[BE]: Apply PERF401 autofixes from ruff (#140980)

* Automatically applies ruff rule 401. Turns loops into equivalent list comprehensions which are faster and do not leak the scope of the loop variables.
* list comprehensions not only often have better typing, but are 50+% faster than for loops on overhead. They also preserve length information etc and are better for the interpreter to optimize.
* Manually went back and made mypy happy after the change.
* Also fixed style lints in files covered by flake8 but not by pyfmt

Pull Request resolved: https://github.com/pytorch/pytorch/pull/140980
Approved by: https://github.com/justinchuby, https://github.com/malfet

torch/_dynamo/backends/distributed.py

@@ -67,8 +67,7 @@ def pretty_print_buckets(buckets: List[Bucket], bucket_bytes_cap: int):
     for idx, bucket in enumerate(reversed(buckets)):
         if len(bucket.params) > 0:
             rows.append((idx, bucket.size, bucket.params[0]))
-            for param in bucket.params[1:]:
+            rows.extend((None, None, param) for param in bucket.params[1:])
-                rows.append((None, None, param))
         if bucket.opcount_increased_to_capture_external_output > 0:
             extended_buckets.append(
                 (

torch/_dynamo/bytecode_transformation.py

@@ -1140,11 +1140,12 @@ def update_offsets(instructions) -> None:
 
 def debug_bytes(*args) -> str:
     index = range(max(map(len, args)))
-    result = []
+    result = [
-    for arg in (
+        " ".join(f"{x:03}" for x in arg)
-        [index] + list(args) + [[int(a != b) for a, b in zip(args[-1], args[-2])]]
+        for arg in [index]
-    ):
+        + list(args)
-        result.append(" ".join(f"{x:03}" for x in arg))
+        + [[int(a != b) for a, b in zip(args[-1], args[-2])]]
+    ]
 
     return "bytes mismatch\n" + "\n".join(result)
 

torch/_dynamo/compiled_autograd.py

@@ -478,10 +478,8 @@ class AutogradCompilerInstance:
 
     @staticmethod
     def get_all_nodes(args):
-        nodes = []
+        # filter out non-Node args, like None
-        for n in args:
+        nodes = [n for n in args if type(n) is torch.fx.Node]
-            if type(n) is torch.fx.Node:  # filter out non-Node args, like None
-                nodes.append(n)
         return nodes
 
     @staticmethod
@@ -671,13 +669,15 @@ class AutogradCompilerInstance:
             input_nodes_and_users = []
             input_nodes_and_users.extend(list(input_nodes))
             for input_node in input_nodes:
-                for user in list(input_node.users.keys()):
+                input_nodes_and_users.extend(
+                    user
+                    for user in list(input_node.users.keys())
                     if not (
                         user.op == "call_function"
                         and user.target == call_hook
                         and node.kwargs.get("hook_type", None) == "post_hook"
-                    ):
+                    )
-                        input_nodes_and_users.append(user)
+                )
 
             arg = max(input_nodes_and_users)  # last input users
             if (

torch/_dynamo/guards.py

@@ -2526,7 +2526,6 @@ def make_torch_function_mode_stack_guard(intial_stack):
 
 
 def recompilation_reason_for_no_tensor_aliasing_guard(guard_manager, scope):
-    duplicate_tensors = []
     global_scope = dict(guard_manager.global_scope)
     ids_to_source = collections.defaultdict(list)
     for tensor_source in guard_manager.no_tensor_aliasing_sources:  # type: ignore[attr-defined]
@@ -2534,9 +2533,9 @@ def recompilation_reason_for_no_tensor_aliasing_guard(guard_manager, scope):
         tensor_id = id(eval(tensor_source, global_scope, scope))
         ids_to_source[tensor_id].append(tensor_source)
 
-    for key in ids_to_source:
+    duplicate_tensors = [
-        if len(ids_to_source[key]) > 1:
+        f"{ids_to_source[key]}" for key in ids_to_source if len(ids_to_source[key]) > 1
-            duplicate_tensors.append(f"{ids_to_source[key]}")
+    ]
 
     reason = ", ".join(duplicate_tensors)
     return [f"Duplicate tensors found: {reason}"]

torch/_dynamo/output_graph.py

@@ -1463,9 +1463,7 @@ class OutputGraph:
         return compiled_fn
 
     def example_inputs(self) -> List[torch.Tensor]:
-        result = []
+        result = [arg.example for arg in self.graphargs]
-        for arg in self.graphargs:
-            result.append(arg.example)
         return result
 
     def remove_unused_graphargs(self) -> None:

torch/_dynamo/resume_execution.py

@@ -252,8 +252,7 @@ def _filter_iter(l1, l2, cond):
 def _load_tuple_and_call(tup):
     insts: List[Instruction] = []
     _initial_push_null(insts)
-    for val in tup:
+    insts.extend(create_instruction("LOAD_CONST", argval=val) for val in tup)
-        insts.append(create_instruction("LOAD_CONST", argval=val))
     insts.extend(create_call_function(len(tup), False))
     return insts
 

torch/_dynamo/testing.py

@@ -95,9 +95,7 @@ def collect_results(
     results.append(buffers)
     for example in example_inputs:
         if isinstance(example, (tuple, list)):
-            for inp in example:
+            results.extend(inp.grad for inp in example if isinstance(inp, torch.Tensor))
-                if isinstance(inp, torch.Tensor):
-                    results.append(inp.grad)
         else:
             if isinstance(example, torch.Tensor):
                 results.append(example.grad)

torch/_dynamo/utils.py

@@ -1535,9 +1535,10 @@ def checkpoint_params(gm):
         rng_state = torch.clone(torch.random.get_rng_state())
         if torch.cuda.is_available():
             cuda_rng_state = torch.clone(torch.cuda.get_rng_state())
-        saved_state = []
+        saved_state = [
-        for param in itertools.chain(gm.parameters(), gm.buffers()):
+            (param, param._version, torch.clone(param))
-            saved_state.append((param, param._version, torch.clone(param)))
+            for param in itertools.chain(gm.parameters(), gm.buffers())
+        ]
 
     def restore():
         with torch.no_grad():

torch/_dynamo/variables/builtin.py

@@ -1765,10 +1765,9 @@ class BuiltinVariable(VariableTracker):
                     # tracked fakes to produce incorrect guards. This is sound because the TensorVariable
                     # coming out of set_() below will be a new one, and get
                     # installed in tracked fakes.
-                    to_remove = []
+                    to_remove = [
-                    for tf in tx.output.tracked_fakes:
+                        tf for tf in tx.output.tracked_fakes if tf.source == obj.source
-                        if tf.source == obj.source:
+                    ]
-                            to_remove.append(tf)
                     for tf in to_remove:
                         tx.output.tracked_fakes.remove(tf)
 

torch/_dynamo/variables/ctx_manager.py

@@ -1026,17 +1026,16 @@ class SDPAKernelVariable(ContextWrappingVariable):
 
     @staticmethod
     def _backends_to_nodes(tx, backends):
-        nodes = []
-        for backend in backends:
         # convert to/from string in order to bake the backend into FX graph
-            nodes.append(
+        nodes = [
             tx.output.create_node(
                 "call_function",
                 torch.nn.attention._backend_from_string,
                 (backend.name,),
                 {},
             )
-            )
+            for backend in backends
+        ]
         return nodes
 
     def enter(self, tx):

torch/_dynamo/variables/iter.py

@@ -48,9 +48,9 @@ class ItertoolsVariable(VariableTracker):
             and all(arg.has_unpack_var_sequence(tx) for arg in args)
         ):
             seqs = [arg.unpack_var_sequence(tx) for arg in args]
-            items = []
+            items = [
-            for item in itertools.product(*seqs):
+                variables.TupleVariable(list(item)) for item in itertools.product(*seqs)
-                items.append(variables.TupleVariable(list(item)))
+            ]
             return variables.ListIteratorVariable(
                 items, mutation_type=ValueMutationNew()
             )

torch/_dynamo/variables/user_defined.py

@@ -776,14 +776,13 @@ class UserDefinedObjectVariable(UserDefinedVariable):
             ):
                 assert self.source  # OrderedDict, dict subtypes must always have source
                 assert not (args or kwargs)
-                items = []
                 keys = self.call_method(tx, "keys", [], {})
-                for key in keys.force_unpack_var_sequence(tx):
+                items = [
-                    items.append(
                     TupleVariable(
                         [key, self.odict_getitem(tx, key)],
                     )
-                    )
+                    for key in keys.force_unpack_var_sequence(tx)
+                ]
                 tx.output.guard_on_key_order.add(self.source.name())
                 return TupleVariable(items)
 

torch/_export/converter.py

@@ -833,9 +833,7 @@ class TS2FXGraphConverter:
         self._convert_prim_iterator(node)
 
     def _convert_prim_iterator(self, node: torch._C.Node):
-        output_list = []
+        output_list = [self.get_fx_value_by_ir_value(inp) for inp in node.inputs()]
-        for inp in node.inputs():
-            output_list.append(self.get_fx_value_by_ir_value(inp))
 
         output_name = node.output().debugName()
         self.name_to_node[output_name] = output_list

torch/_export/db/examples/static_for_loop.py

@@ -7,9 +7,8 @@ class StaticForLoop(torch.nn.Module):
     """
 
     def forward(self, x):
-        ret = []
+        # constant
-        for i in range(10):  # constant
+        ret = [i + x for i in range(10)]
-            ret.append(i + x)
         return ret
 
 example_args = (torch.randn(3, 2),)

torch/_export/serde/serialize.py

@@ -1780,9 +1780,7 @@ class GraphModuleDeserializer(metaclass=Final):
                 ) from e
 
         # Outputs: convert to a single `output` node.
-        outputs = []
+        outputs = [self.deserialize_graph_output(output) for output in serialized_graph.outputs]
-        for output in serialized_graph.outputs:
-            outputs.append(self.deserialize_graph_output(output))
 
         if serialized_graph.is_single_tensor_return:
             assert len(outputs) == 1
@@ -2149,9 +2147,7 @@ class GraphModuleDeserializer(metaclass=Final):
             if len(value) == 0:
                 return []
             elif typ_ == "as_tensors":
-                result = []
+                result = [self.serialized_name_to_node[arg.name] for arg in value]
-                for arg in value:
-                    result.append(self.serialized_name_to_node[arg.name])
                 return result
             elif typ_ in ("as_ints", "as_floats", "as_bools", "as_strings"):
                 # convert from serialized.python.types.List to python list

torch/_functorch/_aot_autograd/runtime_wrappers.py

@@ -1306,8 +1306,9 @@ def merge_view_inputs(
             mutated_input_info[inpt_idx].mutates_data
             for inpt_idx in aliased_input_indices
         ):
-            for curr_idx in aliased_input_indices:
+            other_args.extend(
-                other_args.append(fwd_inputs[curr_idx])
+                fwd_inputs[curr_idx] for curr_idx in aliased_input_indices
+            )
             continue
 
         # Here, we attempt to do a more complicated check to detect false aliasing
@@ -1320,8 +1321,9 @@ def merge_view_inputs(
             fwd_inputs, aliased_input_indices
         )
         if len(aliased_input_indices_no_false_sharing) <= 1:
-            for curr_idx in aliased_input_indices:
+            other_args.extend(
-                other_args.append(fwd_inputs[curr_idx])
+                fwd_inputs[curr_idx] for curr_idx in aliased_input_indices
+            )
             continue
 
         # We detected an input that was mutated, AND aliases with another input.

torch/_higher_order_ops/map.py

@@ -217,9 +217,7 @@ def trace_map(proxy_mode, func_overload, f, xs, pos_args):
 
 @map_impl.py_impl(DispatchKey.CompositeExplicitAutograd)
 def map_dense(f, xs, pos_args):
-    pytrees = []
+    pytrees = [f(*inp, *pos_args) for inp in _unstack_pytree(xs)]
-    for inp in _unstack_pytree(xs):
-        pytrees.append(f(*inp, *pos_args))
     return _stack_pytree(pytrees)
 
 

torch/_higher_order_ops/triton_kernel_wrap.py

@@ -543,8 +543,7 @@ def analyze_kernel_mutations(
                 )
                 stack.extend(arg for arg, mutated in zip(op.args, mutations) if mutated)
             else:
-                for idx in MUTATION_OPS.get(op.name, []):
+                stack.extend(op.args[idx] for idx in MUTATION_OPS.get(op.name, []))
-                    stack.append(op.args[idx])
 
     # The following is an iterative DFS algorithm
     mutated = [False] * num_args

torch/_higher_order_ops/utils.py

@@ -388,9 +388,7 @@ def _unstack_pytree(xs):
 
     a = zip(*flat_xs)
 
-    pytrees = []
+    pytrees = [pytree.tree_unflatten(tuple, inspec) for tuple in a]
-    for tuple in a:
-        pytrees.append(pytree.tree_unflatten(tuple, inspec))
     return pytrees
 
 

torch/_inductor/codecache.py

@@ -2102,13 +2102,14 @@ class AotCodeCompiler:
                 aot_constants = struct.pack("qq", consts_size + 8, magic_number)
 
             consts_o = _compile_consts(aot_constants, sys.platform)
-            kernels_o = []
             gpu_codecache: Union[ROCmCodeCache, CUDACodeCache] = (
                 ROCmCodeCache() if torch.version.hip else CUDACodeCache()
             )
-            for entry in gpu_codecache.cache.values():
+            kernels_o = [
-                if entry.output_path.endswith(".o"):
+                entry.output_path
-                    kernels_o.append(entry.output_path)
+                for entry in gpu_codecache.cache.values()
+                if entry.output_path.endswith(".o")
+            ]
             kernels_o = " ".join(kernels_o)
 
             output_name, output_dir = get_name_and_dir_from_output_file_path(output_so)

torch/_inductor/codegen/cpp.py

@@ -4800,10 +4800,9 @@ class LoopLevel:
 
     def split_with_tiling(self, depth, factor):
         def clone_inner():
-            inner = []
+            inner: List[LoopLevel] = []
             if self.inner:
-                for loop in self.inner:
+                inner.extend(loop.clone() for loop in self.inner)
-                    inner.append(loop.clone())
             return inner
 
         def do_split_with_tiling():

torch/_inductor/codegen/debug_utils.py

@@ -150,18 +150,16 @@ class DebugPrinterManager:
         # get the list of args_to_print_or_save
         # TODO: Find a more reliable way to detect kernel args types to print for extern kernel calls
         if kernel_type == "extern":
-            args_to_print_or_save_extern = []
+            args_to_print_or_save_extern = [
-            for arg in args_to_print_or_save:
+                arg for arg in args_to_print_or_save if arg.startswith(("buf", "arg"))
-                if arg.startswith(("buf", "arg")):
+            ]
-                    args_to_print_or_save_extern.append(arg)
             self.args_to_print_or_save = args_to_print_or_save_extern
         elif kernel_type == "cpp":
-            args_to_print_or_save_cpp = []
+            args_to_print_or_save_cpp = [
-            for arg in args_to_print_or_save:
-                if arg.startswith(("buf", "arg")):
-                    args_to_print_or_save_cpp.append(
                 f"convert_arrayref_tensor_to_tensor({arg})"
-                    )
+                for arg in args_to_print_or_save
+                if arg.startswith(("buf", "arg"))
+            ]
             self.args_to_print_or_save = args_to_print_or_save_cpp
         else:
             self.args_to_print_or_save = args_to_print_or_save

torch/_inductor/codegen/halide.py

@@ -1392,8 +1392,7 @@ class HalideKernel(SIMDKernel):
             result.append((call_str, arg))
             if isinstance(arg, TensorArg):
                 assert arg.offset == 0 and arg.alias_of is None
-                for alias in self.buffer_aliases.get(arg.name, ()):
+                result.extend(
-                    result.append(
                     (
                         None,
                         TensorArg(
@@ -1404,6 +1403,7 @@ class HalideKernel(SIMDKernel):
                             alias_of=arg.name,
                         ),
                     )
+                    for alias in self.buffer_aliases.get(arg.name, ())
                 )
         return result
 

torch/_inductor/codegen/multi_kernel.py

@@ -72,10 +72,11 @@ def get_all_call_args(call_args_list, arg_types_list):
 
 
 def get_numel_argdefs(kernel):
-    numel_argdefs = []
+    numel_argdefs = [
-    for tree in kernel.range_trees:
+        f"{tree.prefix}numel"
-        if tree.prefix != "r" or kernel.inside_reduction:
+        for tree in kernel.range_trees
-            numel_argdefs.append(f"{tree.prefix}numel")
+        if tree.prefix != "r" or kernel.inside_reduction
+    ]
 
     return numel_argdefs
 

torch/_inductor/codegen/triton_combo_kernel.py

@@ -76,7 +76,7 @@ def _default_custom_combo_kernel_horizontal_partition(
     tilings = [node_info_map[n][1] for n in nodes]
 
     max_dims = max(len(t) for t in tilings)
-    nodes_per_ndim = []
+    nodes_per_ndim: List[List[BaseSchedulerNode]] = []
     for i in range(2, max_dims + 1):
         group_per_dim = [n for n, t in zip(nodes, tilings) if len(t) == i]
         reduction = [
@@ -111,12 +111,11 @@ def _default_custom_combo_kernel_horizontal_partition(
                 len(large_pointwise),
             )
             not_reduction = [n for n in not_reduction if n not in large_pointwise]
-            for node in large_pointwise:
+            nodes_per_ndim.extend([node] for node in large_pointwise)
-                nodes_per_ndim.append([node])
 
-        for g in (not_reduction, short_reduction, long_reduction):
+        nodes_per_ndim.extend(
-            if g:
+            g for g in (not_reduction, short_reduction, long_reduction) if g
-                nodes_per_ndim.append(g)
+        )
 
     assert sum(len(p) for p in nodes_per_ndim) == len(nodes)
     return nodes_per_ndim

torch/_inductor/cpp_builder.py

@@ -288,8 +288,7 @@ def get_compiler_version_info(compiler: str) -> str:
 
 # =============================== cpp builder ===============================
 def _append_list(dest_list: List[str], src_list: List[str]) -> None:
-    for item in src_list:
+    dest_list.extend(copy.deepcopy(item) for item in src_list)
-        dest_list.append(copy.deepcopy(item))
 
 
 def _remove_duplication_in_list(orig_list: List[str]) -> List[str]:
@@ -742,12 +741,11 @@ def _setup_standard_sys_libs(
 
 
 def _get_build_args_of_chosen_isa(vec_isa: VecISA) -> Tuple[List[str], List[str]]:
-    macros = []
+    macros: List[str] = []
-    build_flags = []
+    build_flags: List[str] = []
     if vec_isa != invalid_vec_isa:
         # Add Windows support later.
-        for x in vec_isa.build_macro():
+        macros.extend(copy.deepcopy(x) for x in vec_isa.build_macro())
-            macros.append(copy.deepcopy(x))
 
         build_flags = [vec_isa.build_arch_flags()]
 

torch/_inductor/cpu_vec_isa.py

@@ -398,9 +398,11 @@ def valid_vec_isa_list() -> List[VecISA]:
         arch value is x86_64 on Linux, and the value is AMD64 on Windows.
         """
         _cpu_supported_x86_isa = x86_isa_checker()
-        for isa in supported_vec_isa_list:
+        isa_list.extend(
-            if all(flag in _cpu_supported_x86_isa for flag in str(isa).split()) and isa:
+            isa
-                isa_list.append(isa)
+            for isa in supported_vec_isa_list
+            if all(flag in _cpu_supported_x86_isa for flag in str(isa).split()) and isa
+        )
 
     return isa_list
 

torch/_inductor/cudagraph_trees.py

@@ -621,13 +621,12 @@ class CUDAWarmupNode:
         }
 
         def get_non_cudagraph_inps() -> List[weakref.ReferenceType[UntypedStorage]]:
-            non_cudagraph_inps = []
+            non_cudagraph_inps = [
-            for t in itertools.chain(new_inputs, self.wrapped_function.constants):
+                weakref.ref(t.untyped_storage())
-                if (
+                for t in itertools.chain(new_inputs, self.wrapped_function.constants)
-                    isinstance(t, torch.Tensor)
+                if isinstance(t, torch.Tensor)
                 and t.untyped_storage().data_ptr() not in existing_path_data_ptrs
-                ):
+            ]
-                    non_cudagraph_inps.append(weakref.ref(t.untyped_storage()))
             return non_cudagraph_inps
 
         non_cudagraph_inps_storages = get_non_cudagraph_inps()
@@ -1709,12 +1708,10 @@ def get_block_addrs(pool_id: Tuple[int, int], live_only: bool = True) -> List[in
 
 
 def format_tb(frames: List[Any]) -> str:
-    formatted_traceback = []
+    formatted_traceback = [
-
-    for entry in frames:
-        formatted_traceback.append(
         traceback.FrameSummary(entry["filename"], entry["line"], entry["name"])
-        )
+        for entry in frames
+    ]
 
     return "".join(traceback.format_list(formatted_traceback))
 

torch/_inductor/dependencies.py

@@ -127,9 +127,7 @@ class MemoryDep(Dep):
             return None
 
         stride_to_index = {s: i for i, s in enumerate(self_strides)}
-        order = []
+        order = [stride_to_index[s] for s in other_strides]
-        for s in other_strides:
-            order.append(stride_to_index[s])
 
         assert set(order) == set(range(0, self.num_vars))
         return order
@@ -547,9 +545,7 @@ def var_builder(prefix: str) -> Tuple[VarRanges, Callable[[sympy.Expr], sympy.Sy
 
 def index_vars_no_squeeze(*argsizes: Tuple[sympy.Expr, ...], prefix: str):
     var_ranges, add_var = var_builder(prefix)
-    args: List[List[sympy.Symbol]] = []
+    args: List[List[sympy.Symbol]] = [list(map(add_var, size)) for size in argsizes]
-    for size in argsizes:
-        args.append(list(map(add_var, size)))
     return args, var_ranges
 
 

torch/_inductor/fx_passes/split_cat.py

@@ -1120,9 +1120,9 @@ class UnbindCatRemover(SplitCatSimplifier):
             return
         # we need to check if the getitem indices from unbind are consecutive and all go to the same cat node
         # before we do the unbind remove, otherwise it will hit the error when we unbind part of them
-        getitem_indices = []
+        getitem_indices = [
-        for getitem_node in unbind_node.users.keys():
+            getitem_node.args[1] for getitem_node in unbind_node.users.keys()
-            getitem_indices.append(getitem_node.args[1])
+        ]
         if not is_sorted_and_consecutive(getitem_indices) or len(  # type: ignore[arg-type]
             getitem_indices
         ) != len(
@@ -1497,9 +1497,8 @@ def merge_getitem_cat(match: Match, split_sections: List[int], dim: int):
             ):
                 continue
             # find the index of getitems to be cated/stacked
-            indices = []
+            # type: ignore[union-attr]
-            for arg in cat_user.args[0]:  # type: ignore[union-attr]
+            indices = [arg.args[1] for arg in cat_user.args[0]]  # type: ignore[union-attr]
-                indices.append(arg.args[1])  # type: ignore[union-attr]
             # the gettitems to be merged must be consecutive, otherwise
             # returned sliced tensor could be wrong
             if not is_sorted_and_consecutive(indices):

torch/_inductor/graph.py

@@ -1657,14 +1657,13 @@ class GraphLowering(torch.fx.Interpreter):
             new_unbacked_defs |= op.get_unbacked_symbol_defs()
 
         def format_new_defs() -> str:
-            r = []
+            r = [
-            for buf in self.buffers[buffer_watermark:]:
-                r.append(
                 f"unbacked_symbol_defs={buf.get_unbacked_symbol_defs()} in:\n{buf}\n"
-                )
+                for buf in self.buffers[buffer_watermark:]
-            for op in self.operations[operation_watermark:]:
+            ]
-                r.append(
+            r.extend(
                 f"unbacked_symbol_defs={op.get_unbacked_symbol_defs()} in:\n{op}\n"
+                for op in self.operations[operation_watermark:]
             )
             return "***\n".join(r)
 

torch/_inductor/ir.py

@@ -5484,13 +5484,14 @@ class UserDefinedTritonKernel(ExternKernel):
 
         kernel = kernel_side_table.get_kernel(self.kernel_idx)
         configs = []
-        restore_value_args = []
+        restore_value_args: List[str] = []
         if isinstance(kernel, Autotuner):
             # https://github.com/triton-lang/triton/pull/5083
             # changes kernel.restore_idx to kernel.restore_value
             if hasattr(kernel, "restore_idx"):
-                for i in kernel.restore_idx:
+                restore_value_args.extend(
-                    restore_value_args.append(kernel.fn.arg_names[i])
+                    kernel.fn.arg_names[i] for i in kernel.restore_idx
+                )
             else:
                 assert hasattr(kernel, "restore_value")
                 restore_value_args.extend(kernel.restore_value)

torch/_inductor/lowering.py

@@ -1691,9 +1691,7 @@ def split_with_sizes(x, sizes, dim=0):
 def unbind(x, dim=0):
     dim = _validate_dim(x, dim, 0)
     x_size = V.graph.sizevars.evaluate_static_shape(x.get_size()[dim])
-    result = []
+    result = [select(x, dim, i) for i in range(x_size)]
-    for i in range(x_size):
-        result.append(select(x, dim, i))
     return result
 
 

torch/_inductor/mkldnn_ir.py

@@ -87,8 +87,7 @@ def _prepare_convolution_fusion_create(
             weight_size = []
             weight_size.append(prepacked_weight_size[1] * groups)
             weight_size.append(prepacked_weight_size[0] / groups)
-            for d in range(2, dim):
+            weight_size.extend(prepacked_weight_size[d] for d in range(2, dim))
-                weight_size.append(prepacked_weight_size[d])
         else:
             weight_size = prepacked_weight.transpose(0, 1).size()
         return weight_size

torch/_inductor/pattern_matcher.py

@@ -952,9 +952,10 @@ class PatternPrettyPrinter:
         assert hasattr(obj, "pretty_print")
         out_str = obj.pretty_print(pp=pp)
 
-        output = []
+        output = [
-        for key in pp.memoized_objs_names:
+            f"{pp.memoized_objs_names[key]} = {pp.memoized_objs_pp[key]}"
-            output.append(f"{pp.memoized_objs_names[key]} = {pp.memoized_objs_pp[key]}")
+            for key in pp.memoized_objs_names
+        ]
 
         output.append(f"{output_name} = {out_str}")
 
@@ -1361,9 +1362,7 @@ def register_replacement(
             return False
 
     def normalize_args(**kwargs: Any) -> List[Any]:
-        args = []
+        args = [kwargs.pop(name) for name in argnames_static]
-        for name in argnames_static:
-            args.append(kwargs.pop(name))
         for i in range(1, len(kwargs) + 1):
             if f"tangents_{i}" not in kwargs:
                 break

torch/_inductor/runtime/triton_heuristics.py

@@ -120,7 +120,7 @@ def autotune_hints_to_configs(
     configs to try.
     """
     xyz_options: Tuple[Tuple[int, Optional[int], Optional[int]], ...]
-    configs = []
+    configs: List[Config] = []
     warp_size = device_props.warp_size
     # CPU target has no concept of "warp"
     if warp_size is None:
@@ -138,8 +138,7 @@ def autotune_hints_to_configs(
                     (1, block_size // 4, 1),
                     (1, 1, block_size // 4),
                 )
-            for xyz in xyz_options:
+            configs.extend(
-                configs.append(
                 triton_config(
                     size_hints,
                     *xyz,
@@ -147,6 +146,7 @@ def autotune_hints_to_configs(
                         device_props.warp_size if device_props.warp_size else 32
                     ),
                 )
+                for xyz in xyz_options
             )
 
     return configs

torch/_inductor/scheduler.py

@@ -308,9 +308,13 @@ class BaseSchedulerNode:
             dep = deps.pop()
             used_names.add(dep)
             if V.graph.name_to_buffer.get(dep):
-                for alias in V.graph.name_to_buffer[dep].get_inputs_that_alias_output():
+                deps.extend(
-                    if alias not in used_names:
+                    alias
-                        deps.append(alias)
+                    for alias in V.graph.name_to_buffer[
+                        dep
+                    ].get_inputs_that_alias_output()
+                    if alias not in used_names
+                )
         return used_names
 
     def prune_deps(self) -> None:
@@ -3324,10 +3328,11 @@ class Scheduler:
                 node1 = node2
                 node2 = tmp
 
-            deps = []
+            deps = [
-            for dep in node1.read_writes.reads | node1.read_writes.writes:
+                dep
-                if dep in node2.read_writes.reads or dep in node2.read_writes.writes:
+                for dep in node1.read_writes.reads | node1.read_writes.writes
-                    deps.append(dep)
+                if dep in node2.read_writes.reads or dep in node2.read_writes.writes
+            ]
 
             return sum(self.dep_size_hint(dep) for dep in deps)
 

torch/_library/infer_schema.py

@@ -176,14 +176,12 @@ def derived_types(
         ]
 
     if list_base:
-        for seq_typ in derived_seq_types(base_type):
+        result.extend((seq_typ, f"{cpp_type}[]") for seq_typ in derived_seq_types(base_type))  # type: ignore[valid-type]
-            result.append((seq_typ, f"{cpp_type}[]"))  # type: ignore[valid-type]
     if optional_base_list:
-        for seq_typ in derived_seq_types(typing.Optional[base_type]):
+        result.extend((seq_typ, f"{cpp_type}?[]") for seq_typ in derived_seq_types(typing.Optional[base_type]))  # type: ignore[valid-type]
-            result.append((seq_typ, f"{cpp_type}?[]"))  # type: ignore[valid-type]
     if optional_list_base:
-        for seq_typ in derived_seq_types(base_type):  # type: ignore[valid-type]
+        # type: ignore[valid-type]
-            result.append((typing.Optional[seq_typ], f"{cpp_type}[]?"))  # type: ignore[valid-type]
+        result.extend((typing.Optional[seq_typ], f"{cpp_type}[]?") for seq_typ in derived_seq_types(base_type))  # type: ignore[valid-type]
     return result
 
 

torch/_library/utils.py

@@ -279,17 +279,17 @@ def requires_set_python_module() -> bool:
 
 def handle_dispatch_mode(curr_mode, op_overload, *args, **kwargs):
     assert isinstance(curr_mode, torch.utils._python_dispatch.TorchDispatchMode)
-    overload_types = []
     args_flattened, _ = torch.utils._pytree.tree_flatten((args, kwargs.values()))
-    for a in args_flattened:
     # TODO: need to double check the semantics of the "types" argument to torch_dispatch.
     # It's generated in PyInterpreter.cpp, but seems to be generated in two places,
     # where in one case we only include tensors with the python key, and in another
     # we include **all** tensors.
-        if isinstance(a, torch.Tensor) and torch._C._dispatch_keys(a).has(
+    overload_types = [
-            torch._C.DispatchKey.Python
+        type(a)
-        ):
+        for a in args_flattened
-            overload_types.append(type(a))
+        if isinstance(a, torch.Tensor)
+        and torch._C._dispatch_keys(a).has(torch._C.DispatchKey.Python)
+    ]
     # TODO: check that I got these args correct (in C++, we pass in "0000"??)
 
     return curr_mode.__torch_dispatch__(op_overload, overload_types, args, kwargs)

torch/_logging/structured.py

@@ -43,15 +43,14 @@ def dump_file(filename: str) -> None:
 
 
 def from_traceback(tb: Sequence[traceback.FrameSummary]) -> List[Dict[str, Any]]:
-    r = []
-    for frame in tb:
     # dict naming convention here coincides with
     # python/combined_traceback.cpp
-        r.append(
+    r = [
         {
             "line": frame.lineno,
             "name": frame.name,
             "filename": intern_string(frame.filename),
         }
-        )
+        for frame in tb
+    ]
     return r

torch/_prims/__init__.py

@@ -312,10 +312,11 @@ def _make_prim(
         prim_autograd_impl.impl(name, _autograd_impl)
         prim_meta_impl.impl(name, meta)
     else:
-        mutates_args = []
+        mutates_args = [
-        for arg in cpp_schema.arguments:
+            arg.name
-            if arg.alias_info is not None and arg.alias_info.is_write:
+            for arg in cpp_schema.arguments
-                mutates_args.append(arg.name)
+            if arg.alias_info is not None and arg.alias_info.is_write
+        ]
         prim_def = torch.library.custom_op(
             "prims::" + name,
             _prim_impl,

torch/_refs/__init__.py

@@ -3055,9 +3055,7 @@ def chunk(a: TensorLikeType, chunks: int, dim: int = 0) -> Tuple[TensorLikeType,
     full_chunks = math.floor(length / chunk_size)
     tail_chunk_size = length % chunk_size
 
-    result = []
+    result = [narrow(a, dim, i * chunk_size, chunk_size) for i in range(full_chunks)]
-    for i in range(full_chunks):
-        result.append(narrow(a, dim, i * chunk_size, chunk_size))
 
     if tail_chunk_size != 0:
         result.append(narrow(a, dim, full_chunks * chunk_size, tail_chunk_size))

torch/_subclasses/fake_impls.py

@@ -585,14 +585,13 @@ def has_meta(func):
     lambda func: is_builtin(func) and "foreach" in func.name() and has_meta(func)
 )
 def foreach_run_and_map_input_device(fake_mode, func, *args, **kwargs):
-    tensor_lists = []
+    tensor_lists = [
-    for arg in itertools.chain(args, kwargs.values()):
+        arg
-        if (
+        for arg in itertools.chain(args, kwargs.values())
-            isinstance(arg, (list, tuple))
+        if isinstance(arg, (list, tuple))
         and len(arg)
         and isinstance(arg[0], torch.Tensor)
-        ):
+    ]
-            tensor_lists.append(arg)
 
     try:
         with in_kernel_invocation_manager(fake_mode):

torch/_subclasses/fake_tensor.py

@@ -171,8 +171,7 @@ def get_plain_tensors(
             continue
 
         inner_keys, _ = curr.__tensor_flatten__()
-        for key in reversed(inner_keys):
+        todo.extend(getattr(curr, key) for key in reversed(inner_keys))
-            todo.append(getattr(curr, key))
 
     return out
 
@@ -1629,13 +1628,12 @@ class FakeTensorMode(TorchDispatchMode):
             )
 
         if isinstance(output, tuple):
-            output_infos = []
+            output_infos = [
-            for out_elem in output:
-                output_infos.append(
                 self._get_output_info_for_cache_entry(
                     state, key, func, args, kwargs, out_elem
                 )
-                )
+                for out_elem in output
+            ]
             return _DispatchCacheEntry(
                 output_infos=tuple(output_infos), is_output_tuple=True
             )
@@ -1727,9 +1725,7 @@ class FakeTensorMode(TorchDispatchMode):
         """
 
         if entry.is_output_tuple:
-            outputs = []
+            outputs = [
-            for output_info in entry.output_infos:
-                outputs.append(
                 self._get_output_tensor_from_cache_entry(
                     state,
                     output_info,
@@ -1737,7 +1733,8 @@ class FakeTensorMode(TorchDispatchMode):
                     func,
                     args,
                 )
-                )
+                for output_info in entry.output_infos
+            ]
             return tuple(outputs)
         else:
             return self._get_output_tensor_from_cache_entry(

torch/ao/nn/quantizable/modules/rnn.py

@@ -389,8 +389,7 @@ class LSTM(torch.nn.Module):
                 **factory_kwargs,
             )
         ]
-        for layer in range(1, num_layers):
+        layers.extend(
-            layers.append(
             _LSTMLayer(
                 self.hidden_size,
                 self.hidden_size,
@@ -399,6 +398,7 @@ class LSTM(torch.nn.Module):
                 bidirectional=self.bidirectional,
                 **factory_kwargs,
             )
+            for layer in range(1, num_layers)
         )
         self.layers = torch.nn.ModuleList(layers)
 

torch/ao/nn/quantized/modules/conv.py

@@ -32,8 +32,7 @@ def _reverse_repeat_padding(padding: List[int]) -> List[int]:
     _reversed_padding_repeated_twice: List[int] = []
     N = len(padding)
     for idx in range(N):
-        for _ in range(2):
+        _reversed_padding_repeated_twice.extend(padding[N - idx - 1] for _ in range(2))
-            _reversed_padding_repeated_twice.append(padding[N - idx - 1])
     return _reversed_padding_repeated_twice
 
 

torch/ao/ns/fx/n_shadows_utils.py

@@ -568,9 +568,9 @@ def create_one_transformed_and_logged_copy_of_subgraph(
                     and len(arg)
                     and isinstance(arg[0], Node)
                 ):
-                    for inner_arg in arg:
+                    new_args.extend(
-                        if isinstance(inner_arg, Node):
+                        inner_arg for inner_arg in arg if isinstance(inner_arg, Node)
-                            new_args.append(inner_arg)
+                    )
 
             new_kwargs = {}
             for name, old_kwarg in first_node.kwargs.items():

torch/ao/ns/fx/utils.py

@@ -312,10 +312,7 @@ def get_arg_indices_of_inputs_to_log(node: Node) -> List[int]:
         node.target in (torch.add, torch.ops.quantized.add, operator.add)
         or node.target in (torch.mul, torch.ops.quantized.mul, operator.mul)
     ):
-        result = []
+        result = [i for i in range(2) if type(node.args[i]) == Node]
-        for i in range(2):
-            if type(node.args[i]) == Node:
-                result.append(i)
         return result
     return [0]
 

torch/ao/quantization/_equalize.py

@@ -191,8 +191,7 @@ def expand_groups_in_paired_modules_list(paired_modules_list):
         elif len(group) == 2:
             new_list.append(group)
         elif len(group) > 2:
-            for i in range(len(group) - 1):
+            new_list.extend([group[i], group[i + 1]] for i in range(len(group) - 1))
-                new_list.append([group[i], group[i + 1]])
 
     return new_list
 

torch/ao/quantization/backend_config/_common_operator_config_utils.py

@@ -155,13 +155,13 @@ def _get_binary_op_configs(
             (op_with_quantized_bop_scalar_variant, torch.relu),
             op_with_quantized_bop_scalar_variant,
         ]
-        for bop_pattern in bop_patterns:
+        binary_op_configs.extend(
-            binary_op_configs.append(
             BackendPatternConfig(bop_pattern)
             .set_dtype_configs(dtype_configs)  # noqa: E131
             ._set_num_tensor_args_to_observation_type(
                 num_tensor_args_to_observation_type_mapping
             )
+            for bop_pattern in bop_patterns
         )
     # matmul
     binary_op_configs.append(
@@ -502,7 +502,6 @@ def _get_ln_configs(dtype_configs: List[DTypeConfig]) -> List[BackendPatternConf
 def _get_default_op_configs(
     dtype_configs: List[DTypeConfig],
 ) -> List[BackendPatternConfig]:
-    configs = []
     default_ops = [
         torch.nn.ELU,
         torch.nn.LeakyReLU,
@@ -517,14 +516,14 @@ def _get_default_op_configs(
         torch.nn.functional.leaky_relu,
         torch.nn.functional.dropout,
     ]
-    for op in default_ops:
+    configs = [
-        configs.append(
         BackendPatternConfig(op)
         .set_observation_type(
             ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT
         )  # noqa: E131
         .set_dtype_configs(dtype_configs)
-        )
+        for op in default_ops
+    ]
 
     configs.append(
         BackendPatternConfig(torch.nn.functional.group_norm)

torch/ao/quantization/backend_config/_qnnpack_pt2e.py

@@ -159,13 +159,13 @@ def get_binary_op_configs():
             # TODO: remove when functionalization is supported in pt2_mode
             (op_with_quantized_bop_scalar_variant, torch.ops.aten.relu_.default),
         ]
-        for bop_pattern in bop_patterns:
+        binary_op_configs.extend(
-            binary_op_configs.append(
             BackendPatternConfig(bop_pattern)
             .set_dtype_configs(dtype_configs)  # noqa: E131
             ._set_num_tensor_args_to_observation_type(
                 num_tensor_args_to_observation_type_mapping
             )
+            for bop_pattern in bop_patterns
         )
 
     return binary_op_configs

torch/ao/quantization/backend_config/executorch.py

@@ -325,13 +325,13 @@ def _get_binary_ops_configs() -> List[BackendPatternConfig]:
             (op, torch.relu),
             op,
         ]
-        for bop_pattern in bop_patterns:
+        binary_op_configs.extend(
-            binary_op_configs.append(
             BackendPatternConfig(bop_pattern)
             .set_dtype_configs(dtype_configs)  # noqa: E131
             ._set_num_tensor_args_to_observation_type(
                 num_tensor_args_to_observation_type_mapping
             )
+            for bop_pattern in bop_patterns
         )
     return binary_op_configs
 
@@ -385,13 +385,12 @@ def _get_share_qparams_ops_configs() -> List[BackendPatternConfig]:
         "squeeze_",
         "leaky_relu",
     ]
-    share_qparams_op_configs: List[BackendPatternConfig] = []
+    share_qparams_op_configs: List[BackendPatternConfig] = [
-    for op in share_qparams_ops:
-        share_qparams_op_configs.append(
         BackendPatternConfig(op)
         .set_observation_type(observation_type)  # noqa: E131
         .set_dtype_configs(dtype_configs)
-        )
+        for op in share_qparams_ops
+    ]
     return share_qparams_op_configs
 
 

torch/ao/quantization/backend_config/utils.py

@@ -36,18 +36,20 @@ def get_pattern_to_dtype_configs(
 
 
 def get_qat_module_classes(backend_config: BackendConfig) -> Tuple[type, ...]:
-    qat_module_classes = []
+    qat_module_classes = [
-    for config in backend_config.configs:
+        config.qat_module
-        if config.qat_module is not None:
+        for config in backend_config.configs
-            qat_module_classes.append(config.qat_module)
+        if config.qat_module is not None
+    ]
     return tuple(set(qat_module_classes))
 
 
 def get_fused_module_classes(backend_config: BackendConfig) -> Tuple[type, ...]:
-    fused_module_classes = []
+    fused_module_classes = [
-    for config in backend_config.configs:
+        config.fused_module
-        if config.fused_module is not None:
+        for config in backend_config.configs
-            fused_module_classes.append(config.fused_module)
+        if config.fused_module is not None
+    ]
     return tuple(set(fused_module_classes))
 
 

torch/ao/quantization/fuse_modules.py

@@ -92,9 +92,7 @@ def _fuse_modules_helper(
     additional_fuser_method_mapping = fuse_custom_config_dict.get(
         "additional_fuser_method_mapping", {}
     )
-    mod_list = []
+    mod_list = [_get_module(model, item) for item in modules_to_fuse]
-    for item in modules_to_fuse:
-        mod_list.append(_get_module(model, item))
 
     # Fuse list of modules
     new_mod_list = fuser_func(mod_list, is_qat, additional_fuser_method_mapping)

torch/ao/quantization/fuser_method_mappings.py

@@ -266,9 +266,7 @@ def _get_valid_patterns(op_pattern):
     """
     result: List[Any]
     if isinstance(op_pattern, (tuple, list)):
-        sub_combs = []
+        sub_combs = [_get_valid_patterns(sub_pattern) for sub_pattern in op_pattern]
-        for sub_pattern in op_pattern:
-            sub_combs.append(_get_valid_patterns(sub_pattern))
         result = list(itertools.product(*sub_combs))
     else:
         result = [op_pattern, MatchAllNode]

torch/ao/quantization/fx/_model_report/model_report_visualizer.py

@@ -527,8 +527,9 @@ class ModelReportVisualizer:
             # gather the x_data and multiple y_data
             # calculate the number of channels
             num_channels: int = max(row[self.CHANNEL_NUM_INDEX] for row in table) + 1
-            for channel in range(num_channels):
+            y_data.extend(
-                y_data.append([])  # separate data list per channel
+                [] for channel in range(num_channels)
+            )  # separate data list per channel
 
             for table_row_num, row in enumerate(table):
                 # get x_value to append

torch/ao/quantization/fx/convert.py

@@ -1116,10 +1116,9 @@ def convert(
     # for dynamic quant ops or weight only quant ops
     _run_weight_observers(model, backend_config)
 
-    graph_inputs: List[str] = []
+    graph_inputs: List[str] = [
-    for node in model.graph.nodes:
+        node.name for node in model.graph.nodes if node.op == "placeholder"
-        if node.op == "placeholder":
+    ]
-            graph_inputs.append(node.name)
 
     # additional state to override inputs to be quantized, if specified
     # by the user

torch/ao/quantization/fx/fuse_handler.py

@@ -78,8 +78,7 @@ class DefaultFuseHandler(FuseHandler):
                 n, *args = pattern
                 modules: List[torch.nn.Module] = []
                 modules.append(get_modules(n))
-                for a in args:
+                modules.extend(get_modules(a) for a in args)
-                    modules.append(get_modules(a))
                 return tuple(modules)
             else:
                 n = pattern
@@ -111,9 +110,7 @@ class DefaultFuseHandler(FuseHandler):
         # as input
         fused_module = fuser_method(is_qat, *matched_modules)
         setattr(named_modules[module_parent_name], module_name, fused_module)
-        extra_args = []
+        extra_args = [load_arg(input) for input in extra_inputs]
-        for input in extra_inputs:
-            extra_args.append(load_arg(input))
         node = fused_graph.node_copy(root_node, load_arg)
         args = list(node.args)
         args.extend(extra_args)

torch/ao/quantization/fx/prepare.py

@@ -1219,13 +1219,12 @@ def _maybe_insert_observers_before_graph_output(
             else:
                 return maybe_node
         elif isinstance(maybe_node, (list, tuple)):
-            results = []
+            results = [
-            for inner_node in maybe_node:
-                results.append(
                 _recursive_maybe_replace_node_with_obs(
                     inner_node, model, named_modules, graph
                 )
-                )
+                for inner_node in maybe_node
+            ]
             if isinstance(maybe_node, list):
                 return results
             else:
@@ -1244,11 +1243,10 @@ def _maybe_insert_observers_before_graph_output(
                 "Unhandled type for returned node:", maybe_node
             )
 
-    new_args = []
+    new_args = [
-    for old_arg in graph_output_node.args:
-        new_args.append(
         _recursive_maybe_replace_node_with_obs(old_arg, model, named_modules, graph)
-        )
+        for old_arg in graph_output_node.args
+    ]
 
     graph_output_node.args = tuple(new_args)  # type: ignore[assignment]
 

torch/ao/quantization/pt2e/graph_utils.py

@@ -1,7 +1,7 @@
 # mypy: allow-untyped-defs
 import itertools
 import operator
-from typing import Any, Callable, List, Optional, OrderedDict, Set
+from typing import Any, Callable, List, Optional, OrderedDict, Sequence, Set
 
 import torch
 from torch.fx import Node
@@ -58,7 +58,7 @@ def update_equivalent_types_dict(customized_equivalent_types=None):
     _EQUIVALENT_TYPES_DICT = _create_equivalent_types_dict()
 
 
-def _partitions_sequential(partitions: List[SourcePartition]):
+def _partitions_sequential(partitions: Sequence[SourcePartition]):
     prev_partition = None
     for partition in partitions:
         if prev_partition is not None and not check_subgraphs_connected(
@@ -108,8 +108,9 @@ def find_sequential_partitions(
 
     typed_partitions_list = list(typed_partitions.values())
     fusion_candidates = itertools.product(*typed_partitions_list)
-    fused_partitions = []
+    fused_partitions = [
-    for candidate in fusion_candidates:
+        candidate
-        if _partitions_sequential(candidate):  # type: ignore[arg-type]
+        for candidate in fusion_candidates
-            fused_partitions.append(candidate)
+        if _partitions_sequential(candidate)
+    ]
     return fused_partitions

torch/ao/quantization/quantizer/xnnpack_quantizer.py

@@ -93,9 +93,9 @@ def _get_supported_symmetric_config_and_operators() -> List[OperatorConfig]:
         get_symmetric_quantization_config(is_per_channel=True, is_qat=True),
     ]:
         ops = _supported_symmetric_quantized_operators()
-        for pattern_list in ops.values():
+        supported_config_and_operators.extend(
-            supported_config_and_operators.append(
             OperatorConfig(quantization_config, pattern_list)
+            for pattern_list in ops.values()
         )
     return copy.deepcopy(supported_config_and_operators)
 

torch/autograd/gradcheck.py

@@ -55,10 +55,11 @@ def _allocate_jacobians_with_inputs(
     # of `t.numel` and width of `numel_output`. Otherwise, for each tensor, returns
     # a 1-d tensor with size `(t.numel,)`. Each new tensor will be strided and have
     # the same dtype and device as those of the corresponding input.
-    out: List[torch.Tensor] = []
+    out: List[torch.Tensor] = [
-    for t in input_tensors:
+        t.new_zeros((t.numel(), numel_output), layout=torch.strided)
-        if _is_float_or_complex_tensor(t) and t.requires_grad:
+        for t in input_tensors
-            out.append(t.new_zeros((t.numel(), numel_output), layout=torch.strided))
+        if _is_float_or_complex_tensor(t) and t.requires_grad
+    ]
     return tuple(out)
 
 
@@ -69,11 +70,12 @@ def _allocate_jacobians_with_outputs(
     # in `output_tensors`, returns a new zero-filled tensor with height of `dim` and
     # width of `t.numel`. Otherwise, for each tensor, returns a 1-d tensor with size
     # (t.numel,).
-    out: List[torch.Tensor] = []
     options = {"dtype": dtype, "device": device, "layout": torch.strided}
-    for t in output_tensors:
+    out: List[torch.Tensor] = [
-        if _is_float_or_complex_tensor(t):
+        t.new_zeros((numel_input, t.numel()), **options)
-            out.append(t.new_zeros((numel_input, t.numel()), **options))
+        for t in output_tensors
+        if _is_float_or_complex_tensor(t)
+    ]
     return tuple(out)
 
 
@@ -904,10 +906,10 @@ def _compute_analytical_jacobian_rows(
 def _get_analytical_vjps_wrt_specific_output(
     vjp_fn, sample_output, v
 ) -> List[List[Optional[torch.Tensor]]]:
-    vjps: List[List[Optional[torch.Tensor]]] = []
     grad_inputs = vjp_fn(v.reshape(sample_output.shape))
-    for vjp in grad_inputs:
+    vjps: List[List[Optional[torch.Tensor]]] = [
-        vjps.append([vjp.clone() if isinstance(vjp, torch.Tensor) else None])
+        [vjp.clone() if isinstance(vjp, torch.Tensor) else None] for vjp in grad_inputs
+    ]
     return vjps
 
 

torch/autograd/graph.py

@@ -802,9 +802,7 @@ def _register_logging_hooks_on_whole_graph(
         log_str = f"Executing: {node} with grad_outputs: {grad_outputs_str}"
         log.debug(log_str)
 
-    handles = []
+    handles = [node.register_prehook(prehook) for node in iter_graph(grad_fns)]
-    for node in iter_graph(grad_fns):
-        handles.append(node.register_prehook(prehook))
 
     def unregister_hooks() -> None:
         for handle in handles:

torch/autograd/profiler_util.py

@@ -856,10 +856,9 @@ def _build_table(
     flops_column_width = DEFAULT_COLUMN_WIDTH
 
     src_column_width = None
-    stacks = []
+    stacks = [
-    for evt in events:
+        evt.stack for evt in events if evt.stack is not None and len(evt.stack) > 0
-        if evt.stack is not None and len(evt.stack) > 0:
+    ]
-            stacks.append(evt.stack)
     has_stack = len(stacks) > 0
     if has_stack:
         src_column_width = (
@@ -947,10 +946,7 @@ def _build_table(
 
     if with_flops:
         # Auto-scaling of flops header
-        raw_flops = []
+        raw_flops = [evt.flops for evt in events if evt.flops > 0]
-        for evt in events:
-            if evt.flops > 0:
-                raw_flops.append(evt.flops)
         if len(raw_flops) != 0:
             (flops_scale, flops_header) = auto_scale_flops(min(raw_flops))
             headers.append(f"Total {flops_header}")

torch/cuda/__init__.py

@@ -322,9 +322,7 @@ def _lazy_init():
         # However, we must not let any *other* threads in!
         _tls.is_initializing = True
 
-        for calls in _lazy_seed_tracker.get_calls():
+        _queued_calls.extend(calls for calls in _lazy_seed_tracker.get_calls() if calls)
-            if calls:
-                _queued_calls.append(calls)
 
         try:
             for queued_call, orig_traceback in _queued_calls:

torch/cuda/random.py

@@ -44,9 +44,7 @@ def get_rng_state(device: Union[int, str, torch.device] = "cuda") -> Tensor:
 
 def get_rng_state_all() -> List[Tensor]:
     r"""Return a list of ByteTensor representing the random number states of all devices."""
-    results = []
+    results = [get_rng_state(i) for i in range(device_count())]
-    for i in range(device_count()):
-        results.append(get_rng_state(i))
     return results
 
 

torch/distributed/_shard/sharded_optim/api.py

@@ -31,8 +31,9 @@ class ShardedOptimizer(optim.Optimizer):
         tensors: List[Tensor] = []
         for value in named_params.values():
             if isinstance(value, ShardedTensor):
-                for local_shard in value.local_shards():
+                tensors.extend(
-                    tensors.append(local_shard.tensor)
+                    local_shard.tensor for local_shard in value.local_shards()
+                )
             else:
                 tensors.append(value)
 

torch/distributed/_shard/sharded_tensor/_ops/tensor_ops.py

@@ -99,9 +99,10 @@ def sharded_type_as(args, kwargs, pg):
     tensor = args[1]
     if isinstance(tensor, ShardedTensor):
         tensor = tensor.local_tensor()
-    new_local_shards = []
+    new_local_shards = [
-    for shard in st.local_shards():
+        Shard(shard.tensor.type_as(tensor), shard.metadata)
-        new_local_shards.append(Shard(shard.tensor.type_as(tensor), shard.metadata))
+        for shard in st.local_shards()
+    ]
     st_meta = copy.deepcopy(st._metadata)
     st_meta.tensor_properties.dtype = tensor.dtype
     return new_local_shards, st_meta

torch/distributed/_shard/sharded_tensor/reshard.py

@@ -191,9 +191,9 @@ def reshard_local_shard(
     )
 
     # Compute expected size
-    input_split_sizes = []
+    input_split_sizes = [
-    for metadata in shards_metadata:
+        metadata.shard_sizes[reshard_dim] for metadata in shards_metadata
-        input_split_sizes.append(metadata.shard_sizes[reshard_dim])
+    ]
     rearrange_input = any(ranks[i] > ranks[i + 1] for i in range(len(ranks) - 1))
 
     if rearrange_input:

torch/distributed/benchmarks/benchmark_ddp_rpc.py

@@ -142,8 +142,7 @@ def _run_trainer(emb_rref_list, rank):
         )
 
     # model.parameters() only includes local parameters.
-    for param in model.parameters():
+    model_parameter_rrefs.extend(RRef(param) for param in model.parameters())
-        model_parameter_rrefs.append(RRef(param))
 
     # Setup distributed optimizer
     opt = DistributedOptimizer(optim.SGD, model_parameter_rrefs, lr=0.05)

torch/distributed/checkpoint/planner_helpers.py

@@ -207,8 +207,10 @@ def _create_default_metadata_only_plan(state_dict: STATE_DICT_TYPE) -> SavePlan:
         if isinstance(obj, DTensor):
             requests.append(_create_write_items_for_dtensor(fqn, obj))
         elif isinstance(obj, ShardedTensor):
-            for shard_md in obj.metadata().shards_metadata:
+            requests.extend(
-                requests.append(_create_write_item_for_shard(fqn, obj, shard_md))
+                _create_write_item_for_shard(fqn, obj, shard_md)
+                for shard_md in obj.metadata().shards_metadata
+            )
         elif isinstance(obj, torch.Tensor):
             requests.append(_create_write_item_for_tensor(fqn, obj))
         else:

torch/distributed/distributed_c10d.py

@@ -1371,9 +1371,9 @@ def _get_all_pg_configs() -> List[Dict[str, Any]]:
     Return the pg configuration of all the process groups.
 
     """
-    config_info: List[Dict[str, Any]] = []
+    config_info: List[Dict[str, Any]] = [
-    for pg in _world.pg_map.keys():
+        _get_pg_config(pg) for pg in _world.pg_map.keys()
-        config_info.append(_get_pg_config(pg))
+    ]
     return config_info
 
 
@@ -2508,9 +2508,7 @@ def _coalescing_manager(
         # - coalesced `reduce_scatter_tensor`
         op0 = op_list[0].op
         if op0 == all_reduce:
-            tensors = []
+            tensors = [op.tensor for op in op_list]
-            for op in op_list:
-                tensors.append(op.tensor)
             all_reduce_opts = AllreduceCoalescedOptions()
             all_reduce_opts.reduceOp = not_none(op_list[0].redop)
             work = group.allreduce_coalesced(tensors, all_reduce_opts)

torch/distributed/fsdp/_runtime_utils.py

@@ -568,10 +568,11 @@ def _root_pre_forward(
             state._needs_buffer_dtype_restore_check = False
 
         if state.forward_prefetch:
-            handles = []
+            handles = [
-            for fsdp_state in state._all_fsdp_states:
+                fsdp_state._handle
-                if fsdp_state._handle:
+                for fsdp_state in state._all_fsdp_states
-                    handles.append(fsdp_state._handle)
+                if fsdp_state._handle
+            ]
             for handle in handles:
                 handle._needs_pre_forward_unshard = True
                 handle._prefetched = False

torch/distributed/nn/api/remote_module.py

@@ -110,9 +110,9 @@ def _create_module_with_interface(
 
 
 def _param_rrefs(module_rref, recurse) -> List[rpc.RRef[Parameter]]:
-    ret: List[rpc.RRef[Parameter]] = []
+    ret: List[rpc.RRef[Parameter]] = [
-    for param in module_rref.local_value().parameters(recurse):
+        rpc.RRef(param) for param in module_rref.local_value().parameters(recurse)
-        ret.append(rpc.RRef(param))
+    ]
     return ret
 
 

torch/distributed/optim/named_optimizer.py

@@ -146,9 +146,7 @@ class _NamedOptimizer(optim.Optimizer):
 
         ret_groups = []
         for group in param_groups:
-            param_keys = []
+            param_keys = [self.ordered_param_keys[param] for param in group["params"]]
-            for param in group["params"]:
-                param_keys.append(self.ordered_param_keys[param])
             ret_group = {"params": sorted(param_keys)}
             for k, v in group.items():
                 if k != "params":

torch/distributed/optim/optimizer.py

@@ -245,13 +245,12 @@ class DistributedOptimizer:
             else _local_optimizer_step
         )
 
-        rpc_futs = []
+        rpc_futs = [
-        for optimizer in self.remote_optimizers:
-            rpc_futs.append(
             rpc.rpc_async(
                 optimizer.owner(),
                 optimizer_step_func,
                 args=(optimizer, context_id),
             )
-            )
+            for optimizer in self.remote_optimizers
+        ]
         _wait_for_all(rpc_futs)

torch/distributed/optim/zero_redundancy_optimizer.py

@@ -781,14 +781,14 @@ class ZeroRedundancyOptimizer(Optimizer, Joinable):
                 self.process_group, rank
             )
             for param_group in param_groups:
-                for param in param_group["params"]:
+                handles.extend(
-                    handles.append(
                     dist.broadcast(
                         tensor=param.data,
                         src=global_rank,
                         group=self.process_group,
                         async_op=True,
                     )
+                    for param in param_group["params"]
                 )
         return handles
 

torch/distributed/pipelining/microbatch.py

@@ -224,9 +224,7 @@ def _shard_dict_of_args(
     for chunk_idx in range(real_num_chunks):
         chunk_args = {}
         for key, arg in args_sharded_replicated.items():
-            arg_single_chunk = []
+            arg_single_chunk = [v_flat[chunk_idx] for v_flat in arg]
-            for v_flat in arg:
-                arg_single_chunk.append(v_flat[chunk_idx])
             chunk_args[key] = arg_single_chunk
         chunks_flat.append(chunk_args)
 
@@ -340,9 +338,10 @@ def split_args_kwargs_into_chunks(
             f"{len(args_split_dict)}, {len(kwargs_split)}"
         )
 
-    args_split = []
+    args_split = [
-    for chunk_args in args_split_dict:
+        tuple(chunk_args[i] for i in range(len(chunk_args)))
-        args_split.append(tuple(chunk_args[i] for i in range(len(chunk_args))))
+        for chunk_args in args_split_dict
+    ]
 
     return args_split, kwargs_split
 

torch/distributed/pipelining/schedules.py

@@ -1702,15 +1702,13 @@ class ScheduleLoopedBFS(PipelineScheduleMulti):
         )
 
         # Store the list of operations used for that rank
-        rank_ops: List[Optional[_Action]] = []
         # Pre-padding, rank starts with no-ops based on the warmup.
-        for _ in range(rank):
+        rank_ops: List[Optional[_Action]] = [None for _ in range(rank)]
-            rank_ops.append(None)
 
         for stage_index in stage_indices:
-            for mb_index in range(self._n_microbatches):
+            rank_ops.extend(
-                rank_ops.append(
                 _Action(stage_index, _ComputationType.FORWARD, mb_index)
+                for mb_index in range(self._n_microbatches)
             )
 
         # wait for the first backward to trickle up
@@ -1719,9 +1717,9 @@ class ScheduleLoopedBFS(PipelineScheduleMulti):
         rank_ops.extend([None] * post_warmup_ops)
 
         for stage_index in reversed(stage_indices):
-            for mb_index in reversed(range(self._n_microbatches)):
+            rank_ops.extend(
-                rank_ops.append(
                 _Action(stage_index, _ComputationType.FULL_BACKWARD, mb_index)
+                for mb_index in reversed(range(self._n_microbatches))
             )
         return rank_ops
 
@@ -1744,10 +1742,8 @@ def _get_1f1b_rank_ops(
     weight_stage_mb_index: Dict[int, int] = defaultdict(int)
 
     # Store the list of operations used for that rank
-    rank_ops: List[Optional[_Action]] = []
     # Pre-padding, rank starts with no-ops based on the warmup.
-    for _ in range(rank):
+    rank_ops: List[Optional[_Action]] = [None for _ in range(rank)]
-        rank_ops.append(None)
     # These are used to calculate the number of slots to fill with no-ops, to account for the delay in warmup
     # when we want to wait for the backward to trickle back up and start 1f1b to align all ranks.
     # Formula:

torch/distributed/tensor/_collective_utils.py

@@ -195,8 +195,7 @@ def fill_empty_tensor_to_shards(
         size if idx != shard_dim else 0 for idx, size in enumerate(tensor_size)
     ]
     tensor = shards[0].new_zeros(tensor_size)
-    for _ in range(num_empty_tensors):
+    shards.extend(tensor for _ in range(num_empty_tensors))
-        shards.append(tensor)
     return shards
 
 

torch/distributed/tensor/_ops/_einsum_strategy.py

@@ -167,9 +167,7 @@ def gen_einsum_strategies(
     # (i.e. for Shard, tensor dim size must > mesh size)
     all_strategies = []
     for strategy_comb in strategy_combs:
-        spec_list = []
+        spec_list = [DTensorSpec(mesh, tuple(specs)) for specs in zip(*strategy_comb)]
-        for specs in zip(*strategy_comb):
-            spec_list.append(DTensorSpec(mesh, tuple(specs)))
         strat = PlacementStrategy(output_specs=spec_list[0], input_specs=spec_list[1:])
         all_strategies.append(strat)
 

torch/distributed/tensor/_ops/_tensor_ops.py

@@ -43,18 +43,17 @@ def default_strategy(mesh: DeviceMesh, op_schema: OpSchema) -> StrategyType:
     # Default strategy by default just propagate the first input strategy
     select_strategy = op_schema.args_schema[0]
     assert isinstance(select_strategy, OpStrategy)
-    default_strategy = []
-    for strategy in select_strategy.strategies:
     # we create new DTensorSpecs even for default strategy to assure that
     # the tensor metas are distinct between the arguments and outputs
-        default_strategy.append(
+    default_strategy = [
         PlacementStrategy(
             output_specs=DTensorSpec(
                 mesh=strategy.output_spec.mesh,
                 placements=strategy.output_spec.placements,
             )
         )
-        )
+        for strategy in select_strategy.strategies
+    ]
     return OpStrategy(default_strategy)
 
 

torch/distributed/tensor/_ops/utils.py

@@ -209,9 +209,10 @@ def map_placements_after_broadcast(
 def generate_redistribute_costs(
     src_strategy: OpStrategy, dst_spec: DTensorSpec
 ) -> List[float]:
-    redistribute_costs: List[float] = []
+    redistribute_costs: List[float] = [
-    for strat in src_strategy.strategies:
+        redistribute_cost(strat.output_spec, dst_spec)
-        redistribute_costs.append(redistribute_cost(strat.output_spec, dst_spec))
+        for strat in src_strategy.strategies
+    ]
 
     return redistribute_costs
 

torch/export/unflatten.py

@@ -152,9 +152,11 @@ class InterpreterModule(torch.nn.Module):
                 # the keys in the kwarg dict.
                 arg_list = list(args)
                 kwarg_names = self.arg_names[len(arg_list) :]
-                for kwarg_name in kwarg_names:
+                arg_list.extend(
-                    if kwarg_name in kwargs:
+                    kwargs[kwarg_name]
-                        arg_list.append(kwargs[kwarg_name])
+                    for kwarg_name in kwarg_names
+                    if kwarg_name in kwargs
+                )
 
                 # Assert that the kwargs passed in exactly match the positional
                 # arguments specified by the GraphModule. This should be
@@ -933,9 +935,10 @@ class _ModuleFrame:
             assert kwargs_spec.context is not None
 
             with self.graph.inserting_after(None):
-                arg_nodes = []
+                arg_nodes = [
-                for idx in range(args_spec.num_children):
+                    self.graph.placeholder(f"_positional_arg_{idx}")
-                    arg_nodes.append(self.graph.placeholder(f"_positional_arg_{idx}"))
+                    for idx in range(args_spec.num_children)
+                ]
                 kwarg_nodes = {}
                 for name in kwargs_spec.context:
                     kwarg_nodes[name] = self.graph.placeholder(name)

torch/fx/experimental/accelerator_partitioner.py

@@ -1015,10 +1015,11 @@ class Partitioner:
         # Keep tracking the partition pair of node pair
         partition_pair: List[Partition] = []
         # Collect all the op nodes from the graph
-        op_nodes = []
+        op_nodes = [
-        for n in self.graph_module.graph.nodes:
+            n
-            if n.op not in {"placeholder", "get_attr", "output"}:
+            for n in self.graph_module.graph.nodes
-                op_nodes.append(n)
+            if n.op not in {"placeholder", "get_attr", "output"}
+        ]
         for node in op_nodes:
             # Find which partition the current node belongs
             p0_index = self.node_to_partition[node]

torch/fx/experimental/migrate_gradual_types/constraint_transformation.py

@@ -648,11 +648,7 @@ def generate_reshape(constraint, counter):
 
     # then there must be exactly one occurrence of dyn
     else:
-        new_target = []
+        new_target = [n for n in target if n != Dyn]
-
-        for n in target:
-            if n != Dyn:
-                new_target.append(n)
 
         # tensor 1
         c3_tensor1 = Disj(
@@ -886,10 +882,8 @@ def generate_all_int_dyn_dim_possibilities(my_list: List[DVar]):
     neq_possibilities = [
         BinConstraintD(my_list[i], Dyn, op_neq) for i in range(len(my_list))
     ]
-    d_possibilities = []
 
-    for i in zip(eq_possibilities, neq_possibilities):
+    d_possibilities = [list(i) for i in zip(eq_possibilities, neq_possibilities)]
-        d_possibilities.append(list(i))
     all_possibilities = list(itertools.product(*d_possibilities))
     return all_possibilities
 
@@ -1043,13 +1037,11 @@ def apply_padding(
 
         assert len(simulate_padding + d1) == len(d2)
 
-        broadcast_padding = []
-
         # for every padding size, we also consider broadcasting
-        for j in range(len(d2) - i):
+        broadcast_padding = [
-            broadcast_padding.append(
             broadcast_dim(simulate_padding, d2, d11, d12, j, True)
-            )
+            for j in range(len(d2) - i)
+        ]
 
         # we consider the possibilities for broadcasting for every dimension. Since we already
         # padded d1, we do not consider it while broadcasting

torch/fx/experimental/partitioner_utils.py

@@ -302,11 +302,10 @@ def get_latency_of_partitioned_graph(
         """This function is to return all the partitions without parents
         as the starting points of all the paths
         """
-        top_partitions = []
-        for partition in partitions:
         # If a partition has no parents, then it is a top partition
-            if len(partition.parents) == 0:
+        top_partitions = [
-                top_partitions.append(partition)
+            partition for partition in partitions if len(partition.parents) == 0
+        ]
         return top_partitions
 
     top_partitions = get_top_partitions(partitions)

torch/fx/experimental/symbolic_shapes.py

@@ -5201,10 +5201,9 @@ class ShapeEnv:
         symints = {
             s.node.expr for s in symints if isinstance(s.node.expr, sympy.Symbol)
         }
-        guards = []
+        guards = [
-        for g in self.guards:
+            g for g in self.guards if all(s in symints for s in g.expr.free_symbols)
-            if all(s in symints for s in g.expr.free_symbols):
+        ]
-                guards.append(g)
         return guards
 
     def bind_symbols(

torch/fx/passes/pass_manager.py

@@ -220,10 +220,7 @@ class PassManager:
     def remove_pass(self, _passes: List[str]):
         if _passes is None:
             return
-        passes_left = []
+        passes_left = [ps for ps in self.passes if ps.__name__ not in _passes]
-        for ps in self.passes:
-            if ps.__name__ not in _passes:
-                passes_left.append(ps)
         self.passes = passes_left
         self._validated = False
 

torch/jit/_recursive.py

@@ -881,9 +881,7 @@ def infer_methods_to_compile(nn_module):
         uniqued_methods.append(name)
         uniquer.add(name)
 
-    stubs = []
+    stubs = [make_stub_from_method(nn_module, method) for method in uniqued_methods]
-    for method in uniqued_methods:
-        stubs.append(make_stub_from_method(nn_module, method))
     return overload_stubs + stubs
 
 
@@ -959,9 +957,10 @@ def interface_script(mod_interface, nn_module):
 
         It is used to know which methods need to act as starting points for compilation.
         """
-        stubs = []
+        stubs = [
-        for method in mod_interface.getMethodNames():
+            make_stub_from_method(nn_module, method)
-            stubs.append(make_stub_from_method(nn_module, method))
+            for method in mod_interface.getMethodNames()
+        ]
         return stubs
 
     return create_script_module(nn_module, infer_interface_methods_to_compile)

torch/jit/_script.py

@@ -1493,11 +1493,10 @@ def _get_overloads(obj):
             _jit_internal.get_overload_no_implementation_error_message("function", obj)
         )
 
-    compiled_fns = []
+    compiled_fns = [
-    for overload_fn in uncompiled_overloads:
-        compiled_fns.append(
         _compile_function_with_overload(overload_fn, qual_name, obj)
-        )
+        for overload_fn in uncompiled_overloads
+    ]
 
     if existing_compiled_fns:
         compiled_fns = existing_compiled_fns + compiled_fns

torch/mtia/__init__.py

@@ -77,9 +77,7 @@ def _lazy_init() -> None:
         # However, we must not let any *other* threads in!
         _tls.is_initializing = True
 
-        for calls in _lazy_seed_tracker.get_calls():
+        _queued_calls.extend(calls for calls in _lazy_seed_tracker.get_calls() if calls)
-            if calls:
-                _queued_calls.append(calls)
 
         try:
             for queued_call, orig_traceback in _queued_calls:

torch/nn/parallel/_functions.py

@@ -23,8 +23,7 @@ class Broadcast(Function):
         non_differentiables = []
         for idx, input_requires_grad in enumerate(ctx.needs_input_grad[1:]):
             if not input_requires_grad:
-                for output in outputs:
+                non_differentiables.extend(output[idx] for output in outputs)
-                    non_differentiables.append(output[idx])
         ctx.mark_non_differentiable(*non_differentiables)
         return tuple([t for tensors in outputs for t in tensors])
 

torch/nn/utils/_expanded_weights/conv_utils.py

@@ -142,9 +142,7 @@ def conv_backward(func, ctx, grad_output):
         ctx.groups,
     )
 
-    kernel_size = []
+    kernel_size = [weight_shape[i] for i in range(2, conv_picker(func, 3, 4, 5))]
-    for i in range(2, conv_picker(func, 3, 4, 5)):
-        kernel_size.append(weight_shape[i])
 
     batch_size = ctx.batch_size
     results: List[Optional[torch.Tensor]] = []

torch/onnx/symbolic_helper.py

@@ -356,12 +356,14 @@ def quantized_args(
                 return descriptor and _is_value(arg) and _is_tuple_construct(arg)
 
             # Run regular symbolic function if none of the argument is QTensor.
-            is_quantized = []
+            is_quantized: typing.List[bool] = []
             for descriptor, arg in descriptor_args:
                 # ListConstruct
                 if _is_packed_list(arg):
-                    for arg_input in arg.node().inputs():
+                    is_quantized.extend(
-                        is_quantized.append(_is_arg_quantized(descriptor, arg_input))
+                        _is_arg_quantized(descriptor, arg_input)
+                        for arg_input in arg.node().inputs()
+                    )
                 else:
                     is_quantized.append(_is_arg_quantized(descriptor, arg))
 

torch/onnx/symbolic_opset18.py

@@ -67,10 +67,9 @@ def col2im(
     stride: Sequence[int],
 ):
     # convert [i0, i1, ..., in] into [i0, i0, i1, i1, ..., in, in]
-    adjusted_padding = []
+    adjusted_padding: List[int] = []
     for pad in padding:
-        for _ in range(2):
+        adjusted_padding.extend(pad for _ in range(2))
-            adjusted_padding.append(pad)
 
     num_dimensional_axis = symbolic_helper._get_tensor_sizes(output_size)[0]
     if not adjusted_padding:

torch/onnx/verification.py

@@ -1397,8 +1397,6 @@ class GraphInfo:
         original_outputs = list(graph.outputs())
         original_inputs = list(graph.inputs())
 
-        new_outputs = []
-
         def _process_bridge_value_for_lower(
             graph: torch.Graph, bridge_value: torch.Value
         ) -> torch.Value:
@@ -1416,9 +1414,9 @@ class GraphInfo:
             graph, pivot, process_bridge_value_for_lower
         )
 
-        for output in original_outputs:
+        new_outputs = [
-            if _produced_by(output, lower_nodes):
+            output for output in original_outputs if _produced_by(output, lower_nodes)
-                new_outputs.append(output)
+        ]
         for _ in enumerate(original_outputs):
             graph.eraseOutput(0)
         for output in new_outputs:

torch/package/_directory_reader.py

@@ -50,10 +50,11 @@ class DirectoryReader:
     def get_all_records(
         self,
     ):
-        files = []
+        files = [
-        for filename in glob(f"{self.directory}/**", recursive=True):
+            filename[len(self.directory) + 1 :]
-            if not os.path.isdir(filename):
+            for filename in glob(f"{self.directory}/**", recursive=True)
-                files.append(filename[len(self.directory) + 1 :])
+            if not os.path.isdir(filename)
+        ]
         return files
 
     def serialization_id(

torch/package/find_file_dependencies.py

@@ -53,16 +53,16 @@ class _ExtractModuleReferences(ast.NodeVisitor):
         if hasattr(node.func, "id") and node.func.id == "__import__":
             try:
                 name = self._grab_node_str(node.args[0])
-                fromlist = []
+                fromlist: List[str] = []
                 level = 0
                 if len(node.args) > 3:
-                    for v in node.args[3].elts:
+                    fromlist.extend(self._grab_node_str(v) for v in node.args[3].elts)
-                        fromlist.append(self._grab_node_str(v))
                 elif hasattr(node, "keywords"):
                     for keyword in node.keywords:
                         if keyword.arg == "fromlist":
-                            for v in keyword.value.elts:
+                            fromlist.extend(
-                                fromlist.append(self._grab_node_str(v))
+                                self._grab_node_str(v) for v in keyword.value.elts
+                            )
                 if len(node.args) > 4:
                     level = self._grab_node_int(node.args[4])
                 elif hasattr(node, "keywords"):

torch/profiler/_pattern_matcher.py

@@ -97,10 +97,9 @@ class Pattern:
     def matched_events(self):
         if self.skip:
             return []
-        matched_events = []
+        matched_events = [
-        for event in self.eventTreeTraversal():
+            event for event in self.eventTreeTraversal() if self.match(event)
-            if self.match(event):
+        ]
-                matched_events.append(event)
         return matched_events
 
     def root_of(self, event: _ProfilerEvent):