++ b/scripts/release_notes/test_release_notes.py
@@ -34,8 +34,8 @@ class TestCommitList(unittest.TestCase):
             expected.write_to_disk()
 
             commit_list = CommitList.from_existing(commit_list_path)
-            for commit, expected in zip(commit_list.commits, expected.commits):
-                self.assertEqual(commit, expected)
+            for commit, expected_commit in zip(commit_list.commits, expected.commits):
+                self.assertEqual(commit, expected_commit)
 
     def test_update_to(self):
         with tempfile.TemporaryDirectory() as tempdir:
diff --git a/test/distributed/elastic/agent/server/test/local_elastic_agent_test.py b/test/distributed/elastic/agent/server/test/local_elastic_agent_test.py
index 27ea0cc75b1e..59c20e46b232 100644
--- a/test/distributed/elastic/agent/server/test/local_elastic_agent_test.py
+++ b/test/distributed/elastic/agent/server/test/local_elastic_agent_test.py
@@ -880,8 +880,8 @@ class LocalElasticAgentTest(unittest.TestCase):
 
         # global world size == sum of all the role world sizes
         expected_world_size = sum(expected_role_world_sizes.values())
-        for role, run_results in run_results.items():
-            for result in run_results:
+        for role, results in run_results.items():
+            for result in results:
                 res = result.return_values
                 for role_info in res.values():
                     rank = role_info.rank
diff --git a/test/distributed/fsdp/test_fsdp_dtensor_state_dict.py b/test/distributed/fsdp/test_fsdp_dtensor_state_dict.py
index a47fbe12dae4..5e7d84bc790e 100644
--- a/test/distributed/fsdp/test_fsdp_dtensor_state_dict.py
+++ b/test/distributed/fsdp/test_fsdp_dtensor_state_dict.py
@@ -143,11 +143,11 @@ class TestFSDPWithDeviceMeshAndDTensor(DTensorTestBase):
         sharded_tensor_osd = FSDP.optim_state_dict(ref_model, ref_optim)
 
         # Check dtensor and sharded_tensor model state dict values are identical
-        for dtensor_sd, sharded_tensor_sd in zip(
+        for dtensor_sd_item, sharded_tensor_sd_item in zip(
             dtensor_sd.items(), sharded_tensor_sd.items()
         ):
-            k1, v1 = dtensor_sd
-            k2, v2 = sharded_tensor_sd
+            k1, v1 = dtensor_sd_item
+            k2, v2 = sharded_tensor_sd_item
             self.assertEqual(k1, k2)
 
             # if the ShardedTensor is an empty shard,
@@ -227,15 +227,15 @@ class TestFSDPWithDeviceMeshAndDTensor(DTensorTestBase):
         new_optim_state_dict = FSDP.optim_state_dict(model, optim)
 
         # Check whether new_optim_state_dict is the same as ref_optim_state_dict.
-        for new_optim_state_dict, ref_optim_state_dict in zip(
+        for new_optim_state_dict_item, ref_optim_state_dict_item in zip(
             new_optim_state_dict["state"].items(),
             ref_optim_state_dict["state"].items(),
         ):
             # check FQN are the same
-            self.assertEqual(new_optim_state_dict[0], ref_optim_state_dict[0])
+            self.assertEqual(new_optim_state_dict_item[0], ref_optim_state_dict_item[0])
             for new_optim_hyper_param, ref_optim_hyper_param in zip(
-                new_optim_state_dict[1].items(),
-                ref_optim_state_dict[1].items(),
+                new_optim_state_dict_item[1].items(),
+                ref_optim_state_dict_item[1].items(),
             ):
                 k1, v1 = new_optim_hyper_param
                 k2, v2 = ref_optim_hyper_param
diff --git a/test/distributed/fsdp/test_hsdp_dtensor_state_dict.py b/test/distributed/fsdp/test_hsdp_dtensor_state_dict.py
index bf461f133161..23d97d6598bf 100644
--- a/test/distributed/fsdp/test_hsdp_dtensor_state_dict.py
+++ b/test/distributed/fsdp/test_hsdp_dtensor_state_dict.py
@@ -154,11 +154,11 @@ class TestHSDPWithDeviceMeshAndDTensor(DTensorTestBase):
         sharded_tensor_osd = FSDP.optim_state_dict(ref_model, ref_optim)
 
         # Check dtensor and sharded_tensor model state dict values are identical
-        for dtensor_sd, sharded_tensor_sd in zip(
+        for dtensor_sd_item, sharded_tensor_sd_item in zip(
             dtensor_sd.items(), sharded_tensor_sd.items()
         ):
-            k1, v1 = dtensor_sd
-            k2, v2 = sharded_tensor_sd
+            k1, v1 = dtensor_sd_item
+            k2, v2 = sharded_tensor_sd_item
             self.assertEqual(k1, k2)
 
             self.assertEqual(type(v1), DTensor)
@@ -225,15 +225,15 @@ class TestHSDPWithDeviceMeshAndDTensor(DTensorTestBase):
         new_optim_state_dict = FSDP.optim_state_dict(model, optim)
 
         # Check whether new_optim_state_dict is the same as ref_optim_state_dict.
-        for new_optim_state_dict, ref_optim_state_dict in zip(
+        for new_optim_state_dict_item, ref_optim_state_dict_item in zip(
             new_optim_state_dict["state"].items(),
             ref_optim_state_dict["state"].items(),
         ):
             # check FQN are the same
-            self.assertEqual(new_optim_state_dict[0], ref_optim_state_dict[0])
+            self.assertEqual(new_optim_state_dict_item[0], ref_optim_state_dict_item[0])
             for new_optim_hyper_param, ref_optim_hyper_param in zip(
-                new_optim_state_dict[1].items(),
-                ref_optim_state_dict[1].items(),
+                new_optim_state_dict_item[1].items(),
+                ref_optim_state_dict_item[1].items(),
             ):
                 k1, v1 = new_optim_hyper_param
                 k2, v2 = ref_optim_hyper_param
diff --git a/test/functorch/test_vmap.py b/test/functorch/test_vmap.py
index 5b1f6c0e7634..95c0a896e484 100644
--- a/test/functorch/test_vmap.py
+++ b/test/functorch/test_vmap.py
@@ -3423,9 +3423,9 @@ class TestVmapOperatorsOpInfo(TestCase):
                     sample_input = error_input.sample_input
                     args = (sample_input.input,) + tuple(sample_input.args)
                     kwargs = sample_input.kwargs
-                    for args, in_dims, _ in generate_vmap_inputs(args, {}):
+                    for batched_args, in_dims, _ in generate_vmap_inputs(args, {}):
                         with self.assertRaises(Exception):
-                            vmap(op, in_dims)(*args, **kwargs)
+                            vmap(op, in_dims)(*batched_args, **kwargs)
 
             # Sample inputs check
             sample_inputs_op = {
@@ -3455,16 +3455,16 @@ class TestVmapOperatorsOpInfo(TestCase):
                     continue
                 kwargs = sample_input.kwargs
                 is_batch_norm_and_training = is_batch_norm_training(op.name, kwargs)
-                for args, in_dims, _ in generate_vmap_inputs(
+                for batched_args, in_dims, _ in generate_vmap_inputs(
                         args, {}, is_batch_norm_and_training=is_batch_norm_and_training):
                     for func in aliases:
-                        self.vmap_outplace_test(func, args, kwargs, in_dims, check_shape_only, postprocess_fn)
+                        self.vmap_outplace_test(func, batched_args, kwargs, in_dims, check_shape_only, postprocess_fn)
                     if op.name in skip_inplace:
                         continue
                     if not is_valid_inplace_sample_input(sample_input, op, op.inplace_variant):
                         continue
                     for func in inplace_aliases:
-                        self.vmap_inplace_test(func, args, kwargs, in_dims, postprocess_fn)
+                        self.vmap_inplace_test(func, batched_args, kwargs, in_dims, postprocess_fn)
 
         if check_has_batch_rule:
             check_vmap_fallback(self, test, op)
@@ -4195,11 +4195,11 @@ class TestVmapOperatorsOpInfo(TestCase):
         gout = torch.randn(2, 2, device=device)
         args = (leaf, gout)
 
-        for args, in_dims, _, in generate_vmap_inputs(args, {}):
+        for batched_args, in_dims, _, in generate_vmap_inputs(args, {}):
             if in_dims[1] is None:
                 # triggers some composite compliance problem
                 continue
-            self.vmap_outplace_test(push_vjp, args, {}, in_dims)
+            self.vmap_outplace_test(push_vjp, batched_args, {}, in_dims)
 
     def test_advanced_indexing(self, device):
         def test(f, args):
diff --git a/test/jit/test_freezing.py b/test/jit/test_freezing.py
index b207860dd0de..d5d25082bdca 100644
--- a/test/jit/test_freezing.py
+++ b/test/jit/test_freezing.py
@@ -2691,9 +2691,9 @@ class TestFrozenOptimizations(JitTestCase):
         with set_default_dtype(torch.float):
             conv_bias = [True, False]
             conv_ops = [nn.Conv2d, nn.Conv3d]
-            add_z = [True, False]
+            use_add_z = [True, False]
             use_tracing = [True, False]
-            for use_bias, conv, add_z, tracing in product(conv_bias, conv_ops, add_z, use_tracing):
+            for use_bias, conv, add_z, tracing in product(conv_bias, conv_ops, use_add_z, use_tracing):
                 class Net(nn.Module):
                     def __init__(self, in_channels, out_channels, **kwargs):
                         super().__init__()
diff --git a/test/nn/test_embedding.py b/test/nn/test_embedding.py
index d9ec6753f0ee..5d876a155699 100644
--- a/test/nn/test_embedding.py
+++ b/test/nn/test_embedding.py
@@ -944,10 +944,10 @@ class TestEmbeddingNNDeviceType(NNTestCase):
                                  atol=dtype2prec_DONTUSE[dtypes[2]], rtol=0)
 
         trainable_scale = (True, False)
-        include_last_offset = (True, False)
+        include_last_offset_list = (True, False)
         modes = (('sum', False), ('sum', True), ('max', False), ('mean', False))
         for (mode, has_weight), trainable, include_last_offset in itertools.product(
-            modes, trainable_scale, include_last_offset
+            modes, trainable_scale, include_last_offset_list
         ):
             test_per_sample_weights_new_offsets(
                 mode, trainable, include_last_offset, has_weight
diff --git a/test/quantization/core/test_workflow_ops.py b/test/quantization/core/test_workflow_ops.py
index a3528098b256..2e22d49af4ca 100644
--- a/test/quantization/core/test_workflow_ops.py
+++ b/test/quantization/core/test_workflow_ops.py
@@ -364,8 +364,8 @@ class TestFakeQuantizeOps(TestCase):
     def _test_backward_per_tensor_cachemask_impl(self, device):
         float_types = (torch.float32, torch.float16, torch.float64)
         torch_types = (torch.qint8, torch.quint8)
-        tensor_qparam = (True, False)
-        for float_type, torch_type, tensor_qparam in itertools.product(float_types, torch_types, tensor_qparam):
+        tensor_qparams = (True, False)
+        for float_type, torch_type, tensor_qparam in itertools.product(float_types, torch_types, tensor_qparams):
             X = torch.randn(4, 8).to(device).to(float_type)
             X.requires_grad_()
             # pick the scale + zp so that some values get clipped
diff --git a/test/test_linalg.py b/test/test_linalg.py
index c7b816ee5c1e..86a5aece3ff8 100644
--- a/test/test_linalg.py
+++ b/test/test_linalg.py
@@ -2413,7 +2413,7 @@ class TestLinalg(TestCase):
                 self.assertEqual(v.mT.matmul(V).det().abs(), torch.ones(batches, device=device, dtype=dtype))
 
         all_batches = [(), (1,), (3,), (2, 3)]
-        for actual_rank, size, all_batches in [
+        for actual_rank, size, all_batches in [  # noqa: B020
                 (2, (17, 4), all_batches),
                 (4, (17, 4), all_batches),
                 (4, (17, 17), all_batches),
@@ -7415,7 +7415,7 @@ scipy_lobpcg  | {eq_err_scipy:10.2e}  | {eq_err_general_scipy:10.2e}  | {iters2:
                 self.assertEqual(s[..., :actual_rank], S[..., :actual_rank])
 
         all_batches = [(), (1,), (3,), (2, 3)]
-        for actual_rank, size, all_batches in [
+        for actual_rank, size, all_batches in [  # noqa: B020
                 (2, (17, 4), all_batches),
                 (2, (100, 4), all_batches),
                 (6, (100, 40), all_batches),
diff --git a/test/test_nn.py b/test/test_nn.py
index d12469ee91fd..2815ddcf2d4d 100644
--- a/test/test_nn.py
+++ b/test/test_nn.py
@@ -11760,7 +11760,7 @@ class TestNNDeviceType(NNTestCase):
         reductions = ['none', 'sum', 'mean']
         label_smoothings = [0.05, 0.15]
 
-        weight = torch.tensor([0.3, 0.6], device=device)
+        wgt = torch.tensor([0.3, 0.6], device=device)
         inp1 = torch.tensor([[0.3, 0.4], [1, 2]], device=device)
         inp2 = torch.tensor([[0.3, 0.6], [1, 2]], device=device)
 
@@ -11768,7 +11768,7 @@ class TestNNDeviceType(NNTestCase):
         targ_negative_ignore_index = torch.tensor([-2, 1], device=device)
         targ_positive_ignore_index = torch.tensor([2, 1], device=device)
 
-        for reduction, label_smoothing, weight in product(reductions, label_smoothings, (None, weight)):
+        for reduction, label_smoothing, weight in product(reductions, label_smoothings, (None, wgt)):
             def check_equal(loss, inp_targ_1, inp_targ_2):
                 inp1, targ1 = inp_targ_1
                 inp2, targ2 = inp_targ_2
diff --git a/test/torch_np/numpy_tests/lib/test_function_base.py b/test/torch_np/numpy_tests/lib/test_function_base.py
index 0ada4214e1e2..3934613a64fc 100644
--- a/test/torch_np/numpy_tests/lib/test_function_base.py
+++ b/test/torch_np/numpy_tests/lib/test_function_base.py
@@ -734,9 +734,9 @@ class TestDiff(TestCase):
         x = list(range(3))
         assert_raises(ValueError, diff, x, n=-1)
         output = [diff(x, n=n) for n in range(1, 5)]
-        expected = [[1, 1], [0], [], []]
+        expected_output = [[1, 1], [0], [], []]
         # assert_(diff(x, n=0) is x)
-        for n, (expected, out) in enumerate(zip(expected, output), start=1):
+        for n, (expected, out) in enumerate(zip(expected_output, output), start=1):
             assert_(type(out) is np.ndarray)
             assert_array_equal(out, expected)
             assert_equal(out.dtype, np.int_)
diff --git a/torch/_inductor/lowering.py b/torch/_inductor/lowering.py
index 621710df1b7e..a45e3e3e7953 100644
--- a/torch/_inductor/lowering.py
+++ b/torch/_inductor/lowering.py
@@ -323,19 +323,19 @@ def broadcast_symbolic_shapes(a, b):
     are symbolic sympy formulas.
     """
     output = []
-    for a, b in itertools.zip_longest(
+    for x, y in itertools.zip_longest(
         reversed(a), reversed(b), fillvalue=sympy.Integer(1)
     ):
-        if b == 1:
-            output.append(a)
-        elif a == 1:
-            output.append(b)
+        if y == 1:
+            output.append(x)
+        elif x == 1:
+            output.append(y)
         else:
-            V.graph.sizevars.guard_equals(a, b)
-            if len(sympy.expand(b).free_symbols) < len(sympy.expand(a).free_symbols):
-                output.append(b)  # prefer shorter formula
+            V.graph.sizevars.guard_equals(x, y)
+            if len(sympy.expand(y).free_symbols) < len(sympy.expand(x).free_symbols):
+                output.append(y)  # prefer shorter formula
             else:
-                output.append(a)
+                output.append(x)
     return tuple(reversed(output))
 
 
diff --git a/torch/_inductor/triton_heuristics.py b/torch/_inductor/triton_heuristics.py
index 7fecff7cdb1d..be062e8eb3cb 100644
--- a/torch/_inductor/triton_heuristics.py
+++ b/torch/_inductor/triton_heuristics.py
@@ -540,8 +540,9 @@ def _find_names(obj):
     import inspect
 
     frame = inspect.currentframe()
-    for frame in iter(lambda: frame.f_back, None):  # type: ignore[union-attr]
+    while frame is not None:
         frame.f_locals
+        frame = frame.f_back
     obj_names = []
     for referrer in gc.get_referrers(obj):
         if isinstance(referrer, dict):
diff --git a/torch/autograd/functional.py b/torch/autograd/functional.py
index 30045bc8671c..bab85498bce8 100644
--- a/torch/autograd/functional.py
+++ b/torch/autograd/functional.py
@@ -550,9 +550,9 @@ def _jacfwd(func, inputs, strict=False, vectorize=False):
         is_outputs_tuple, outputs = output_info
         # Step 3: for each of the output tangents, split along dim 0
         jacobian_input_output = []
-        for jac, output_i in zip(outputs_before_split, outputs):
+        for jac_output_i, output_i in zip(outputs_before_split, outputs):
             jacobian_output_i_output = []
-            for jac, input_j in zip(jac.split(input_numels, dim=0), inputs):
+            for jac, input_j in zip(jac_output_i.split(input_numels, dim=0), inputs):
                 # We need to transpose the Jacobian because in forward AD, the
                 # batch dimension represents that of the inputs
                 jacobian_input_i_output_j = jac.permute(*range(1, jac.ndim), 0).reshape(
@@ -758,9 +758,11 @@ def jacobian(
             # Step 3: The returned jacobian is one big tensor per input. In this step,
             # we split each Tensor by output.
             jacobian_input_output = []
-            for jac, input_i in zip(jacobians_of_flat_output, inputs):
+            for jac_input_i, input_i in zip(jacobians_of_flat_output, inputs):
                 jacobian_input_i_output = []
-                for jac, output_j in zip(jac.split(output_numels, dim=0), outputs):
+                for jac, output_j in zip(
+                    jac_input_i.split(output_numels, dim=0), outputs
+                ):
                     jacobian_input_i_output_j = jac.view(output_j.shape + input_i.shape)
                     jacobian_input_i_output.append(jacobian_input_i_output_j)
                 jacobian_input_output.append(jacobian_input_i_output)
diff --git a/torch/fx/experimental/migrate_gradual_types/constraint_generator.py b/torch/fx/experimental/migrate_gradual_types/constraint_generator.py
index 7f83a639ff48..0b5307521533 100644
--- a/torch/fx/experimental/migrate_gradual_types/constraint_generator.py
+++ b/torch/fx/experimental/migrate_gradual_types/constraint_generator.py
@@ -1262,8 +1262,8 @@ class ConstraintGenerator:
             if isinstance(t, torch.Tensor):
                 if len(t.shape) > 0:
                     res = []
-                    for t in t.shape:
-                        res.append(t)
+                    for d in t.shape:
+                        res.append(d)
                     attr_type = TensorType(res)
                     output, counter = gen_tvar(counter)
                     self.symbol_dict[n] = output
diff --git a/torch/nn/parallel/distributed.py b/torch/nn/parallel/distributed.py
index 7b2005bdd593..dd674864d688 100644
--- a/torch/nn/parallel/distributed.py
+++ b/torch/nn/parallel/distributed.py
@@ -1286,8 +1286,8 @@ class DistributedDataParallel(Module, Joinable):
             )
             yield from ps
 
-        for m in m.modules() if recurse else [m]:
-            yield from model_parameters(m)
+        for mod in m.modules() if recurse else [m]:
+            yield from model_parameters(mod)
 
     def _check_default_group(self):
         pickle_not_supported = False
diff --git a/torch/testing/_internal/common_device_type.py b/torch/testing/_internal/common_device_type.py
index 934218bfff88..96b7817b5c4a 100644
--- a/torch/testing/_internal/common_device_type.py
+++ b/torch/testing/_internal/common_device_type.py
@@ -458,7 +458,7 @@ class DeviceTypeTestBase(TestCase):
             parametrize_fn = compose_parametrize_fns(dtype_parametrize_fn, parametrize_fn)
 
         # Instantiate the parametrized tests.
-        for (test, test_suffix, param_kwargs, decorator_fn) in parametrize_fn(test, generic_cls, cls):
+        for (test, test_suffix, param_kwargs, decorator_fn) in parametrize_fn(test, generic_cls, cls):  # noqa: B020
             test_suffix = '' if test_suffix == '' else '_' + test_suffix
             device_suffix = '_' + cls.device_type
 
diff --git a/torch/testing/_internal/common_methods_invocations.py b/torch/testing/_internal/common_methods_invocations.py
index 9683c02753fd..e24145a865da 100644
--- a/torch/testing/_internal/common_methods_invocations.py
+++ b/torch/testing/_internal/common_methods_invocations.py
@@ -4926,7 +4926,7 @@ def sample_inputs_put(op_info, device, dtype, requires_grad, **kwargs):
     tgt_gen = (make_arg(size) for size in tgt_sizes)
     idx = make_idx((0,), high=1)
     src = make_arg((0,))
-    for tgt, acc in product(tgt, (True, False)):
+    for tgt, acc in product(tgt_gen, (True, False)):
         yield SampleInput(input=tgt.clone().requires_grad_(requires_grad),
                           args=(idx.clone(),
                                 src.clone().requires_grad_(requires_grad),
@@ -8190,9 +8190,9 @@ def sample_inputs_scaled_dot_product_attention(op_info, device, dtype, requires_
 
     qkv_shapes = [(dim_3_q_shape, dim_3_kv_shape), (dim_4_q_shape, dim_4_kv_shape), broadcast_tuple]
     samples = []
-    for qkv_shapes, is_causal, dropout_p in product(
+    for qkv_shape, is_causal, dropout_p in product(
             qkv_shapes, [True, False], [0.0, 0.5]):
-        shape_q, shape_kv = qkv_shapes
+        shape_q, shape_kv = qkv_shape
         samples.append(SampleInput(
             make(shape_q),
             make(shape_kv),
diff --git a/torch/utils/data/datapipes/_hook_iterator.py b/torch/utils/data/datapipes/_hook_iterator.py
index dae358fd0143..7319f9e96657 100644
--- a/torch/utils/data/datapipes/_hook_iterator.py
+++ b/torch/utils/data/datapipes/_hook_iterator.py
@@ -40,9 +40,9 @@ def _generate_input_args_string(obj):
     for param_name in signature.parameters.keys():
         input_param_names.add(param_name)
     result = []
-    for name, obj in inspect.getmembers(obj):
+    for name, value in inspect.getmembers(obj):
         if name in input_param_names:
-            result.append((name, _simplify_obj_name(obj)))
+            result.append((name, _simplify_obj_name(value)))
     return ', '.join([f'{name}={value}' for name, value in result])