diff --git a/.lintrunner.toml b/.lintrunner.toml
index 13e6393dbecf71..4d47a8b3b4f8b3 100644
--- a/.lintrunner.toml
+++ b/.lintrunner.toml
@@ -2666,6 +2666,6 @@ init_command = [
     'python3',
     'tools/linter/adapters/pip_init.py',
     '--dry-run={{DRYRUN}}',
-    'ruff==0.0.285',
+    'ruff==0.0.280',
 ]
 is_formatter = true
diff --git a/benchmarks/functional_autograd_benchmark/torchvision_models.py b/benchmarks/functional_autograd_benchmark/torchvision_models.py
index 595259bd247d8b..3c1f2b471381cd 100644
--- a/benchmarks/functional_autograd_benchmark/torchvision_models.py
+++ b/benchmarks/functional_autograd_benchmark/torchvision_models.py
@@ -168,7 +168,7 @@ def __init__(
         if len(replace_stride_with_dilation) != 3:
             raise ValueError(
                 "replace_stride_with_dilation should be None "
-                f"or a 3-element tuple, got {replace_stride_with_dilation}"
+                "or a 3-element tuple, got {}".format(replace_stride_with_dilation)
             )
         self.groups = groups
         self.base_width = width_per_group
diff --git a/benchmarks/operator_benchmark/benchmark_core.py b/benchmarks/operator_benchmark/benchmark_core.py
index 439cdba8b36211..73a43d65ff9884 100644
--- a/benchmarks/operator_benchmark/benchmark_core.py
+++ b/benchmarks/operator_benchmark/benchmark_core.py
@@ -200,10 +200,10 @@ def __init__(self, args):
     def _print_header(self):
         DASH_LINE = "-" * 40
         print(
-            f"# {DASH_LINE}\n"
+            "# {}\n"
             "# PyTorch/Caffe2 Operator Micro-benchmarks\n"
-            f"# {DASH_LINE}\n"
-            f"# Tag : {self.args.tag_filter}\n"
+            "# {}\n"
+            "# Tag : {}\n".format(DASH_LINE, DASH_LINE, self.args.tag_filter)
         )
         if self.args.list_tests:
             print("# List of tests:")
diff --git a/benchmarks/overrides_benchmark/bench.py b/benchmarks/overrides_benchmark/bench.py
index 93c7f1b130d4a7..0811d0e0d76221 100644
--- a/benchmarks/overrides_benchmark/bench.py
+++ b/benchmarks/overrides_benchmark/bench.py
@@ -57,8 +57,10 @@ def main():
 
         bench_min, bench_std = bench(tensor_1, tensor_2)
         print(
-            f"Type {t.__name__} had a minimum time of {10**6 * bench_min} us"
-            f" and a standard deviation of {(10**6) * bench_std} us."
+            "Type {} had a minimum time of {} us"
+            " and a standard deviation of {} us.".format(
+                t.__name__, (10**6 * bench_min), (10**6) * bench_std
+            )
         )
 
 
diff --git a/pyproject.toml b/pyproject.toml
index eb764cb895fcb9..220986d1160ea7 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -74,7 +74,6 @@ select = [
     "PIE807",
     "PIE810",
     "PLE",
-    "RUF017",
     "TRY302",
 ]
 
diff --git a/test/cpp_api_parity/functional_impl_check.py b/test/cpp_api_parity/functional_impl_check.py
index ad56e917ef6d5f..828f57e7e69812 100644
--- a/test/cpp_api_parity/functional_impl_check.py
+++ b/test/cpp_api_parity/functional_impl_check.py
@@ -173,7 +173,7 @@ def write_test_to_test_class(
 
     assert not ('cpp_options_args' in test_params_dict and 'cpp_function_call' in test_params_dict), (
         "Only one of `cpp_options_args` and `cpp_function_call` entries "
-        f"should be present in test params dict:\n{pprint.pformat(test_params_dict)}")
+        "should be present in test params dict:\n{}").format(pprint.pformat(test_params_dict))
 
     functional_name = compute_functional_name(test_params_dict)
 
diff --git a/test/cpp_api_parity/module_impl_check.py b/test/cpp_api_parity/module_impl_check.py
index 1aa6273f0d99de..aa18798940ae28 100644
--- a/test/cpp_api_parity/module_impl_check.py
+++ b/test/cpp_api_parity/module_impl_check.py
@@ -209,11 +209,11 @@ def process_test_params_for_module(test_params_dict, device, test_instance_class
     if 'constructor_args' in test_params_dict:
         assert 'cpp_constructor_args' in test_params_dict, (
             "If `constructor_args` is present in test params dict, to enable C++ API parity test, "
-            f"`cpp_constructor_args` must be present in:\n{pprint.pformat(test_params_dict)}"
+            "`cpp_constructor_args` must be present in:\n{}"
             "If you are interested in adding the C++ API parity test, please see:\n"
             "NOTE [How to check NN module / functional API parity between Python and C++ frontends]. \n"
             "If not, please add `test_cpp_api_parity=False` to the test params dict and file an issue about this."
-        )
+        ).format(pprint.pformat(test_params_dict))
 
     return TorchNNModuleTestParams(
         module_name=module_name,
@@ -233,16 +233,16 @@ def write_test_to_test_class(
     module_name = compute_module_name(test_params_dict)
 
     assert hasattr(torch.nn, module_name), (
-        f"`torch.nn` doesn't have module `{module_name}`. "
+        "`torch.nn` doesn't have module `{}`. "
         "If you are adding a new test, please set `fullname` using format `ModuleName_desc` "
-        f"or set `module_name` using format `ModuleName` in the module test dict:\n{pprint.pformat(test_params_dict)}"
-    )
+        "or set `module_name` using format `ModuleName` in the module test dict:\n{}"
+    ).format(module_name, pprint.pformat(test_params_dict))
 
     module_full_name = 'torch::nn::' + module_name
 
     assert module_full_name in parity_table['torch::nn'], (
-        f"Please add `{module_full_name}` entry to `torch::nn` section of `test/cpp_api_parity/parity-tracker.md`. "
-        f"(Discovered while processing\n{pprint.pformat(test_params_dict)}.)")
+        "Please add `{}` entry to `torch::nn` section of `test/cpp_api_parity/parity-tracker.md`. "
+        "(Discovered while processing\n{}.)").format(module_full_name, pprint.pformat(test_params_dict))
 
     for device in devices:
         test_params = process_test_params_for_module(
diff --git a/test/nn/test_multihead_attention.py b/test/nn/test_multihead_attention.py
index d5ae098b1d25ae..7c1ba2084c29b5 100644
--- a/test/nn/test_multihead_attention.py
+++ b/test/nn/test_multihead_attention.py
@@ -329,7 +329,7 @@ def test_multihead_attn_3d_attn_mask(self):
         key = torch.rand(batch_size, src_len, embed_dim)  # [N, S, D]
         value = key  # [N, S, D]
         attn_mask = torch.randint(0, 2, (batch_size, tgt_len, src_len)).float()  # [N, T, S]
-        attn_mask = attn_mask.masked_fill(attn_mask == 0, float('-inf')).masked_fill(attn_mask == 1, 0.0)
+        attn_mask = attn_mask.masked_fill(attn_mask == 0, float('-inf')).masked_fill(attn_mask == 1, float(0.0))
 
         mta_model = torch.nn.MultiheadAttention(embed_dim, num_heads)
 
diff --git a/test/onnx/model_defs/squeezenet.py b/test/onnx/model_defs/squeezenet.py
index a97b399c7dac55..b6d83cd5305904 100644
--- a/test/onnx/model_defs/squeezenet.py
+++ b/test/onnx/model_defs/squeezenet.py
@@ -32,7 +32,8 @@ def __init__(self, version=1.0, num_classes=1000, ceil_mode=False):
         super().__init__()
         if version not in [1.0, 1.1]:
             raise ValueError(
-                f"Unsupported SqueezeNet version {version}:" "1.0 or 1.1 expected"
+                "Unsupported SqueezeNet version {version}:"
+                "1.0 or 1.1 expected".format(version=version)
             )
         self.num_classes = num_classes
         if version == 1.0:
diff --git a/test/quantization/core/test_docs.py b/test/quantization/core/test_docs.py
index a5a49f2ff2fa61..82ef4d277805bb 100644
--- a/test/quantization/core/test_docs.py
+++ b/test/quantization/core/test_docs.py
@@ -82,8 +82,10 @@ def get_correct_path(path_from_pytorch):
 
             # want to make sure we are actually getting some code,
             assert last_line_num - line_num_start > 3 or short_snippet, (
-                f"The code in {path_to_file} identified by {unique_identifier} seems suspiciously short:"
-                f"\n\n###code-start####\n{code}###code-end####"
+                "The code in {} identified by {} seems suspiciously short:"
+                "\n\n###code-start####\n{}###code-end####".format(
+                    path_to_file, unique_identifier, code
+                )
             )
             return code
 
diff --git a/test/quantization/core/test_quantized_op.py b/test/quantization/core/test_quantized_op.py
index 1d84f29cd66dca..da7ef834727104 100644
--- a/test/quantization/core/test_quantized_op.py
+++ b/test/quantization/core/test_quantized_op.py
@@ -806,11 +806,11 @@ def _test_binary_op_scalar_relu(self, A, b, binary_op_name, binary_op, quantized
             C_relu, C_relu_hat.q_scale(), C_relu_hat.q_zero_point(), dtype)
 
         self.assertEqual(C_ref.dequantize(), C_hat.dequantize(),
-                         msg=f"{binary_op_name}_scalar results don't match: "
-                         f"{C_ref.dequantize()} vs {C_hat.dequantize()}")
+                         msg="{}_scalar results don't match: "
+                         "{} vs {}".format(binary_op_name, C_ref.dequantize(), C_hat.dequantize()))
         self.assertEqual(C_relu_ref.dequantize(), C_relu_hat.dequantize(),
-                         msg=f"{binary_op_name}_scalar_relu results don't match: "
-                         f"{C_relu_ref.dequantize()} vs {C_relu_hat.dequantize()}")
+                         msg="{}_scalar_relu results don't match: "
+                         "{} vs {}".format(binary_op_name, C_relu_ref.dequantize(), C_relu_hat.dequantize()))
 
     @unittest.skipIf(IS_MACOS, "skipping macos test")
     @given(A=hu.tensor(shapes=hu.array_shapes(1, 4, 1, 5),
diff --git a/test/test_autocast.py b/test/test_autocast.py
index 79499d8cb50d5e..7e449df11fefc5 100644
--- a/test/test_autocast.py
+++ b/test/test_autocast.py
@@ -67,7 +67,7 @@ def compare(first, second):
             if (output is not None) and (output_method is not None):
                 self.assertTrue(type(output) == type(output_method))
                 comparison = compare(output, output_method)
-                self.assertTrue(comparison, f"torch.{op} result did not match Tensor.{op} result")
+                self.assertTrue(comparison, "torch.{0} result did not match Tensor.{0} result".format(op))
 
             # Compare numerics to Python-side "autocasting" that (we expect) does the same thing
             # as the C++-side autocasting, and should be bitwise accurate.
diff --git a/test/test_binary_ufuncs.py b/test/test_binary_ufuncs.py
index c4bb5a1ba0b54a..ada4005d128c4d 100644
--- a/test/test_binary_ufuncs.py
+++ b/test/test_binary_ufuncs.py
@@ -168,9 +168,9 @@ def _numel(x):
             if _numel(l) <= 100 and _numel(r) <= 100:
                 msg = (
                     "Failed to produce expected results! Input lhs tensor was"
-                    f" {l}, rhs tensor was {r}, torch result is {actual}, and reference result is"
-                    f" {expected}."
-                )
+                    " {}, rhs tensor was {}, torch result is {}, and reference result is"
+                    " {}."
+                ).format(l, r, actual, expected)
             else:
                 msg = None
 
diff --git a/test/test_cpp_extensions_jit.py b/test/test_cpp_extensions_jit.py
index c7ebe5aef74bdb..85b5a410742ff6 100644
--- a/test/test_cpp_extensions_jit.py
+++ b/test/test_cpp_extensions_jit.py
@@ -150,14 +150,17 @@ def _check_cuobjdump_output(expected_values, is_ptx=False):
             err = err.decode("ascii")
 
             if not p.returncode == 0 or not err == '':
-                raise AssertionError(f"Flags: {flags}\nReturncode: {p.returncode}\nStderr: {err}\n"
-                                     f"Output: {output} ")
+                raise AssertionError("Flags: {}\nReturncode: {}\nStderr: {}\n"
+                                     "Output: {} ".format(flags, p.returncode,
+                                                          err, output))
 
             actual_arches = sorted(re.findall(r'sm_\d\d', output))
             expected_arches = sorted(['sm_' + xx for xx in expected_values])
             self.assertEqual(actual_arches, expected_arches,
-                             msg=f"Flags: {flags},  Actual: {actual_arches},  Expected: {expected_arches}\n"
-                                 f"Stderr: {err}\nOutput: {output}")
+                             msg="Flags: {},  Actual: {},  Expected: {}\n"
+                                 "Stderr: {}\nOutput: {}".format(
+                                     flags, actual_arches, expected_arches,
+                                     err, output))
 
         temp_dir = tempfile.mkdtemp()
         old_envvar = os.environ.get('TORCH_CUDA_ARCH_LIST', None)
diff --git a/test/test_cuda.py b/test/test_cuda.py
index 5f637ac9fb8580..43914649992c51 100644
--- a/test/test_cuda.py
+++ b/test/test_cuda.py
@@ -168,7 +168,7 @@ def test_out_of_memory_retry(self):
     def test_set_per_process_memory_fraction(self):
         # test invalid fraction value.
         with self.assertRaisesRegex(TypeError, "Invalid type"):
-            torch.cuda.set_per_process_memory_fraction(1)
+            torch.cuda.set_per_process_memory_fraction(int(1))
         with self.assertRaisesRegex(ValueError, "Invalid fraction value"):
             torch.cuda.set_per_process_memory_fraction(-0.1)
         with self.assertRaisesRegex(ValueError, "Invalid fraction value"):
@@ -1765,7 +1765,7 @@ def compare(first, second):
             if (output is not None) and (output_method is not None):
                 self.assertTrue(type(output) == type(output_method))
                 comparison = compare(output, output_method)
-                self.assertTrue(comparison, f"torch.{op} result did not match Tensor.{op} result")
+                self.assertTrue(comparison, "torch.{0} result did not match Tensor.{0} result".format(op))
 
             # Compare numerics to Python-side "autocasting" that (we expect) does the same thing
             # as the C++-side autocasting, and should be bitwise accurate.
diff --git a/test/test_dispatch.py b/test/test_dispatch.py
index 5b03c49c3090d0..cb485bda7af49c 100644
--- a/test/test_dispatch.py
+++ b/test/test_dispatch.py
@@ -152,10 +152,10 @@ def check_invariants(actual_provenance):
                 # NB: this finally test asserts that if a registrations fails,
                 # the dispatcher is left in the same state *that it was before*!
                 check_invariants(
-                    f"running ctors {ctor_order[:i]} and then failing to run ctor {op_ix} "
+                    "running ctors {} and then failing to run ctor {} "
                     "(did this failure leave the dispatcher in a wedged state? "
                     "it shouldn't!)"
-                )
+                    .format(ctor_order[:i], op_ix))
                 break
         last_ctor = i
         if expect_raises and len(active_ops) == len(ops):
@@ -165,7 +165,7 @@ def check_invariants(actual_provenance):
             self.assertTrue(
                 False,
                 "expected exception to be raised, but nothing was raised "
-                f"(after running ctors {ctor_order})")
+                "(after running ctors {})".format(ctor_order))
         # In the order specified by dtor_order, run deregistrations
         for i, op_ix in enumerate(dtor_order):
             # Trigger a destruction
diff --git a/test/test_jit.py b/test/test_jit.py
index cb71f5c888a830..03fc3679e2c3d6 100644
--- a/test/test_jit.py
+++ b/test/test_jit.py
@@ -5810,19 +5810,19 @@ def test_dispatch(op, expects, dtype, binary=False):
                 raise RuntimeError('Unknown dtype')
 
             if binary:
-                code = f'''
+                code = '''
                     graph(%3 : Tensor, %4 : Tensor):
-                        %2 : {dtype_str}(*, *) = aten::{op}(%3, %4)
-                        %1 : {dtype_str}(*, *) = aten::relu(%2)
+                        %2 : {dtype}(*, *) = aten::{op}(%3, %4)
+                        %1 : {dtype}(*, *) = aten::relu(%2)
                         return (%1)
-                '''
+                '''.format(op=op, dtype=dtype_str)
             else:
-                code = f'''
+                code = '''
                     graph(%3 : Tensor):
-                        %2 : {dtype_str}(*, *) = aten::{op}(%3)
-                        %1 : {dtype_str}(*, *) = aten::relu(%2)
+                        %2 : {dtype}(*, *) = aten::{op}(%3)
+                        %1 : {dtype}(*, *) = aten::relu(%2)
                         return (%1)
-                '''
+                '''.format(op=op, dtype=dtype_str)
 
             graph = parse_ir(code)
             inputs = (2 if binary else 1) * [torch.rand(26, 2048, dtype=dtype)]
@@ -14936,7 +14936,7 @@ def jit_multihead_attn_forward(query,                   # type: Tensor
         value = torch.rand((src_l, bsz, embed_size))
 
         mask = (torch.triu(torch.ones(src_l, src_l)) == 1).transpose(0, 1)
-        mask = mask.float().masked_fill(mask == 0, float('-inf')).masked_fill(mask == 1, 0.0).to(torch.get_default_dtype())
+        mask = mask.float().masked_fill(mask == 0, float('-inf')).masked_fill(mask == 1, float(0.0)).to(torch.get_default_dtype())
 
         jit_out = jit_multihead_attn_forward(query, key, value,
                                              embed_size, nhead,
diff --git a/test/test_mobile_optimizer.py b/test/test_mobile_optimizer.py
index e672d69ab5dd2d..eeb62ecc4bc444 100644
--- a/test/test_mobile_optimizer.py
+++ b/test/test_mobile_optimizer.py
@@ -582,7 +582,7 @@ def dummy_method_ref_attr_pqr(self):
         self.assertTrue(
             cloned.qualified_name.startswith('__torch__.'),
             ("Expected the cloned module's name to start with the string "
-             f"'__torch__.', but got: {cloned.qualified_name}"),
+             "'__torch__.', but got: {}").format(cloned.qualified_name),
         )
 
 
diff --git a/test/test_mps.py b/test/test_mps.py
index e4cdd1dbfa0a17..b2f7d649be50cb 100644
--- a/test/test_mps.py
+++ b/test/test_mps.py
@@ -9436,8 +9436,8 @@ def get_grid(device='cpu', data=None):
                     output = F.grid_sample(input, grid, mode=mode, padding_mode=padding_mode,
                                            align_corners=align_corners)
                     self.assertEqual(output, groundtruth, atol=1e-5, rtol=0,
-                                     msg=f"groundtruth comparison failed for mode={mode}, "
-                                     f"padding_mode={padding_mode}")
+                                     msg="groundtruth comparison failed for mode={}, "
+                                     "padding_mode={}".format(mode, padding_mode))
 
 class TestAdvancedIndexing(TestCaseMPS):
     supported_dtypes = [torch.float32, torch.float16, torch.int64, torch.int32, torch.int16, torch.uint8]
diff --git a/test/test_nn.py b/test/test_nn.py
index 556e184c99406b..85aa6034600bbb 100644
--- a/test/test_nn.py
+++ b/test/test_nn.py
@@ -5959,8 +5959,8 @@ def get_grid(device='cpu', data=None):
                     output = F.grid_sample(input, grid, mode=mode, padding_mode=padding_mode,
                                            align_corners=align_corners)
                     self.assertEqual(output, groundtruth, atol=1e-5, rtol=0,
-                                     msg=f"groundtruth comparison failed for mode={mode}, "
-                                     f"padding_mode={padding_mode}")
+                                     msg="groundtruth comparison failed for mode={}, "
+                                     "padding_mode={}".format(mode, padding_mode))
 
                     # See NOTE [ grid_sample CPU fallback ]
                     output = torch._grid_sampler_2d_cpu_fallback(
@@ -6047,8 +6047,8 @@ def get_grid(device='cpu', data=None):
                         F.grid_sample(input, grid, mode=mode, padding_mode=padding_mode,
                                       align_corners=align_corners).sum().backward()
                         self.assertEqual(grid.grad, groundtruth, atol=1e-5, rtol=0,
-                                         msg=f"gradient groundtruth comparison failed for mode={mode}, "
-                                         f"padding_mode={padding_mode}, input_requires_grad={input_requires_grad}")
+                                         msg="gradient groundtruth comparison failed for mode={}, "
+                                         "padding_mode={}, input_requires_grad={}".format(mode, padding_mode, input_requires_grad))
                         grid.grad.zero_()
 
                     # See NOTE [ grid_sample CPU fallback ]
diff --git a/test/test_ops.py b/test/test_ops.py
index a3b14857cae0f4..ab2d1aa599a701 100644
--- a/test/test_ops.py
+++ b/test/test_ops.py
@@ -1098,8 +1098,10 @@ def _test_consistency_helper(samples, variants):
                             RuntimeError,
                             msg=(
                                 "inplace variant either incorrectly allowed "
-                                f"resizing or you have marked the sample {sample.summary()}"
-                                " incorrectly with `broadcasts_self=True"
+                                "resizing or you have marked the sample {}"
+                                " incorrectly with `broadcasts_self=True".format(
+                                    sample.summary()
+                                )
                             ),
                         ):
                             variant_forward = variant(
diff --git a/test/test_reductions.py b/test/test_reductions.py
index 7ba000f4760235..8cb0a64ce2987d 100644
--- a/test/test_reductions.py
+++ b/test/test_reductions.py
@@ -3503,8 +3503,8 @@ def to_numpy(input):
             expected = np.asarray(expected)  # transform numpy scalars to numpy.ndarray instances
 
             msg = ("Failed to produce expected results! Input tensor was"
-                   f" {t}, torch result is {actual}, and reference result is"
-                   f" {expected}.") if t.numel() < 10 else None
+                   " {}, torch result is {}, and reference result is"
+                   " {}.").format(t, actual, expected) if t.numel() < 10 else None
 
             self.assertEqual(actual, expected, msg, exact_dtype=exact_dtype)
 
diff --git a/test/test_tensor_creation_ops.py b/test/test_tensor_creation_ops.py
index 815e28282c3ba6..cc419adf8d2c24 100644
--- a/test/test_tensor_creation_ops.py
+++ b/test/test_tensor_creation_ops.py
@@ -451,8 +451,9 @@ def dtype_name(dtype):
             other_dtype = torch.float64 if dtype == torch.float32 else torch.float32
             a = torch.tensor([1, 2], device=device, dtype=dtype)
             b = torch.tensor([3, 4], device=device, dtype=other_dtype)
-            error = f"Expected object of scalar type {dtype_name(dtype)} but got scalar type " \
-                    f"{dtype_name(other_dtype)} for second argument"
+            error = "Expected object of scalar type {} but got scalar type " \
+                    "{} for second argument".format(dtype_name(dtype),
+                                                    dtype_name(other_dtype))
             with self.assertRaisesRegex(RuntimeError, error):
                 op(a, b)
 
@@ -471,8 +472,9 @@ def complex_dtype_name(dtype):
             b = torch.tensor([3, 4], device=device, dtype=dtype)
             out = torch.zeros(2, device=device, dtype=dtype)
             expected_dtype = torch.complex64 if dtype == torch.float32 else torch.complex128
-            error = f"Expected object of scalar type {complex_dtype_name(expected_dtype)} but got scalar type " \
-                    f"{dtype_name(dtype)} for argument 'out'"
+            error = "Expected object of scalar type {} but got scalar type " \
+                    "{} for argument 'out'".format(
+                        complex_dtype_name(expected_dtype), dtype_name(dtype))
             with self.assertRaisesRegex(RuntimeError, error):
                 op(a, b, out=out)
 
diff --git a/test/test_torch.py b/test/test_torch.py
index 2d0e02de141c26..ab1a92bdb43771 100644
--- a/test/test_torch.py
+++ b/test/test_torch.py
@@ -8682,8 +8682,8 @@ def test_doc_template(self) -> None:
 
             for common_args in [multi_dim_common, single_dim_common, factory_common_args, factory_like_common_args]:
                 for k, v in common_args.items():
-                    self.assertNotIn(v, desc, f'The argument description "{v}" in {func} can be '
-                                              f'replaced by {{{k}}}')
+                    self.assertNotIn(v, desc, 'The argument description "{}" in {} can be '
+                                              'replaced by {{{}}}'.format(v, func, k))
 
     def test_doc(self):
         checked_types = (types.MethodType, types.FunctionType,
@@ -8719,8 +8719,8 @@ def _test_namespace(ns, *skips):
                 full_name = ns_name + '.' + name
                 if any(r.match(name) for r in skip_regexes):
                     self.assertFalse(has_doc,
-                                     f'New docs have been added for {full_name}, please remove '
-                                     'it from the skipped list in TestTorch.test_doc')
+                                     'New docs have been added for {}, please remove '
+                                     'it from the skipped list in TestTorch.test_doc'.format(full_name))
                 else:
                     self.assertTrue(has_doc, f'{full_name} is missing documentation')
 
diff --git a/test/test_transformers.py b/test/test_transformers.py
index 8a57af902c6e66..52eabc0b9a0552 100644
--- a/test/test_transformers.py
+++ b/test/test_transformers.py
@@ -146,7 +146,7 @@ def test_self_attn_TxT_attn_mask(self, device):
 
         query = torch.rand(batch_size, tgt_len, embed_dim, device=device)  # [N, T, D]
         attn_mask = torch.randint(0, 2, (tgt_len, tgt_len)).cuda().float()  # [T, T]
-        attn_mask = attn_mask.masked_fill(attn_mask == 0, float('-inf')).masked_fill(attn_mask == 1, 0.0)
+        attn_mask = attn_mask.masked_fill(attn_mask == 0, float('-inf')).masked_fill(attn_mask == 1, float(0.0))
 
         attn_mask_4d = attn_mask.expand(batch_size, num_heads, tgt_len, tgt_len)
 
diff --git a/test/test_unary_ufuncs.py b/test/test_unary_ufuncs.py
index eb6e270cde57d9..6c6744b6b781e6 100644
--- a/test/test_unary_ufuncs.py
+++ b/test/test_unary_ufuncs.py
@@ -101,9 +101,9 @@ def test_float_domains(self, device, dtype, op):
                     result.item(),
                     float("nan"),
                     msg=(
-                        f"input of {lower_tensor.item()} outside lower domain boundary"
-                        f" {low} produced {result.item()}, not nan!"
-                    ),
+                        "input of {} outside lower domain boundary"
+                        " {} produced {}, not nan!"
+                    ).format(lower_tensor.item(), low, result.item()),
                 )
 
         if high is not None:
@@ -120,9 +120,9 @@ def test_float_domains(self, device, dtype, op):
                     result.item(),
                     float("nan"),
                     msg=(
-                        f"input of {higher_tensor.item()} outside upper domain boundary"
-                        f" {high} produced {result.item()}, not nan!"
-                    ),
+                        "input of {} outside upper domain boundary"
+                        " {} produced {}, not nan!"
+                    ).format(higher_tensor.item(), high, result.item()),
                 )
 
     # Helper for comparing torch tensors and numpy arrays
@@ -245,9 +245,9 @@ def _helper_reference_numerics(
             if t.numel() < 10:
                 msg = (
                     "Failed to produce expected results! Input tensor was"
-                    f" {t}, torch result is {actual}, and reference result is"
-                    f" {expected}."
-                )
+                    " {}, torch result is {}, and reference result is"
+                    " {}."
+                ).format(t, actual, expected)
             else:
                 msg = None
 
diff --git a/test/torch_np/numpy_tests/core/test_multiarray.py b/test/torch_np/numpy_tests/core/test_multiarray.py
index 5f48d1c848ed1e..65bd6edb353b35 100644
--- a/test/torch_np/numpy_tests/core/test_multiarray.py
+++ b/test/torch_np/numpy_tests/core/test_multiarray.py
@@ -6431,7 +6431,7 @@ def test_exotic_2(self):
         e = float("-Infinity")
         assert_equal(np.where(True, d, e).dtype, np.float32)
         # also check upcast
-        e = 1e150
+        e = float(1e150)
         assert_equal(np.where(True, d, e).dtype, np.float64)
 
     def test_ndim(self):
diff --git a/test/torch_np/numpy_tests/core/test_scalarmath.py b/test/torch_np/numpy_tests/core/test_scalarmath.py
index 6e9f70392f5d78..6819dc640d01a6 100644
--- a/test/torch_np/numpy_tests/core/test_scalarmath.py
+++ b/test/torch_np/numpy_tests/core/test_scalarmath.py
@@ -264,7 +264,9 @@ def test_mixed_types(self):
                 a = t1(3)
                 b = t2(2)
                 result = a**b
-                msg = f"error with {t1!r} and {t2!r}:" f"got {result!r}, expected {9!r}"
+                msg = ("error with {!r} and {!r}:" "got {!r}, expected {!r}").format(
+                    t1, t2, result, 9
+                )
                 if np.issubdtype(np.dtype(result), np.integer):
                     assert_(result == 9, msg)
                 else:
diff --git a/test/torch_np/numpy_tests/lib/test_histograms.py b/test/torch_np/numpy_tests/lib/test_histograms.py
index 156a61b2d0a1e0..e3c0d5cbbff6f0 100644
--- a/test/torch_np/numpy_tests/lib/test_histograms.py
+++ b/test/torch_np/numpy_tests/lib/test_histograms.py
@@ -485,8 +485,8 @@ def test_simple(self):
                 assert_equal(
                     len(a),
                     numbins,
-                    err_msg=f"For the {estimator} estimator "
-                    f"with datasize of {testlen}",
+                    err_msg="For the {} estimator "
+                    "with datasize of {}".format(estimator, testlen),
                 )
 
     def test_small(self):
@@ -532,8 +532,8 @@ def test_small(self):
                 assert_equal(
                     len(a),
                     expbins,
-                    err_msg=f"For the {estimator} estimator "
-                    f"with datasize of {testlen}",
+                    err_msg="For the {} estimator "
+                    "with datasize of {}".format(estimator, testlen),
                 )
 
     def test_incorrect_methods(self):
@@ -566,7 +566,7 @@ def test_novariance(self):
             assert_equal(
                 len(a),
                 numbins,
-                err_msg=f"{estimator} estimator, " "No Variance test",
+                err_msg="{} estimator, " "No Variance test".format(estimator),
             )
 
     def test_limited_variance(self):
diff --git a/tools/pyi/gen_pyi.py b/tools/pyi/gen_pyi.py
index 1f01d96ad5ccfa..d255852be11215 100644
--- a/tools/pyi/gen_pyi.py
+++ b/tools/pyi/gen_pyi.py
@@ -462,7 +462,7 @@ def gen_nn_functional(fm: FileManager) -> None:
         "pdist",
         "cosine_similarity",
     ]
-    imported_hints = [f"from .. import {_} as {_}" for _ in torch_imports]
+    imported_hints = ["from .. import {0} as {0}".format(_) for _ in torch_imports]
 
     # Functions imported into `torch.nn.functional` from `torch._C._nn`
     c_nn_imports = [
@@ -479,7 +479,9 @@ def gen_nn_functional(fm: FileManager) -> None:
         "one_hot",
         "scaled_dot_product_attention",
     ]
-    imported_hints += [f"from .._C._nn import {_} as {_}" for _ in c_nn_imports]
+    imported_hints += [
+        "from .._C._nn import {0} as {0}".format(_) for _ in c_nn_imports
+    ]
     # This is from `torch._C._nn` but renamed
     imported_hints.append("from .._C._nn import log_sigmoid\nlogsigmoid = log_sigmoid")
 
@@ -873,13 +875,15 @@ def gen_pyi(
     )
     for binop in ["mul", "true_divide", "floor_divide"]:
         unsorted_function_hints[binop].append(
-            f"def {binop}(input: Union[Tensor, Number], other: Union[Tensor, Number], "
-            "*, out: Optional[Tensor] = None) -> Tensor: ..."
+            "def {}(input: Union[Tensor, Number], other: Union[Tensor, Number], "
+            "*, out: Optional[Tensor] = None) -> Tensor: ...".format(binop)
         )
     for binop in ["add", "sub"]:
         unsorted_function_hints[binop].append(
-            f"def {binop}(input: Union[Tensor, Number], other: Union[Tensor, Number], "
-            "*, alpha: Optional[Number] = 1, out: Optional[Tensor] = None) -> Tensor: ..."
+            "def {}(input: Union[Tensor, Number], other: Union[Tensor, Number], "
+            "*, alpha: Optional[Number] = 1, out: Optional[Tensor] = None) -> Tensor: ...".format(
+                binop
+            )
         )
 
     native_functions = parse_native_yaml(
@@ -1082,8 +1086,8 @@ def replace_special_case(hint: str) -> str:
                 binop += "_"
                 out_suffix = ""
             unsorted_tensor_method_hints[binop].append(
-                f"def {binop}(self, other: Union[Tensor, Number, torch.SymInt, torch.SymFloat]{out_suffix})"
-                " -> Tensor: ..."
+                "def {}(self, other: Union[Tensor, Number, torch.SymInt, torch.SymFloat]{})"
+                " -> Tensor: ...".format(binop, out_suffix)
             )
     for binop in ["add", "sub"]:
         for inplace in [False, True]:
@@ -1092,9 +1096,9 @@ def replace_special_case(hint: str) -> str:
                 binop += "_"
                 out_suffix = ""
             unsorted_tensor_method_hints[binop].append(
-                f"def {binop}(self, other: Union[Tensor, Number, torch.SymInt, torch.SymFloat], "
-                f"*, alpha: Optional[Number] = 1{out_suffix})"
-                " -> Tensor: ..."
+                "def {}(self, other: Union[Tensor, Number, torch.SymInt, torch.SymFloat], "
+                "*, alpha: Optional[Number] = 1{})"
+                " -> Tensor: ...".format(binop, out_suffix)
             )
     simple_conversions = [
         "byte",
diff --git a/torch/__init__.py b/torch/__init__.py
index 55bb56033376ea..f3caf80fd081b9 100644
--- a/torch/__init__.py
+++ b/torch/__init__.py
@@ -1733,8 +1733,8 @@ def _register_device_module(device_type, module):
     device_type = torch.device(device_type).type
     m = sys.modules[__name__]
     if hasattr(m, device_type):
-        raise RuntimeError(f"The runtime module of '{device_type}' has already "
-                           f"been registered with '{getattr(m, device_type)}'")
+        raise RuntimeError("The runtime module of '{}' has already "
+                           "been registered with '{}'".format(device_type, getattr(m, device_type)))
     setattr(m, device_type, module)
     torch_module_name = '.'.join([__name__, device_type])
     sys.modules[torch_module_name] = module
diff --git a/torch/_lobpcg.py b/torch/_lobpcg.py
index a5ed5cf8fcfd26..16f37cdf6a4a8e 100644
--- a/torch/_lobpcg.py
+++ b/torch/_lobpcg.py
@@ -614,8 +614,8 @@ def _lobpcg(
 
     if m < 3 * n:
         raise ValueError(
-            f"LPBPCG algorithm is not applicable when the number of A rows (={m})"
-            f" is smaller than 3 x the number of requested eigenpairs (={n})"
+            "LPBPCG algorithm is not applicable when the number of A rows (={})"
+            " is smaller than 3 x the number of requested eigenpairs (={})".format(m, n)
         )
 
     method = "ortho" if method is None else method
@@ -1151,7 +1151,9 @@ def _get_ortho(self, U, V):
                 assert B is not None
                 raise ValueError(
                     "Overdetermined shape of U:"
-                    f" #B-cols(={B.shape[-1]}) >= #U-cols(={U.shape[-1]}) + #V-cols(={V.shape[-1]}) must hold"
+                    " #B-cols(={}) >= #U-cols(={}) + #V-cols(={}) must hold".format(
+                        B.shape[-1], U.shape[-1], V.shape[-1]
+                    )
                 )
         self.ivars["ortho_i"] = i
         self.ivars["ortho_j"] = j
diff --git a/torch/_namedtensor_internals.py b/torch/_namedtensor_internals.py
index cbc9de2de091d0..47bdcd82d14ab7 100644
--- a/torch/_namedtensor_internals.py
+++ b/torch/_namedtensor_internals.py
@@ -50,8 +50,8 @@ def single_ellipsis_index(names, fn_name):
     ellipsis_indices = [i for i, name in enumerate(names) if is_ellipsis(name)]
     if len(ellipsis_indices) >= 2:
         raise RuntimeError(
-            f"{fn_name}: More than one Ellipsis ('...') found in names ("
-            f"{names}). This function supports up to one Ellipsis."
+            "{}: More than one Ellipsis ('...') found in names ("
+            "{}). This function supports up to one Ellipsis.".format(fn_name, names)
         )
     if len(ellipsis_indices) == 1:
         return ellipsis_indices[0]
@@ -97,8 +97,15 @@ def update_names_with_mapping(tensor, rename_map, inplace):
             dim_map[old_dim] = new_dim
         else:
             raise RuntimeError(
-                f"{namer_api_name(inplace)}: Tried to rename dim '{old_dim}' to dim "
-                f"{new_dim} in Tensor[{tensor.names}] but dim '{old_dim}' does not exist"
+                (
+                    "{api_name}: Tried to rename dim '{old_dim}' to dim "
+                    "{new_dim} in Tensor[{dims}] but dim '{old_dim}' does not exist"
+                ).format(
+                    old_dim=old_dim,
+                    new_dim=new_dim,
+                    dims=tensor.names,
+                    api_name=namer_api_name(inplace),
+                )
             )
     return tensor._update_names(tuple(dim_map.values()), inplace)
 
@@ -142,10 +149,10 @@ def update_names(tensor, names, rename_map, inplace):
     has_rename_pairs = bool(rename_map)
     if has_names and has_rename_pairs:
         raise RuntimeError(
-            f"{namer_api_name(inplace)}: This function takes either positional "
-            f"args or keyword args, but not both. Use tensor.{namer_api_name(inplace)}(*names) "
-            f"to name dims and tensor.{namer_api_name(inplace)}(**rename_map) to rename "
-            "dims."
+            "{api_name}: This function takes either positional "
+            "args or keyword args, but not both. Use tensor.{api_name}(*names) "
+            "to name dims and tensor.{api_name}(**rename_map) to rename "
+            "dims.".format(api_name=namer_api_name(inplace))
         )
 
     # Special case for tensor.rename(*[]), which is valid for a 0 dim tensor.
diff --git a/torch/_numpy/testing/utils.py b/torch/_numpy/testing/utils.py
index 03f0e99b93d470..d48809dc14653c 100644
--- a/torch/_numpy/testing/utils.py
+++ b/torch/_numpy/testing/utils.py
@@ -1410,8 +1410,8 @@ def assert_array_max_ulp(a, b, maxulp=1, dtype=None):
     ret = nulp_diff(a, b, dtype)
     if not np.all(ret <= maxulp):
         raise AssertionError(
-            f"Arrays are not almost equal up to {maxulp:g} "
-            f"ULP (max difference is {np.max(ret):g} ULP)"
+            "Arrays are not almost equal up to {:g} "
+            "ULP (max difference is {:g} ULP)".format(maxulp, np.max(ret))
         )
     return ret
 
diff --git a/torch/_ops.py b/torch/_ops.py
index 19d7df3aab237b..329d8fa534a950 100644
--- a/torch/_ops.py
+++ b/torch/_ops.py
@@ -646,8 +646,10 @@ def __getattr__(self, key):
             # an object name different from the one the attribute
             # query was performed on.
             raise AttributeError(
-                f"'{str(self)}' can't have an overload name beginning with '__' and the "
-                f"underlying op {str(self._op)} has no attribute {key} either."
+                "'{}' can't have an overload name beginning with '__' and the "
+                "underlying op {} has no attribute {} either.".format(
+                    str(self), str(self._op), key
+                )
             ) from None
 
         try:
diff --git a/torch/_prims/__init__.py b/torch/_prims/__init__.py
index 8b1a58eef78648..0e7d06a1e5bd43 100644
--- a/torch/_prims/__init__.py
+++ b/torch/_prims/__init__.py
@@ -1483,8 +1483,10 @@ def _slice_meta(
             raise ValueError(msg)
         if x > y:
             msg = (
-                f"Attempting to slice a tensor but a start index in {start_indices} is greater than"
-                f" the length of its corresponding dimension in shape {a.shape}"
+                "Attempting to slice a tensor but a start index in {} is greater than"
+                " the length of its corresponding dimension in shape {}".format(
+                    start_indices, a.shape
+                )
             )
             raise ValueError(msg)
 
@@ -1494,14 +1496,16 @@ def _slice_meta(
             raise ValueError(msg)
         if x > y:
             msg = (
-                f"Attempting to slice a tensor but a stop index in {limit_indices} is greater than the length of "
-                f" its corresponding dimension in shape {a.shape}"
+                "Attempting to slice a tensor but a stop index in {} is greater than the length of "
+                " its corresponding dimension in shape {}".format(
+                    limit_indices, a.shape
+                )
             )
             raise ValueError(msg)
         if x < z:
             msg = (
-                f"Attempting to slice a tensor but a start index in {x} is greater than "
-                f" its corresponding stop index {z}"
+                "Attempting to slice a tensor but a start index in {} is greater than "
+                " its corresponding stop index {}".format(x, z)
             )
 
     for x in _strides:
diff --git a/torch/_refs/nn/functional/__init__.py b/torch/_refs/nn/functional/__init__.py
index 8eafe93e635060..a6427977474036 100644
--- a/torch/_refs/nn/functional/__init__.py
+++ b/torch/_refs/nn/functional/__init__.py
@@ -605,7 +605,9 @@ def margin_ranking_loss(
     if input1.ndim != input2.ndim or input1.ndim != target.ndim:
         raise RuntimeError(
             "margin_ranking_loss : All input tensors should have same dimension but got sizes: "
-            f"input1: {input1.shape}, input2: {input2.shape}, target: {target.shape} "
+            "input1: {}, input2: {}, target: {} ".format(
+                input1.shape, input2.shape, target.shape
+            )
         )
     _check_reduction_value(reduction)
     loss = torch.clamp_min(-target * (input1 - input2) + margin, 0)
diff --git a/torch/ao/nn/quantizable/modules/rnn.py b/torch/ao/nn/quantizable/modules/rnn.py
index 882c63d699dcf8..b4de0fd1ed9ca6 100644
--- a/torch/ao/nn/quantizable/modules/rnn.py
+++ b/torch/ao/nn/quantizable/modules/rnn.py
@@ -318,8 +318,8 @@ def __init__(self, input_size: int, hidden_size: int,
             if num_layers == 1:
                 warnings.warn("dropout option adds dropout after all but last "
                               "recurrent layer, so non-zero dropout expects "
-                              f"num_layers greater than 1, but got dropout={dropout} "
-                              f"and num_layers={num_layers}")
+                              "num_layers greater than 1, but got dropout={} "
+                              "and num_layers={}".format(dropout, num_layers))
 
         layers = [_LSTMLayer(self.input_size, self.hidden_size,
                              self.bias, batch_first=False,
diff --git a/torch/ao/nn/quantized/dynamic/modules/linear.py b/torch/ao/nn/quantized/dynamic/modules/linear.py
index a8a366e57f53cd..22f483f32fd7a8 100644
--- a/torch/ao/nn/quantized/dynamic/modules/linear.py
+++ b/torch/ao/nn/quantized/dynamic/modules/linear.py
@@ -105,7 +105,7 @@ def from_float(cls, mod):
             weight_observer = default_dynamic_qconfig.weight()
         dtype = weight_observer.dtype
         assert dtype in [torch.qint8, torch.float16], "The only supported dtypes for " \
-            f"dynamic quantized linear are qint8 and float16 got: {dtype}"
+            "dynamic quantized linear are qint8 and float16 got: {}".format(dtype)
         weight_observer(mod.weight)
         if dtype == torch.qint8:
             qweight = _quantize_weight(mod.weight.float(), weight_observer)
diff --git a/torch/ao/nn/quantized/dynamic/modules/rnn.py b/torch/ao/nn/quantized/dynamic/modules/rnn.py
index f8c68c841ef290..af50d51bb37099 100644
--- a/torch/ao/nn/quantized/dynamic/modules/rnn.py
+++ b/torch/ao/nn/quantized/dynamic/modules/rnn.py
@@ -93,8 +93,8 @@ def __init__(self, mode, input_size, hidden_size,
         if dropout > 0 and num_layers == 1:  # type: ignore[operator]
             warnings.warn("dropout option adds dropout after all but last "
                           "recurrent layer, so non-zero dropout expects "
-                          f"num_layers greater than 1, but got dropout={dropout} and "
-                          f"num_layers={num_layers}")
+                          "num_layers greater than 1, but got dropout={} and "
+                          "num_layers={}".format(dropout, num_layers))
 
         if mode == 'LSTM':
             gate_size = 4 * hidden_size
diff --git a/torch/ao/quantization/fx/convert.py b/torch/ao/quantization/fx/convert.py
index 917735529b7dde..a85036f51be1d4 100644
--- a/torch/ao/quantization/fx/convert.py
+++ b/torch/ao/quantization/fx/convert.py
@@ -977,8 +977,8 @@ def convert(
             assert k in convert_node_name_to_qconfig, f'Expected key {k} in convert node_name_to_qconfig'
             if convert_node_name_to_qconfig[k] is not None:
                 assert qconfig_equals(v, convert_node_name_to_qconfig[k]), \
-                    f"Expected k {k} to have the same value in prepare and convert QConfigMappings, " \
-                    f"but {v} was updated to {convert_node_name_to_qconfig[k]}"
+                    "Expected k {} to have the same value in prepare and convert QConfigMappings, " \
+                    "but {} was updated to {}".format(k, v, convert_node_name_to_qconfig[k])
         node_name_to_qconfig = convert_node_name_to_qconfig
 
     custom_module_classes = get_custom_module_class_keys(convert_custom_config.observed_to_quantized_mapping)
diff --git a/torch/ao/quantization/fx/prepare.py b/torch/ao/quantization/fx/prepare.py
index abdfba6ba1f6e0..12a29ebbb86e85 100644
--- a/torch/ao/quantization/fx/prepare.py
+++ b/torch/ao/quantization/fx/prepare.py
@@ -189,7 +189,7 @@ def _create_obs_or_fq_from_qspec(
         edge_or_node = quantization_spec.edge_or_node
         assert edge_or_node in obs_or_fq_map, \
             "please make sure only refer to edge or node that has " \
-            f"observer/fake_quant inserted: '{edge_or_node}' not in\n{obs_or_fq_map.keys()}"
+            "observer/fake_quant inserted: '{}' not in\n{}".format(edge_or_node, obs_or_fq_map.keys())
         return obs_or_fq_map[edge_or_node]
     elif isinstance(quantization_spec, DerivedQuantizationSpec):
         # can't use asdict, so not calling get_observer_kwargs here
diff --git a/torch/ao/quantization/pt2e/qat_utils.py b/torch/ao/quantization/pt2e/qat_utils.py
index ee4982942d93fb..837e19b16028f3 100644
--- a/torch/ao/quantization/pt2e/qat_utils.py
+++ b/torch/ao/quantization/pt2e/qat_utils.py
@@ -226,10 +226,10 @@ def _quantized_qat_conv2d_bn_pattern(
             )
         else:
             scaled_weight = torch.ops.quantized_decomposed.quantize_per_tensor(
-                scaled_weight, 1.0, 0, weight_quant_min, weight_quant_max, torch.int8,
+                scaled_weight, 1.0, int(0), weight_quant_min, weight_quant_max, torch.int8,
             )
             scaled_weight = torch.ops.quantized_decomposed.dequantize_per_tensor(
-                scaled_weight, 1.0, 0, weight_quant_min, weight_quant_max, torch.int8,
+                scaled_weight, 1.0, int(0), weight_quant_min, weight_quant_max, torch.int8,
             )
         if has_bias:
             zero_bias = torch.zeros_like(kwargs["conv_bias"], dtype=x.dtype)
@@ -283,10 +283,10 @@ def _folded_quantized_qat_conv2d_bn_pattern(
             )
         else:
             conv_weight = torch.ops.quantized_decomposed.quantize_per_tensor(
-                conv_weight, 1.0, 0, weight_quant_min, weight_quant_max, torch.int8,
+                conv_weight, 1.0, int(0), weight_quant_min, weight_quant_max, torch.int8,
             )
             conv_weight = torch.ops.quantized_decomposed.dequantize_per_tensor(
-                conv_weight, 1.0, 0, weight_quant_min, weight_quant_max, torch.int8,
+                conv_weight, 1.0, int(0), weight_quant_min, weight_quant_max, torch.int8,
             )
         if has_bias:
             x = F.conv2d(x, conv_weight, kwargs["conv_bias"])
diff --git a/torch/ao/quantization/utils.py b/torch/ao/quantization/utils.py
index 308fce44a40c6f..154372ef9a8597 100644
--- a/torch/ao/quantization/utils.py
+++ b/torch/ao/quantization/utils.py
@@ -201,7 +201,7 @@ def get_swapped_custom_module_class(custom_module, custom_module_class_mapping,
     quant_type = get_quant_type(qconfig)
     class_mapping = custom_module_class_mapping.get(quant_type, {})
     assert type(custom_module) in class_mapping, "did not find corresponding observed " \
-        f"module class for {type(custom_module)} in mapping: {class_mapping}"
+        "module class for {} in mapping: {}".format(type(custom_module), class_mapping)
     return class_mapping[type(custom_module)]
 
 def activation_dtype(qconfig):
@@ -298,8 +298,8 @@ def get_quant_type(qconfig):
         elif activation.dtype == torch.float16:
             return QuantType.STATIC
 
-    raise Exception(f"Unrecognized dtype combination in get_quant_type: activation({activation.dtype}),"
-                    f"weight({weight.dtype})")
+    raise Exception("Unrecognized dtype combination in get_quant_type: activation({}),"
+                    "weight({})".format(activation.dtype, weight.dtype))
 
 def check_min_max_valid(min_val: torch.Tensor, max_val: torch.Tensor) -> bool:
     """ Checks if the given minimum and maximum values are valid, meaning that
diff --git a/torch/autograd/functional.py b/torch/autograd/functional.py
index 30045bc8671cf5..755494a88ade4c 100644
--- a/torch/autograd/functional.py
+++ b/torch/autograd/functional.py
@@ -33,13 +33,17 @@ def _as_tuple(inp, arg_name=None, fn_name=None):
         if not isinstance(el, torch.Tensor):
             if is_inp_tuple:
                 raise TypeError(
-                    f"The {arg_name} given to {fn_name} must be either a Tensor or a tuple of Tensors but the"
-                    f" value at index {i} has type {type(el)}."
+                    "The {} given to {} must be either a Tensor or a tuple of Tensors but the"
+                    " value at index {} has type {}.".format(
+                        arg_name, fn_name, i, type(el)
+                    )
                 )
             else:
                 raise TypeError(
-                    f"The {arg_name} given to {fn_name} must be either a Tensor or a tuple of Tensors but the"
-                    f" given {arg_name} has type {type(el)}."
+                    "The {} given to {} must be either a Tensor or a tuple of Tensors but the"
+                    " given {} has type {}.".format(
+                        arg_name, fn_name, arg_name, type(el)
+                    )
                 )
 
     return is_inp_tuple, inp
@@ -130,35 +134,37 @@ def _check_requires_grad(inputs, input_type, strict):
         if inp is None:
             # This can only be reached for grad_inputs.
             raise RuntimeError(
-                f"The output of the user-provided function is independent of input {i}."
-                " This is not allowed in strict mode."
+                "The output of the user-provided function is independent of input {}."
+                " This is not allowed in strict mode.".format(i)
             )
         if not inp.requires_grad:
             if input_type == "hessian":
                 raise RuntimeError(
-                    f"The hessian of the user-provided function with respect to input {i}"
+                    "The hessian of the user-provided function with respect to input {}"
                     " is independent of the input. This is not allowed in strict mode."
                     " You should ensure that your function is thrice differentiable and that"
-                    " the hessian depends on the inputs."
+                    " the hessian depends on the inputs.".format(i)
                 )
             elif input_type == "jacobian":
                 raise RuntimeError(
                     "While computing the hessian, found that the jacobian of the user-provided"
-                    f" function with respect to input {i} is independent of the input. This is not"
+                    " function with respect to input {} is independent of the input. This is not"
                     " allowed in strict mode. You should ensure that your function is twice"
                     " differentiable and that the jacobian depends on the inputs (this would be"
-                    " violated by a linear function for example)."
+                    " violated by a linear function for example).".format(i)
                 )
             elif input_type == "grad_inputs":
                 raise RuntimeError(
-                    f"The gradient with respect to input {i} is independent of the inputs of the"
-                    " user-provided function. This is not allowed in strict mode."
+                    "The gradient with respect to input {} is independent of the inputs of the"
+                    " user-provided function. This is not allowed in strict mode.".format(
+                        i
+                    )
                 )
             else:
                 raise RuntimeError(
-                    f"Output {i} of the user-provided function does not require gradients."
+                    "Output {} of the user-provided function does not require gradients."
                     " The outputs must be computed in a differentiable manner from the input"
-                    " when running in strict mode."
+                    " when running in strict mode.".format(i)
                 )
 
 
@@ -215,25 +221,27 @@ def _fill_in_zeros(grads, refs, strict, create_graph, stage):
                 if stage == "back":
                     raise RuntimeError(
                         "The output of the user-provided function is independent of "
-                        f"input {i}. This is not allowed in strict mode."
+                        "input {}. This is not allowed in strict mode.".format(i)
                     )
                 elif stage == "back_trick":
                     raise RuntimeError(
-                        f"The gradient with respect to the input is independent of entry {i}"
+                        "The gradient with respect to the input is independent of entry {}"
                         " in the grad_outputs when using the double backward trick to compute"
-                        " forward mode gradients. This is not allowed in strict mode."
+                        " forward mode gradients. This is not allowed in strict mode.".format(
+                            i
+                        )
                     )
                 elif stage == "double_back":
                     raise RuntimeError(
                         "The jacobian of the user-provided function is independent of "
-                        f"input {i}. This is not allowed in strict mode."
+                        "input {}. This is not allowed in strict mode.".format(i)
                     )
                 else:
                     raise RuntimeError(
                         "The hessian of the user-provided function is independent of "
-                        f"entry {i} in the grad_jacobian. This is not allowed in strict "
+                        "entry {} in the grad_jacobian. This is not allowed in strict "
                         "mode as it prevents from using the double backward trick to "
-                        "replace forward mode AD."
+                        "replace forward mode AD.".format(i)
                     )
 
             grads_i = torch.zeros_like(refs[i])
@@ -242,12 +250,16 @@ def _fill_in_zeros(grads, refs, strict, create_graph, stage):
                 if "double" not in stage:
                     raise RuntimeError(
                         "The jacobian of the user-provided function is independent of "
-                        f"input {i}. This is not allowed in strict mode when create_graph=True."
+                        "input {}. This is not allowed in strict mode when create_graph=True.".format(
+                            i
+                        )
                     )
                 else:
                     raise RuntimeError(
                         "The hessian of the user-provided function is independent of "
-                        f"input {i}. This is not allowed in strict mode when create_graph=True."
+                        "input {}. This is not allowed in strict mode when create_graph=True.".format(
+                            i
+                        )
                     )
 
         res += (grads_i,)
@@ -799,17 +811,17 @@ def vjp(grad_output):
                         if strict and create_graph and not vj_el.requires_grad:
                             msg = (
                                 "The jacobian of the user-provided function is "
-                                f"independent of input {i}. This is not allowed in "
-                                "strict mode when create_graph=True."
+                                "independent of input {}. This is not allowed in "
+                                "strict mode when create_graph=True.".format(i)
                             )
                             raise RuntimeError(msg)
                         jac_i_el.append(vj_el)
                     else:
                         if strict:
                             msg = (
-                                f"Output {i} of the user-provided function is "
-                                f"independent of input {el_idx}. This is not allowed in "
-                                "strict mode."
+                                "Output {} of the user-provided function is "
+                                "independent of input {}. This is not allowed in "
+                                "strict mode.".format(i, el_idx)
                             )
                             raise RuntimeError(msg)
                         jac_i_el.append(torch.zeros_like(inp_el))
diff --git a/torch/autograd/profiler_util.py b/torch/autograd/profiler_util.py
index de330f10a4fe21..a6944d6a4840e5 100644
--- a/torch/autograd/profiler_util.py
+++ b/torch/autograd/profiler_util.py
@@ -246,14 +246,20 @@ def export_chrome_trace(self, path):
                     # 's' and 'f' draw Flow arrows from
                     # the CPU launch to the GPU kernel
                     f.write(
-                        f'{{"name": "{evt.trace_name}", '
+                        '{{"name": "{}", '
                         '"ph": "s", '
-                        f'"ts": {evt.time_range.start}, '
-                        f'"tid": {evt.thread}, '
+                        '"ts": {}, '
+                        '"tid": {}, '
                         '"pid": "CPU functions", '
-                        f'"id": {next_id}, '
-                        f'"cat": "cpu_to_{device_name}", '
-                        '"args": {{}}}}, '
+                        '"id": {}, '
+                        '"cat": "cpu_to_{}", '
+                        '"args": {{}}}}, '.format(
+                            evt.trace_name,
+                            evt.time_range.start,
+                            evt.thread,
+                            next_id,
+                            device_name,
+                        )
                     )
                     # Note: use torch.profiler to get device kernel trace
                     next_id += 1
diff --git a/torch/backends/cuda/__init__.py b/torch/backends/cuda/__init__.py
index 54ecd2681387a7..ab63b15f32e074 100644
--- a/torch/backends/cuda/__init__.py
+++ b/torch/backends/cuda/__init__.py
@@ -94,8 +94,10 @@ def __getitem__(self, device):
         index = torch.cuda._utils._get_device_index(device)
         if index < 0 or index >= torch.cuda.device_count():
             raise RuntimeError(
-                f"cufft_plan_cache: expected 0 <= device index < {torch.cuda.device_count()}, but got "
-                f"device with index {index}"
+                (
+                    "cufft_plan_cache: expected 0 <= device index < {}, but got "
+                    "device with index {}"
+                ).format(torch.cuda.device_count(), index)
             )
         if len(self.caches) == 0:
             self.caches.extend(
diff --git a/torch/distributed/distributed_c10d.py b/torch/distributed/distributed_c10d.py
index 59e5a71f76c7fa..8269caec4421fa 100644
--- a/torch/distributed/distributed_c10d.py
+++ b/torch/distributed/distributed_c10d.py
@@ -1123,9 +1123,9 @@ def init_process_group(
     if backend == Backend.MPI:
         if world_size != -1 or rank != -1:
             warnings.warn(
-                f"For MPI backend, world_size ({world_size}) and rank ({rank}) "
+                "For MPI backend, world_size ({}) and rank ({}) "
                 "are ignored since they are assigned by the "
-                "MPI runtime."
+                "MPI runtime.".format(world_size, rank)
             )
 
         default_pg, _ = _new_process_group_helper(
diff --git a/torch/distributed/nn/api/remote_module.py b/torch/distributed/nn/api/remote_module.py
index 935e27436ea28d..75d7e6215d5620 100644
--- a/torch/distributed/nn/api/remote_module.py
+++ b/torch/distributed/nn/api/remote_module.py
@@ -500,8 +500,8 @@ def _check_attribute_picklability(self):
                 and k not in _REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING
             ):
                 raise AttributeError(
-                    f"Attribute {k} must be either in ``_REMOTE_MODULE_PICKLED_ATTRIBUTES`` or "
-                    "``_REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING``."
+                    "Attribute {} must be either in ``_REMOTE_MODULE_PICKLED_ATTRIBUTES`` or "
+                    "``_REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING``.".format(k)
                 )
 
     def _install_generated_methods(self):
@@ -729,9 +729,11 @@ def _remote_module_reducer(remote_module):
         # Check if unpickled attributes are all in _REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING.
         elif k not in _REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING:
             print(
-                f"The new attribute ``{k}`` of RemoteModule is ignored during RPC pickling. "
+                "The new attribute ``{}`` of RemoteModule is ignored during RPC pickling. "
                 "To pickle this attribute, please add it to ``_REMOTE_MODULE_PICKLED_ATTRIBUTES``. "
-                "Otherwise, please explicitly add it to ``_REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING``.",
+                "Otherwise, please explicitly add it to ``_REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING``.".format(
+                    k
+                ),
                 file=sys.stderr,
             )
 
diff --git a/torch/distributed/pipeline/sync/pipe.py b/torch/distributed/pipeline/sync/pipe.py
index 65063e9b1c8246..707f56c9d2eee9 100644
--- a/torch/distributed/pipeline/sync/pipe.py
+++ b/torch/distributed/pipeline/sync/pipe.py
@@ -112,8 +112,8 @@ def _retrieve_device(module: nn.Module) -> torch.device:
             device = parameter.device
         elif device != parameter.device:
             raise ValueError(
-                f'nn.Module: {module}, should have all parameters on a single device,'
-                ' please use .to() to place the module on a single device')
+                'nn.Module: {}, should have all parameters on a single device,'
+                ' please use .to() to place the module on a single device'.format(module))
 
     return device if device is not None else torch.device("cpu")
 
diff --git a/torch/distributions/independent.py b/torch/distributions/independent.py
index 35b705fd0f29c7..a58e81b7562e4b 100644
--- a/torch/distributions/independent.py
+++ b/torch/distributions/independent.py
@@ -45,7 +45,9 @@ def __init__(
         if reinterpreted_batch_ndims > len(base_distribution.batch_shape):
             raise ValueError(
                 "Expected reinterpreted_batch_ndims <= len(base_distribution.batch_shape), "
-                f"actual {reinterpreted_batch_ndims} vs {len(base_distribution.batch_shape)}"
+                "actual {} vs {}".format(
+                    reinterpreted_batch_ndims, len(base_distribution.batch_shape)
+                )
             )
         shape = base_distribution.batch_shape + base_distribution.event_shape
         event_dim = reinterpreted_batch_ndims + len(base_distribution.event_shape)
diff --git a/torch/fx/experimental/unification/multipledispatch/dispatcher.py b/torch/fx/experimental/unification/multipledispatch/dispatcher.py
index d2a8e6bfc7ffff..65ad70a3e1abb4 100644
--- a/torch/fx/experimental/unification/multipledispatch/dispatcher.py
+++ b/torch/fx/experimental/unification/multipledispatch/dispatcher.py
@@ -205,9 +205,9 @@ def add(self, signature, func):
             if not isinstance(typ, (type, list)):
                 str_sig = ', '.join(c.__name__ if isinstance(c, type)
                                     else str(c) for c in signature)
-                raise TypeError(f"Tried to dispatch on non-type: {typ}\n"
-                                f"In signature: <{str_sig}>\n"
-                                f"In function: {self.name}")
+                raise TypeError("Tried to dispatch on non-type: {}\n"
+                                "In signature: <{}>\n"
+                                "In function: {}".format(typ, str_sig, self.name))
 
             # handle variadic signatures
             if isinstance(typ, list):
@@ -272,7 +272,8 @@ def __call__(self, *args, **kwargs):
 
             raise NotImplementedError(
                 "Matching functions for "
-                f"{self.name}: <{str_signature(types)}> found, but none completed successfully",) from e
+                "{}: <{}> found, but none completed successfully".format(
+                    self.name, str_signature(types),),) from e
 
     def __str__(self):
         return f"<dispatched {self.name}>"
diff --git a/torch/jit/_recursive.py b/torch/jit/_recursive.py
index c7d10750377120..aa4e370e3b25d1 100644
--- a/torch/jit/_recursive.py
+++ b/torch/jit/_recursive.py
@@ -310,17 +310,17 @@ def infer_type(name, item):
                 )
 
             warnings.warn(
-                f"'{name}' was found in ScriptModule constants, "
-                f" but it is a non-constant {hint}. Consider removing it."
+                "'{}' was found in ScriptModule constants, "
+                " but it is a non-constant {}. Consider removing it.".format(name, hint)
             )
             continue
         if not hasattr(nn_module, name):
             # TODO: We should really error in this case, but its bc-breaking so
             # we need to warn for at least one release
             warnings.warn(
-                f"'{name}' was found in ScriptModule constants, "
+                "'{}' was found in ScriptModule constants, "
                 "but was not actually set in __init__. "
-                "Consider removing it."
+                "Consider removing it.".format(name)
             )
             continue
         value = getattr(nn_module, name)
@@ -370,8 +370,8 @@ def infer_type(name, item):
                 hint = (
                     "(This function exists as an attribute on the Python module, "
                     "but we failed to compile it to a TorchScript function. "
-                    f"\nThe error stack is reproduced here:\n{e}"
-                )
+                    "\nThe error stack is reproduced here:\n{}"
+                ).format(e)
                 concrete_type_builder.add_failed_attribute(name, hint)
                 pass
 
@@ -998,9 +998,9 @@ def try_compile_fn(fn, loc):
 
     if not inspect.isfunction(fn) and not inspect.ismethod(fn):
         raise RuntimeError(
-            f"`{fn}` is not a function. Recursive scripting only supports "
+            "`{}` is not a function. Recursive scripting only supports "
             "Python functions or methods currently.\n"
-            f"Consider manually annotating `{fn}` with @torch.jit.script."
+            "Consider manually annotating `{}` with @torch.jit.script.".format(fn, fn)
         )
 
     # We don't have the actual scope where the function was defined, but we can
diff --git a/torch/jit/_script.py b/torch/jit/_script.py
index acedc52786f709..bcfe27b9663c1d 100644
--- a/torch/jit/_script.py
+++ b/torch/jit/_script.py
@@ -257,7 +257,7 @@ def __setitem__(self, k, v):
         else:
             raise RuntimeError(
                 "Cannot re-assign modules in a ScriptModule with non-scripted "
-                f"module, tried to replace existing module '{k}': {v}"
+                "module, tried to replace existing module '{}': {}".format(k, v)
             )
 
     def __getitem__(self, k):
@@ -1402,7 +1402,7 @@ def _check_overload_defaults(impl_defaults, overload_defaults, loc):
                 loc,
                 "Default parameters on overloads do not affect the runtime so they "
                 "must equal to the default parameter on the implementation function. Found on "
-                f"parameter {name}",
+                "parameter {name}".format(name=name),
             )
 
 
@@ -1461,9 +1461,9 @@ def _check_directly_compile_overloaded(obj):
     qual_name = _qualified_name(obj)
     if _jit_internal._get_fn_overloads(qual_name) or _try_get_jit_cached_overloads(obj):
         raise RuntimeError(
-            f"Function {qual_name} cannot be directly compiled because it"
+            "Function {} cannot be directly compiled because it"
             " is overloaded. It must be used in a context of a function"
-            " where its inputs can determine which overload to call."
+            " where its inputs can determine which overload to call.".format(qual_name)
         )
 
 
diff --git a/torch/jit/_trace.py b/torch/jit/_trace.py
index 2eaa79cfe0bcfc..11e8de464aebc8 100644
--- a/torch/jit/_trace.py
+++ b/torch/jit/_trace.py
@@ -254,8 +254,10 @@ def run_fwd_bwd(args, force_trace=False, assert_compiled=False):
             out = (out,)
         if loss_fn == torch.sum and len(out) != 1:
             raise ValueError(
-                f"Model returns {len(out)} outputs, but default loss function "
-                "(torch.sum) can only handle a single output"
+                (
+                    "Model returns {} outputs, but default loss function "
+                    "(torch.sum) can only handle a single output"
+                ).format(len(out))
             )
         out_vars, _ = _flatten(out)
         saved_outs = [
diff --git a/torch/nn/functional.py b/torch/nn/functional.py
index 9abf844acce309..0aa729d92124aa 100644
--- a/torch/nn/functional.py
+++ b/torch/nn/functional.py
@@ -903,7 +903,9 @@ def _unpool_output_size(
         if len(output_size) != len(kernel_size):
             raise ValueError(
                 "output_size should be a sequence containing "
-                f"{len(kernel_size)} or {len(kernel_size) + 2} elements, but it has a length of '{len(output_size)}'"
+                "{} or {} elements, but it has a length of '{}'".format(
+                    len(kernel_size), len(kernel_size) + 2, len(output_size)
+                )
             )
         for d in range(len(kernel_size)):
             min_size = default_size[d] - stride[d]
@@ -2354,8 +2356,8 @@ def embedding_bag(
 
     if per_sample_weights is not None and input.size() != per_sample_weights.size():
         raise ValueError(
-            f"embedding_bag: If per_sample_weights ({per_sample_weights.shape}) is not None, "
-            f"then it must have the same shape as the input ({input.shape})"
+            "embedding_bag: If per_sample_weights ({}) is not None, "
+            "then it must have the same shape as the input ({})".format(per_sample_weights.shape, input.shape)
         )
 
     if not weight.dim() == 2:
@@ -2373,7 +2375,7 @@ def embedding_bag(
                 "if input is 2D, then offsets has to be None"
                 ", as input is treated is a mini-batch of"
                 " fixed length sequences. However, found "
-                f"offsets of type {type_str}"
+                "offsets of type {}".format(type_str)
             )
         offsets = torch.arange(0, input.numel(), input.size(1), dtype=input.dtype, device=input.device)
 
@@ -2414,7 +2416,7 @@ def embedding_bag(
         raise NotImplementedError(
             "embedding_bag: per_sample_weights was not None. "
             "per_sample_weights is only supported for mode='sum' "
-            f"(got mode='{mode}'). Please open a feature request on GitHub."
+            "(got mode='{}'). Please open a feature request on GitHub.".format(mode)
         )
 
     ret, _, _, _ = torch.embedding_bag(
@@ -3221,9 +3223,9 @@ def smooth_l1_loss(
         )
     if not (target.size() == input.size()):
         warnings.warn(
-            f"Using a target size ({target.size()}) that is different to the input size ({input.size()}). "
+            "Using a target size ({}) that is different to the input size ({}). "
             "This will likely lead to incorrect results due to broadcasting. "
-            "Please ensure they have the same size.",
+            "Please ensure they have the same size.".format(target.size(), input.size()),
             stacklevel=2,
         )
     if size_average is not None or reduce is not None:
@@ -3258,9 +3260,9 @@ def huber_loss(
             delta=delta,
         )
     if not (target.size() == input.size()):
-        warnings.warn(f"Using a target size ({target.size()}) that is different to the input size ({input.size()}). "
+        warnings.warn("Using a target size ({}) that is different to the input size ({}). "
                       "This will likely lead to incorrect results due to broadcasting. "
-                      "Please ensure they have the same size.",
+                      "Please ensure they have the same size.".format(target.size(), input.size()),
                       stacklevel=2)
 
     expanded_input, expanded_target = torch.broadcast_tensors(input, target)
@@ -3286,9 +3288,9 @@ def l1_loss(
         )
     if not (target.size() == input.size()):
         warnings.warn(
-            f"Using a target size ({target.size()}) that is different to the input size ({input.size()}). "
+            "Using a target size ({}) that is different to the input size ({}). "
             "This will likely lead to incorrect results due to broadcasting. "
-            "Please ensure they have the same size.",
+            "Please ensure they have the same size.".format(target.size(), input.size()),
             stacklevel=2,
         )
     if size_average is not None or reduce is not None:
@@ -3317,9 +3319,9 @@ def mse_loss(
         )
     if not (target.size() == input.size()):
         warnings.warn(
-            f"Using a target size ({target.size()}) that is different to the input size ({input.size()}). "
+            "Using a target size ({}) that is different to the input size ({}). "
             "This will likely lead to incorrect results due to broadcasting. "
-            "Please ensure they have the same size.",
+            "Please ensure they have the same size.".format(target.size(), input.size()),
             stacklevel=2,
         )
     if size_average is not None or reduce is not None:
@@ -4042,8 +4044,8 @@ def interpolate(input: Tensor, size: Optional[int] = None, scale_factor: Optiona
 
     raise NotImplementedError(
         "Input Error: Only 3D, 4D and 5D input Tensors supported"
-        f" (got {input.dim()}D) for the modes: nearest | linear | bilinear | bicubic | trilinear | area | nearest-exact"
-        f" (got {mode})"
+        " (got {}D) for the modes: nearest | linear | bilinear | bicubic | trilinear | area | nearest-exact"
+        " (got {})".format(input.dim(), mode)
     )
 
 
@@ -4275,7 +4277,7 @@ def grid_sample(
         raise ValueError(
             "nn.functional.grid_sample(): expected padding_mode "
             "to be 'zeros', 'border', or 'reflection', "
-            f"but got: '{padding_mode}'"
+            "but got: '{}'".format(padding_mode)
         )
 
     if mode == "bilinear":
@@ -4383,7 +4385,7 @@ def affine_grid(theta: Tensor, size: List[int], align_corners: Optional[bool] =
         raise NotImplementedError(
             "affine_grid only supports 4D and 5D sizes, "
             "for 2D and 3D affine transforms, respectively. "
-            f"Got size {size}."
+            "Got size {}.".format(size)
         )
     # check for empty span
     if align_corners and min(spatial_size) == 1:
diff --git a/torch/nn/init.py b/torch/nn/init.py
index 31e78442796b92..21768a65a9304f 100644
--- a/torch/nn/init.py
+++ b/torch/nn/init.py
@@ -543,13 +543,14 @@ def deprecated_init(*args, **kwargs):
         warnings.warn(f"nn.init.{old_name} is now deprecated in favor of nn.init.{new_name}.", stacklevel=2)
         return meth(*args, **kwargs)
 
-    deprecated_init.__doc__ = fr"""
+    deprecated_init.__doc__ = r"""
     {old_name}(...)
 
     .. warning::
         This method is now deprecated in favor of :func:`torch.nn.init.{new_name}`.
 
-    See :func:`~torch.nn.init.{new_name}` for details."""
+    See :func:`~torch.nn.init.{new_name}` for details.""".format(
+        old_name=old_name, new_name=new_name)
     deprecated_init.__name__ = old_name
     return deprecated_init
 
diff --git a/torch/nn/modules/adaptive.py b/torch/nn/modules/adaptive.py
index dd5539bec33020..cf2f56efa8178d 100644
--- a/torch/nn/modules/adaptive.py
+++ b/torch/nn/modules/adaptive.py
@@ -224,9 +224,11 @@ def forward(self, input_: Tensor, target_: Tensor) -> _ASMoutput:
             used_rows += row_indices.numel()
 
         if used_rows != batch_size:
-            raise RuntimeError(f"Target values should be in [0, {self.n_classes - 1}], "
-                               f"but values in range [{target.min().item()}, {target.max().item()}] "
-                               "were found. ")
+            raise RuntimeError("Target values should be in [0, {}], "
+                               "but values in range [{}, {}] "
+                               "were found. ".format(self.n_classes - 1,
+                                                     target.min().item(),
+                                                     target.max().item()))
 
         head_output = self.head(input)
         head_logprob = log_softmax(head_output, dim=1)
diff --git a/torch/nn/modules/container.py b/torch/nn/modules/container.py
index 2b8cd1bab3b267..bbd253530e2c43 100644
--- a/torch/nn/modules/container.py
+++ b/torch/nn/modules/container.py
@@ -148,7 +148,8 @@ def __add__(self, other) -> 'Sequential':
             return ret
         else:
             raise ValueError('add operator supports only objects '
-                             f'of Sequential class, but {str(type(other))} is given.')
+                             'of Sequential class, but {} is given.'.format(
+                                 str(type(other))))
 
     def pop(self, key: Union[int, slice]) -> Module:
         v = self[key]
@@ -163,7 +164,8 @@ def __iadd__(self, other) -> 'Sequential':
             return self
         else:
             raise ValueError('add operator supports only objects '
-                             f'of Sequential class, but {str(type(other))} is given.')
+                             'of Sequential class, but {} is given.'.format(
+                                 str(type(other))))
 
     def __mul__(self, other: int) -> 'Sequential':
         if not isinstance(other, int):
diff --git a/torch/nn/modules/conv.py b/torch/nn/modules/conv.py
index cb018d98dc20ef..9fb1135eb9d6ab 100644
--- a/torch/nn/modules/conv.py
+++ b/torch/nn/modules/conv.py
@@ -656,9 +656,10 @@ def _output_padding(self, input: Tensor, output_size: Optional[List[int]],
                 min_size = min_sizes[i]
                 max_size = max_sizes[i]
                 if size < min_size or size > max_size:
-                    raise ValueError(
-                        f"requested an output size of {output_size}, but valid sizes range "
-                        f"from {min_sizes} to {max_sizes} (for an input of {input.size()[2:]})")
+                    raise ValueError((
+                        "requested an output size of {}, but valid sizes range "
+                        "from {} to {} (for an input of {})").format(
+                            output_size, min_sizes, max_sizes, input.size()[2:]))
 
             res = torch.jit.annotate(List[int], [])
             for d in range(num_spatial_dims):
diff --git a/torch/nn/modules/module.py b/torch/nn/modules/module.py
index 8fad740545008e..a0255076c5613a 100644
--- a/torch/nn/modules/module.py
+++ b/torch/nn/modules/module.py
@@ -461,8 +461,8 @@ def __init__(self, *args, **kwargs) -> None:
                             "".format(type(self).__name__, next(iter(kwargs))))
 
         if self.call_super_init is False and bool(args):
-            raise TypeError(f"{type(self).__name__}.__init__() takes 1 positional argument but {len(args) + 1} were"
-                            " given")
+            raise TypeError("{}.__init__() takes 1 positional argument but {} were"
+                            " given".format(type(self).__name__, len(args) + 1))
 
         """
         Calls super().__setattr__('a', a) instead of the typical self.a = a
@@ -537,9 +537,9 @@ def register_buffer(self, name: str, tensor: Optional[Tensor], persistent: bool
         elif hasattr(self, name) and name not in self._buffers:
             raise KeyError(f"attribute '{name}' already exists")
         elif tensor is not None and not isinstance(tensor, torch.Tensor):
-            raise TypeError(f"cannot assign '{torch.typename(tensor)}' object to buffer '{name}' "
+            raise TypeError("cannot assign '{}' object to buffer '{}' "
                             "(torch Tensor or None required)"
-                            )
+                            .format(torch.typename(tensor), name))
         else:
             for hook in _global_buffer_registration_hooks.values():
                 output = hook(self, name, tensor)
@@ -580,15 +580,15 @@ def register_parameter(self, name: str, param: Optional[Parameter]) -> None:
         if param is None:
             self._parameters[name] = None
         elif not isinstance(param, Parameter):
-            raise TypeError(f"cannot assign '{torch.typename(param)}' object to parameter '{name}' "
+            raise TypeError("cannot assign '{}' object to parameter '{}' "
                             "(torch.nn.Parameter or None required)"
-                            )
+                            .format(torch.typename(param), name))
         elif param.grad_fn:
             raise ValueError(
-                f"Cannot assign non-leaf Tensor to parameter '{name}'. Model "
-                f"parameters must be created explicitly. To express '{name}' "
+                "Cannot assign non-leaf Tensor to parameter '{0}'. Model "
+                "parameters must be created explicitly. To express '{0}' "
                 "as a function of another Tensor, compute the value in "
-                "the forward() method.")
+                "the forward() method.".format(name))
         else:
             for hook in _global_parameter_registration_hooks.values():
                 output = hook(self, name, param)
@@ -1143,7 +1143,7 @@ def to(self, *args, **kwargs):
         if dtype is not None:
             if not (dtype.is_floating_point or dtype.is_complex):
                 raise TypeError('nn.Module.to only accepts floating point or complex '
-                                f'dtypes, but got desired dtype={dtype}')
+                                'dtypes, but got desired dtype={}'.format(dtype))
             if dtype.is_complex:
                 warnings.warn(
                     "Complex modules are a new feature under active development whose design may change, "
@@ -1712,9 +1712,9 @@ def remove_from(*dicts_or_sets):
             self.register_parameter(name, value)
         elif params is not None and name in params:
             if value is not None:
-                raise TypeError(f"cannot assign '{torch.typename(value)}' as parameter '{name}' "
+                raise TypeError("cannot assign '{}' as parameter '{}' "
                                 "(torch.nn.Parameter or None expected)"
-                                )
+                                .format(torch.typename(value), name))
             self.register_parameter(name, value)
         else:
             modules = self.__dict__.get('_modules')
@@ -1730,9 +1730,9 @@ def remove_from(*dicts_or_sets):
                 modules[name] = value
             elif modules is not None and name in modules:
                 if value is not None:
-                    raise TypeError(f"cannot assign '{torch.typename(value)}' as child module '{name}' "
+                    raise TypeError("cannot assign '{}' as child module '{}' "
                                     "(torch.nn.Module or None expected)"
-                                    )
+                                    .format(torch.typename(value), name))
                 for hook in _global_module_registration_hooks.values():
                     output = hook(self, name, value)
                     if output is not None:
@@ -1742,9 +1742,9 @@ def remove_from(*dicts_or_sets):
                 buffers = self.__dict__.get('_buffers')
                 if buffers is not None and name in buffers:
                     if value is not None and not isinstance(value, torch.Tensor):
-                        raise TypeError(f"cannot assign '{torch.typename(value)}' as buffer '{name}' "
+                        raise TypeError("cannot assign '{}' as buffer '{}' "
                                         "(torch.Tensor or None expected)"
-                                        )
+                                        .format(torch.typename(value), name))
                     for hook in _global_buffer_registration_hooks.values():
                         output = hook(self, name, value)
                         if output is not None:
@@ -2000,10 +2000,10 @@ def _load_from_state_dict(self, state_dict, prefix, local_metadata, strict,
             if key in state_dict:
                 input_param = state_dict[key]
                 if not torch.overrides.is_tensor_like(input_param):
-                    error_msgs.append(f'While copying the parameter named "{key}", '
+                    error_msgs.append('While copying the parameter named "{}", '
                                       'expected torch.Tensor or Tensor-like object from checkpoint but '
-                                      f'received {type(input_param)}'
-                                      )
+                                      'received {}'
+                                      .format(key, type(input_param)))
                     continue
 
                 # This is used to avoid copying uninitialized parameters into
@@ -2039,11 +2039,11 @@ def _load_from_state_dict(self, state_dict, prefix, local_metadata, strict,
                         else:
                             param.copy_(input_param)
                 except Exception as ex:
-                    error_msgs.append(f'While copying the parameter named "{key}", '
-                                      f'whose dimensions in the model are {param.size()} and '
-                                      f'whose dimensions in the checkpoint are {input_param.size()}, '
-                                      f'an exception occurred : {ex.args}.'
-                                      )
+                    error_msgs.append('While copying the parameter named "{}", '
+                                      'whose dimensions in the model are {} and '
+                                      'whose dimensions in the checkpoint are {}, '
+                                      'an exception occurred : {}.'
+                                      .format(key, param.size(), input_param.size(), ex.args))
             elif strict:
                 missing_keys.append(key)
 
diff --git a/torch/nn/modules/rnn.py b/torch/nn/modules/rnn.py
index b49d804fb45ae6..1ea44a6e4d7379 100644
--- a/torch/nn/modules/rnn.py
+++ b/torch/nn/modules/rnn.py
@@ -70,8 +70,8 @@ def __init__(self, mode: str, input_size: int, hidden_size: int,
         if dropout > 0 and num_layers == 1:
             warnings.warn("dropout option adds dropout after all but last "
                           "recurrent layer, so non-zero dropout expects "
-                          f"num_layers greater than 1, but got dropout={dropout} and "
-                          f"num_layers={num_layers}")
+                          "num_layers greater than 1, but got dropout={} and "
+                          "num_layers={}".format(dropout, num_layers))
 
         if not isinstance(hidden_size, int):
             raise TypeError(f"hidden_size should be of type int, got: {type(hidden_size).__name__}")
diff --git a/torch/nn/parallel/data_parallel.py b/torch/nn/parallel/data_parallel.py
index 2551d43fdc9e1e..40253c4b43c218 100644
--- a/torch/nn/parallel/data_parallel.py
+++ b/torch/nn/parallel/data_parallel.py
@@ -169,8 +169,8 @@ def forward(self, *inputs: Any, **kwargs: Any) -> Any:
             for t in chain(self.module.parameters(), self.module.buffers()):
                 if t.device != self.src_device_obj:
                     raise RuntimeError("module must have its parameters and buffers "
-                                       f"on device {self.src_device_obj} (device_ids[0]) but found one of "
-                                       f"them on device: {t.device}")
+                                       "on device {} (device_ids[0]) but found one of "
+                                       "them on device: {}".format(self.src_device_obj, t.device))
 
             inputs, module_kwargs = self.scatter(inputs, kwargs, self.device_ids)
             # for forward function without any inputs, empty list and dict will be created
@@ -249,8 +249,8 @@ def data_parallel(
     for t in chain(module.parameters(), module.buffers()):
         if t.device != src_device_obj:
             raise RuntimeError("module must have its parameters and buffers "
-                               f"on device {src_device_obj} (device_ids[0]) but found one of "
-                               f"them on device: {t.device}")
+                               "on device {} (device_ids[0]) but found one of "
+                               "them on device: {}".format(src_device_obj, t.device))
 
     inputs, module_kwargs = scatter_kwargs(inputs, module_kwargs, device_ids, dim)
     # for module without any inputs, empty list and dict will be created
diff --git a/torch/nn/parallel/distributed.py b/torch/nn/parallel/distributed.py
index 24f63e18a29ee0..d72346f2376019 100644
--- a/torch/nn/parallel/distributed.py
+++ b/torch/nn/parallel/distributed.py
@@ -697,7 +697,9 @@ def __init__(
             self._log_and_throw(
                 ValueError,
                 "DistributedDataParallel's input module must be on "
-                f"the same type of devices, but input module parameters locate in {distinct_device_types}.",
+                "the same type of devices, but input module parameters locate in {}.".format(
+                    distinct_device_types
+                ),
             )
 
         self.device_type = list(distinct_device_types)[0]
diff --git a/torch/nn/parallel/parallel_apply.py b/torch/nn/parallel/parallel_apply.py
index 36de8845e56b49..fc14a968620dca 100644
--- a/torch/nn/parallel/parallel_apply.py
+++ b/torch/nn/parallel/parallel_apply.py
@@ -71,8 +71,8 @@ def _worker(
             if t is None:
                 with lock:
                     results[i] = ExceptionWrapper(
-                        where=f"in replica {i}, no device was provided and no tensor input was found; "
-                        "device cannot be resolved")
+                        where="in replica {}, no device was provided and no tensor input was found; "
+                        "device cannot be resolved".format(i))
                 return
             device = t.get_device()
         if stream is None:
diff --git a/torch/nn/parameter.py b/torch/nn/parameter.py
index c8716b6c20d3b2..c15ad0c863c94b 100644
--- a/torch/nn/parameter.py
+++ b/torch/nn/parameter.py
@@ -154,11 +154,11 @@ def __torch_function__(cls, func, types, args=(), kwargs=None):
                 kwargs = {}
             return super().__torch_function__(func, types, args, kwargs)
         raise ValueError(
-            f'Attempted to use an uninitialized parameter in {func}. '
+            'Attempted to use an uninitialized parameter in {}. '
             'This error happens when you are using a `LazyModule` or '
-            f'explicitly manipulating `torch.nn.parameter.{cls.__name__}` '
+            'explicitly manipulating `torch.nn.parameter.{}` '
             'objects. When using LazyModules Call `forward` with a dummy batch '
-            'to initialize the parameters before calling torch functions')
+            'to initialize the parameters before calling torch functions'.format(func, cls.__name__))
 
 
 def is_lazy(param):
diff --git a/torch/nn/utils/parametrizations.py b/torch/nn/utils/parametrizations.py
index 40dc763269be33..c451be6dd7920c 100644
--- a/torch/nn/utils/parametrizations.py
+++ b/torch/nn/utils/parametrizations.py
@@ -379,7 +379,7 @@ def __init__(
 
         if n_power_iterations <= 0:
             raise ValueError('Expected n_power_iterations to be positive, but '
-                             f'got n_power_iterations={n_power_iterations}')
+                             'got n_power_iterations={}'.format(n_power_iterations))
         self.dim = dim if dim >= 0 else dim + ndim
         self.eps = eps
         if ndim > 1:
diff --git a/torch/nn/utils/prune.py b/torch/nn/utils/prune.py
index 15e244afaf1da8..1e16f11b6b626e 100644
--- a/torch/nn/utils/prune.py
+++ b/torch/nn/utils/prune.py
@@ -288,7 +288,9 @@ def add_pruning_method(self, method):
         elif method is not None and self._tensor_name != method._tensor_name:
             raise ValueError(
                 "Can only add pruning methods acting on "
-                f"the parameter named '{self._tensor_name}' to PruningContainer {self}."
+                "the parameter named '{}' to PruningContainer {}.".format(
+                    self._tensor_name, self
+                )
                 + f" Found '{method._tensor_name}'"
             )
         # if all checks passed, add to _pruning_methods tuple
@@ -1090,7 +1092,9 @@ def global_unstructured(parameters, pruning_method, importance_scores=None, **kw
     if method.PRUNING_TYPE != "unstructured":
         raise TypeError(
             'Only "unstructured" PRUNING_TYPE supported for '
-            f"the `pruning_method`. Found method {pruning_method} of type {method.PRUNING_TYPE}"
+            "the `pruning_method`. Found method {} of type {}".format(
+                pruning_method, method.PRUNING_TYPE
+            )
         )
 
     container.add_pruning_method(method)
@@ -1276,7 +1280,7 @@ def _validate_structured_pruning(t):
         raise ValueError(
             "Structured pruning can only be applied to "
             "multidimensional tensors. Found tensor of shape "
-            f"{shape} with {len(shape)} dims"
+            "{} with {} dims".format(shape, len(shape))
         )
 
 
diff --git a/torch/nn/utils/rnn.py b/torch/nn/utils/rnn.py
index fd7a12d933df1c..4f84cfb9d5a9a3 100644
--- a/torch/nn/utils/rnn.py
+++ b/torch/nn/utils/rnn.py
@@ -327,8 +327,8 @@ def pad_packed_sequence(
         if total_length < max_seq_length:
             raise ValueError("Expected total_length to be at least the length "
                              "of the longest sequence in input, but got "
-                             f"total_length={total_length} and max sequence length being {max_seq_length}"
-                             )
+                             "total_length={} and max sequence length being {}"
+                             .format(total_length, max_seq_length))
         max_seq_length = total_length
     padded_output, lengths = _VF._pad_packed_sequence(
         sequence.data, sequence.batch_sizes, batch_first, padding_value, max_seq_length)
diff --git a/torch/nn/utils/spectral_norm.py b/torch/nn/utils/spectral_norm.py
index 0dd5973abb6783..ea29d095ea1c0f 100644
--- a/torch/nn/utils/spectral_norm.py
+++ b/torch/nn/utils/spectral_norm.py
@@ -29,7 +29,7 @@ def __init__(self, name: str = 'weight', n_power_iterations: int = 1, dim: int =
         self.dim = dim
         if n_power_iterations <= 0:
             raise ValueError('Expected n_power_iterations to be positive, but '
-                             f'got n_power_iterations={n_power_iterations}')
+                             'got n_power_iterations={}'.format(n_power_iterations))
         self.n_power_iterations = n_power_iterations
         self.eps = eps
 
diff --git a/torch/optim/lr_scheduler.py b/torch/optim/lr_scheduler.py
index 337537ebe5a514..d78fdbf18580b1 100644
--- a/torch/optim/lr_scheduler.py
+++ b/torch/optim/lr_scheduler.py
@@ -40,7 +40,7 @@ def __init__(self, optimizer, last_epoch=-1, verbose=False):
             for i, group in enumerate(optimizer.param_groups):
                 if 'initial_lr' not in group:
                     raise KeyError("param 'initial_lr' is not specified "
-                                   f"in param_groups[{i}] when resuming an optimizer")
+                                   "in param_groups[{}] when resuming an optimizer".format(i))
         self.base_lrs = [group['initial_lr'] for group in optimizer.param_groups]
         self.last_epoch = last_epoch
 
@@ -645,8 +645,8 @@ def __init__(self, optimizer, schedulers, milestones, last_epoch=-1, verbose=Fal
         if (len(milestones) != len(schedulers) - 1):
             raise ValueError(
                 "Sequential Schedulers expects number of schedulers provided to be one more "
-                f"than the number of milestone points, but got number of schedulers {len(schedulers)} and the "
-                f"number of milestones to be equal to {len(milestones)}"
+                "than the number of milestone points, but got number of schedulers {} and the "
+                "number of milestones to be equal to {}".format(len(schedulers), len(milestones))
             )
         self._schedulers = schedulers
         self._milestones = milestones
@@ -862,7 +862,7 @@ def __init__(self, schedulers):
             if (schedulers[scheduler_idx].optimizer != schedulers[0].optimizer):
                 raise ValueError(
                     "ChainedScheduler expects all schedulers to belong to the same optimizer, but "
-                    f"got schedulers at index {0} and {scheduler_idx} to be different"
+                    "got schedulers at index {} and {} to be different".format(0, scheduler_idx)
                 )
         self._schedulers = list(schedulers)
         self.optimizer = schedulers[0].optimizer
diff --git a/torch/overrides.py b/torch/overrides.py
index 1fd26d5312ad8e..80d4a8844a75e1 100644
--- a/torch/overrides.py
+++ b/torch/overrides.py
@@ -1577,9 +1577,9 @@ def handle_torch_function(
 
     func_name = f'{public_api.__module__}.{public_api.__name__}'
     msg = (
-        f"no implementation found for '{func_name}' on types that implement "
-        f'__torch_function__: {[type(arg) for arg in overloaded_args]}'
-    )
+        "no implementation found for '{}' on types that implement "
+        '__torch_function__: {}'
+    ).format(func_name, [type(arg) for arg in overloaded_args])
     if _is_torch_function_mode_enabled():
         msg += f" nor in mode {_get_current_function_mode()}"
     raise TypeError(msg)
diff --git a/torch/quasirandom.py b/torch/quasirandom.py
index 1c9b949c55651c..c5086da283a4d2 100644
--- a/torch/quasirandom.py
+++ b/torch/quasirandom.py
@@ -122,11 +122,11 @@ def draw_base2(self, m: int, out: Optional[torch.Tensor] = None,
         total_n = self.num_generated + n
         if not (total_n & (total_n - 1) == 0):
             raise ValueError("The balance properties of Sobol' points require "
-                             f"n to be a power of 2. {self.num_generated} points have been "
-                             f"previously generated, then: n={self.num_generated}+2**{m}={total_n}. "
+                             "n to be a power of 2. {0} points have been "
+                             "previously generated, then: n={0}+2**{1}={2}. "
                              "If you still want to do this, please use "
                              "'SobolEngine.draw()' instead."
-                             )
+                             .format(self.num_generated, m, total_n))
         return self.draw(n=n, out=out, dtype=dtype)
 
     def reset(self):
diff --git a/torch/serialization.py b/torch/serialization.py
index aeb1d8444612ba..875c6abe030a81 100644
--- a/torch/serialization.py
+++ b/torch/serialization.py
@@ -202,8 +202,10 @@ def check_module_version_greater_or_equal(module, req_version_tuple, error_if_ma
 
     except Exception as e:
         message = (
-            f"'{module.__name__}' module version string is malformed '{module.__version__}' and cannot be compared"
-            f" with tuple {str(req_version_tuple)}"
+            "'{}' module version string is malformed '{}' and cannot be compared"
+            " with tuple {}"
+        ).format(
+            module.__name__, module.__version__, str(req_version_tuple)
         )
         if error_if_malformed:
             raise RuntimeError(message) from e
diff --git a/torch/testing/_internal/common_device_type.py b/torch/testing/_internal/common_device_type.py
index c1d19c02b43bf1..7f045b0e658b31 100644
--- a/torch/testing/_internal/common_device_type.py
+++ b/torch/testing/_internal/common_device_type.py
@@ -1195,7 +1195,7 @@ def __init__(self, *args, device_type="all"):
                 assert isinstance(arg, (list, tuple)), \
                     "When one dtype variant is a tuple or list, " \
                     "all dtype variants must be. " \
-                    f"Received non-list non-tuple dtype {str(arg)}"
+                    "Received non-list non-tuple dtype {}".format(str(arg))
                 assert all(isinstance(dtype, torch.dtype) for dtype in arg), f"Unknown dtype in {str(arg)}"
         else:
             assert all(isinstance(arg, torch.dtype) for arg in args), f"Unknown dtype in {str(args)}"
diff --git a/torch/testing/_internal/common_utils.py b/torch/testing/_internal/common_utils.py
index 1e18ca2afec56e..8d2a5e1ac8525f 100644
--- a/torch/testing/_internal/common_utils.py
+++ b/torch/testing/_internal/common_utils.py
@@ -506,8 +506,9 @@ def test_wrapper(*args, **kwargs):
 
                 values = list(values) if len(self.arg_names) > 1 else [values]
                 if len(values) != len(self.arg_names):
-                    raise RuntimeError(f'Expected # values == # arg names, but got: {len(values)} '
-                                       f'values and {len(self.arg_names)} names for test "{test.__name__}"')
+                    raise RuntimeError('Expected # values == # arg names, but got: {} '
+                                       'values and {} names for test "{}"'.format(
+                                           len(values), len(self.arg_names), test.__name__))
 
                 param_kwargs = dict(zip(self.arg_names, values))
 
@@ -3466,9 +3467,9 @@ def accept_output(update_type):
                 return accept_output("output")
             else:
                 raise RuntimeError(
-                      f"I got this output for {munged_id}{subname_output}:\n\n{s}\n\n"
-                      "No expect file exists; to accept the current output, run:\n"
-                      f"python {__main__.__file__} {munged_id} --accept") from None
+                    ("I got this output for {}{}:\n\n{}\n\n"
+                     "No expect file exists; to accept the current output, run:\n"
+                     "python {} {} --accept").format(munged_id, subname_output, s, __main__.__file__, munged_id)) from None
 
         # a hack for JIT tests
         if IS_WINDOWS:
@@ -4071,9 +4072,10 @@ def check_test_defined_in_running_script(test_case):
     if running_script_path is None:
         return
     test_case_class_file = os.path.abspath(os.path.realpath(inspect.getfile(test_case.__class__)))
-    assert test_case_class_file == running_script_path, f"Class of loaded TestCase \"{test_case.id()}\" " \
-        f"is not defined in the running script \"{running_script_path}\", but in \"{test_case_class_file}\". Did you " \
-        "accidentally import a unittest.TestCase from another file?"
+    assert test_case_class_file == running_script_path, "Class of loaded TestCase \"{}\" " \
+        "is not defined in the running script \"{}\", but in \"{}\". Did you " \
+        "accidentally import a unittest.TestCase from another file?".format(
+            test_case.id(), running_script_path, test_case_class_file)
 
 def load_tests(loader, tests, pattern):
     set_running_script_path()
diff --git a/torch/testing/_internal/hypothesis_utils.py b/torch/testing/_internal/hypothesis_utils.py
index 0654a64b96b268..4ace78f7594e33 100644
--- a/torch/testing/_internal/hypothesis_utils.py
+++ b/torch/testing/_internal/hypothesis_utils.py
@@ -362,7 +362,7 @@ def assert_deadline_disabled():
         warning_message = (
             "Your version of hypothesis is outdated. "
             "To avoid `DeadlineExceeded` errors, please update. "
-            f"Current hypothesis version: {hypothesis.__version__}"
+            "Current hypothesis version: {}".format(hypothesis.__version__)
         )
         warnings.warn(warning_message)
     else:
diff --git a/torch/testing/_internal/jit_utils.py b/torch/testing/_internal/jit_utils.py
index 447a425b9a5bfb..f96c2fb436be9d 100644
--- a/torch/testing/_internal/jit_utils.py
+++ b/torch/testing/_internal/jit_utils.py
@@ -66,7 +66,7 @@ def get_execution_plan(graph_executor_state):
     num_plans = len(execution_plans)
     if num_plans != 1:
         raise RuntimeError('This test assumes this GraphExecutor should '
-                           f'only have one execution plan, got: {num_plans}')
+                           'only have one execution plan, got: {}'.format(num_plans))
     return execution_plans[0]
 
 class _AssertRaisesRegexWithHighlightContext:
diff --git a/torch/utils/bundled_inputs.py b/torch/utils/bundled_inputs.py
index b18857b0e716ee..df2d771f6f51d1 100644
--- a/torch/utils/bundled_inputs.py
+++ b/torch/utils/bundled_inputs.py
@@ -334,14 +334,17 @@ def get_all_bundled_inputs_for_{name}(self):
 
         # Add to the high level helper methods
         inputs_info = repr(info[function]) if info and function in info else '[]'
-        get_bundled_inputs_functions_and_info_template += f"""
+        get_bundled_inputs_functions_and_info_template += """
             temp_dict : Dict[str,List[str]] = {{}}
-            info: List[str] = {inputs_info}
+            info: List[str] = {info}
 
             temp_dict['info'] = info
-            temp_dict['get_inputs_function_name'] = ['get_all_bundled_inputs_for_{function_name}']
-            all_inputs['{function_name}'] = temp_dict
-            """
+            temp_dict['get_inputs_function_name'] = ['get_all_bundled_inputs_for_{name}']
+            all_inputs['{name}'] = temp_dict
+            """.format(
+            name=function_name,
+            info=inputs_info,
+        )
 
         # To ensure backwards compatibility and a streamlined api for forward these wrappers are provided
         if function_name == 'forward':
diff --git a/torch/utils/cpp_extension.py b/torch/utils/cpp_extension.py
index 6600eb3cb54037..0ed858a96cb464 100644
--- a/torch/utils/cpp_extension.py
+++ b/torch/utils/cpp_extension.py
@@ -1611,9 +1611,11 @@ def load_inline(name,
             raise ValueError(f"Expected 'functions' to be a list or dict, but was {type(functions)}")
         for function_name, docstring in functions.items():
             if with_pytorch_error_handling:
-                module_def.append(f'm.def("{function_name}", torch::wrap_pybind_function({function_name}), "{docstring}");')
+                module_def.append(
+                    'm.def("{0}", torch::wrap_pybind_function({0}), "{1}");'
+                    .format(function_name, docstring))
             else:
-                module_def.append(f'm.def("{function_name}", {function_name}, "{docstring}");')
+                module_def.append('m.def("{0}", {0}, "{1}");'.format(function_name, docstring))
         module_def.append('}')
         cpp_sources += module_def
 
diff --git a/torch/utils/data/dataloader.py b/torch/utils/data/dataloader.py
index f3cd8e3d331a41..7f50905617957f 100644
--- a/torch/utils/data/dataloader.py
+++ b/torch/utils/data/dataloader.py
@@ -315,7 +315,7 @@ def __init__(self, dataset: Dataset[T_co], batch_size: Optional[int] = 1,
                 # See NOTE [ Custom Samplers and IterableDataset ]
                 raise ValueError(
                     "DataLoader with IterableDataset: expected unspecified "
-                    f"batch_sampler option, but got batch_sampler={batch_sampler}")
+                    "batch_sampler option, but got batch_sampler={}".format(batch_sampler))
         else:
             shuffle = bool(shuffle)
             self._dataset_kind = _DatasetKind.Map
@@ -397,19 +397,19 @@ def multiprocessing_context(self, multiprocessing_context):
                     valid_start_methods = multiprocessing.get_all_start_methods()
                     if multiprocessing_context not in valid_start_methods:
                         raise ValueError(
-                            'multiprocessing_context option '
-                            f'should specify a valid start method in {valid_start_methods!r}, but got '
-                            f'multiprocessing_context={multiprocessing_context!r}')
+                            ('multiprocessing_context option '
+                             'should specify a valid start method in {!r}, but got '
+                             'multiprocessing_context={!r}').format(valid_start_methods, multiprocessing_context))
                     multiprocessing_context = multiprocessing.get_context(multiprocessing_context)
 
                 if not isinstance(multiprocessing_context, python_multiprocessing.context.BaseContext):
-                    raise TypeError('multiprocessing_context option should be a valid context '
-                                    'object or a string specifying the start method, but got '
-                                    f'multiprocessing_context={multiprocessing_context}')
+                    raise TypeError(('multiprocessing_context option should be a valid context '
+                                     'object or a string specifying the start method, but got '
+                                     'multiprocessing_context={}').format(multiprocessing_context))
             else:
-                raise ValueError('multiprocessing_context can only be used with '
-                                 'multi-process loading (num_workers > 0), but got '
-                                 f'num_workers={self.num_workers}')
+                raise ValueError(('multiprocessing_context can only be used with '
+                                  'multi-process loading (num_workers > 0), but got '
+                                  'num_workers={}').format(self.num_workers))
 
         self.__multiprocessing_context = multiprocessing_context
 
diff --git a/torch/utils/data/datapipes/_decorator.py b/torch/utils/data/datapipes/_decorator.py
index b6e93776aa38ec..d2afd996f54ba8 100644
--- a/torch/utils/data/datapipes/_decorator.py
+++ b/torch/utils/data/datapipes/_decorator.py
@@ -70,7 +70,7 @@ def __init__(self, arg: Union[Type[IterDataPipe], Callable[[], bool]]) -> None:
         if isinstance(arg, Type):  # type: ignore[arg-type]
             if not issubclass(arg, IterDataPipe):  # type: ignore[arg-type]
                 raise TypeError("Only `IterDataPipe` can be decorated with `non_deterministic`"
-                                f", but {arg.__name__} is found")
+                                ", but {} is found".format(arg.__name__))
             self.cls = arg  # type: ignore[assignment]
         # 2. Decorator has an argument of a function
         #    This class should behave differently given different inputs. Use this
@@ -103,13 +103,13 @@ def deterministic_wrapper_fn(self, *args, **kwargs) -> IterDataPipe:
         res = self.deterministic_fn(*args, **kwargs)  # type: ignore[call-arg, misc]
         if not isinstance(res, bool):
             raise TypeError("deterministic_fn of `non_deterministic` decorator is required "
-                            f"to return a boolean value, but {type(res)} is found")
+                            "to return a boolean value, but {} is found".format(type(res)))
         global _determinism
         if _determinism and res:
-            raise TypeError(f"{self.cls.__name__} is non-deterministic with the inputs, but you set "  # type: ignore[union-attr]
+            raise TypeError("{} is non-deterministic with the inputs, but you set "
                             "'guaranteed_datapipes_determinism'. You can turn off determinism "
                             "for this DataPipe if that is acceptable for your application"
-                            )
+                            .format(self.cls.__name__))  # type: ignore[union-attr]
         return self.cls(*args, **kwargs)  # type: ignore[call-arg, misc]
 
 
@@ -130,9 +130,9 @@ def wrapper(*args, **kwargs):
                 if not isinstance(value, IterDataPipe):
                     raise TypeError(f"Expected argument '{argument_name}' as a IterDataPipe, but found {type(value)}")
                 if not value.type.issubtype(hint.type):
-                    raise TypeError(f"Expected type of argument '{argument_name}' as a subtype of "
-                                    f"hint {hint.type}, but found {value.type}"
-                                    )
+                    raise TypeError("Expected type of argument '{}' as a subtype of "
+                                    "hint {}, but found {}"
+                                    .format(argument_name, hint.type, value.type))
 
         return f(*args, **kwargs)
 
diff --git a/torch/utils/data/sampler.py b/torch/utils/data/sampler.py
index 1ba1dfc665efac..1e56df5adc8a43 100644
--- a/torch/utils/data/sampler.py
+++ b/torch/utils/data/sampler.py
@@ -223,7 +223,7 @@ def __init__(self, weights: Sequence[float], num_samples: int,
         weights_tensor = torch.as_tensor(weights, dtype=torch.double)
         if len(weights_tensor.shape) != 1:
             raise ValueError("weights should be a 1d sequence but given "
-                             f"weights have shape {tuple(weights_tensor.shape)}")
+                             "weights have shape {}".format(tuple(weights_tensor.shape)))
 
         self.weights = weights_tensor
         self.num_samples = num_samples
diff --git a/torch/utils/hipify/hipify_python.py b/torch/utils/hipify/hipify_python.py
index 4d7bf7233cfccc..34361d09221c5e 100755
--- a/torch/utils/hipify/hipify_python.py
+++ b/torch/utils/hipify/hipify_python.py
@@ -502,7 +502,7 @@ def hip_header_magic(input_string):
 
     # Check if one of the following headers is already included.
     headers = ["hip/hip_runtime.h", "hip/hip_runtime_api.h"]
-    if any(re.search(fr'#include ("{ext}"|<{ext}>)', output_string) for ext in headers):
+    if any(re.search(r'#include ("{0}"|<{0}>)'.format(ext), output_string) for ext in headers):
         return output_string
 
     # Rough logic to detect if we're inside device code
diff --git a/torch/utils/hooks.py b/torch/utils/hooks.py
index e0ab3618242af1..1cc83911d32de1 100644
--- a/torch/utils/hooks.py
+++ b/torch/utils/hooks.py
@@ -83,10 +83,10 @@ def warn_if_has_hooks(tensor):
         for k in tensor._backward_hooks:
             hook = tensor._backward_hooks[k]
             if not hasattr(k, "__torch_unserializable__"):
-                warnings.warn(f"backward hook {repr(hook)} on tensor will not be "
+                warnings.warn("backward hook {} on tensor will not be "
                               "serialized.  If this is expected, you can "
                               "decorate the function with @torch.utils.hooks.unserializable_hook "
-                              "to suppress this warning")
+                              "to suppress this warning".format(repr(hook)))
 
 class BackwardHook:
     """
@@ -140,7 +140,7 @@ def hook(grad_input, _):
 
                 if len(out) != len(res):
                     raise RuntimeError("Backward hook returned an invalid number of grad_input, "
-                                       f"got {len(out)}, but expected {len(res)}")
+                                       "got {}, but expected {}".format(len(out), len(res)))
 
                 res = out
 
@@ -209,7 +209,7 @@ def hook(_, grad_output):
                         actual_len = len(hook_grad_outputs)
                         if actual_len != expected_len:
                             raise RuntimeError("Backward pre hook returned an invalid number of grad_output, "
-                                               f"got {actual_len}, but expected {expected_len}")
+                                               "got {}, but expected {}".format(actual_len, expected_len))
                         self.grad_outputs = hook_grad_outputs
 
                 # Special case if no input required gradients, this hook should call the user
diff --git a/torch/utils/mobile_optimizer.py b/torch/utils/mobile_optimizer.py
index b89169a80a93e1..fae6efe265f988 100644
--- a/torch/utils/mobile_optimizer.py
+++ b/torch/utils/mobile_optimizer.py
@@ -95,9 +95,9 @@ def generate_mobile_module_lints(script_module: torch.jit.ScriptModule):
 
     for name, param in script_module.named_parameters():
         if param.requires_grad:
-            lint_list.append({"name": LintCode.REQUIRES_GRAD.name, "message": f"Param {name} requires grad, "
+            lint_list.append({"name": LintCode.REQUIRES_GRAD.name, "message": "Param {} requires grad, "
                              "please set torch.no_grad() to reduce memory usage and improve computation speed during "
-                              "inference phase."})
+                              "inference phase.".format(name)})
 
     op_names = torch.jit.export_opnames(script_module)
     for op_name in op_names: