Use pyupgrade --py39-plus to improve code

Author: cyyever

src/transformers/cache_utils.py

@@ -90,7 +90,7 @@ def get_relative_imports(module_file: Union[str, os.PathLike]) -> list[str]:
         module_file (`str` or `os.PathLike`): The module file to inspect.
 
     Returns:
-        `List[str]`: The list of relative imports in the module.
+        `list[str]`: The list of relative imports in the module.
     """
     with open(module_file, encoding="utf-8") as f:
         content = f.read()
@@ -112,7 +112,7 @@ def get_relative_import_files(module_file: Union[str, os.PathLike]) -> list[str]
         module_file (`str` or `os.PathLike`): The module file to inspect.
 
     Returns:
-        `List[str]`: The list of all relative imports a given module needs (recursively), which will give us the list
+        `list[str]`: The list of all relative imports a given module needs (recursively), which will give us the list
         of module files a given module needs.
     """
     no_change = False
@@ -144,7 +144,7 @@ def get_imports(filename: Union[str, os.PathLike]) -> list[str]:
         filename (`str` or `os.PathLike`): The module file to inspect.
 
     Returns:
-        `List[str]`: The list of all packages required to use the input module.
+        `list[str]`: The list of all packages required to use the input module.
     """
     with open(filename, encoding="utf-8") as f:
         content = f.read()
@@ -175,7 +175,7 @@ def check_imports(filename: Union[str, os.PathLike]) -> list[str]:
         filename (`str` or `os.PathLike`): The module file to check.
 
     Returns:
-        `List[str]`: The list of relative imports in the file.
+        `list[str]`: The list of relative imports in the file.
     """
     imports = get_imports(filename)
     missing_packages = []

src/transformers/dynamic_module_utils.py

@@ -1,4 +1,3 @@
-# coding=utf-8
 # Copyright 2025 The HuggingFace Inc. team.
 # All rights reserved.
 #

src/transformers/model_debugging_utils.py

@@ -1,5 +1,4 @@
 #!/usr/bin/env python
-# coding=utf-8
 
 # Copyright 2021 The HuggingFace Inc. team. All rights reserved.
 #
@@ -16,7 +15,6 @@
 # limitations under the License.
 
 from functools import lru_cache
-from typing import FrozenSet
 
 from huggingface_hub import get_full_repo_name  # for backward compatibility
 from huggingface_hub.constants import HF_HUB_DISABLE_TELEMETRY as DISABLE_TELEMETRY  # for backward compatibility
@@ -299,8 +297,8 @@ def check_min_version(min_version):
         )
 
 
-@lru_cache()
-def get_available_devices() -> FrozenSet[str]:
+@lru_cache
+def get_available_devices() -> frozenset[str]:
     """
     Returns a frozenset of devices available for the current PyTorch installation.
     """

src/transformers/modeling_outputs.py

@@ -1,4 +1,3 @@
-# coding=utf-8
 # Copyright 2025 The HuggingFace Inc. team.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");

src/transformers/modeling_utils.py

@@ -1,4 +1,3 @@
-# coding=utf-8
 # Copyright 2023 The HuggingFace Inc. team.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -17,7 +16,8 @@
 
 import enum
 import inspect
-from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, Union
+from collections.abc import Iterable
+from typing import TYPE_CHECKING, Optional, Union
 
 
 if TYPE_CHECKING:
@@ -75,9 +75,9 @@ def verify_out_features_out_indices(
 
 
 def _align_output_features_output_indices(
-    out_features: Optional[List[str]],
-    out_indices: Optional[Union[List[int], Tuple[int]]],
-    stage_names: List[str],
+    out_features: Optional[list[str]],
+    out_indices: Optional[Union[list[int], tuple[int]]],
+    stage_names: list[str],
 ):
     """
     Finds the corresponding `out_features` and `out_indices` for the given `stage_names`.
@@ -106,10 +106,10 @@ def _align_output_features_output_indices(
 
 
 def get_aligned_output_features_output_indices(
-    out_features: Optional[List[str]],
-    out_indices: Optional[Union[List[int], Tuple[int]]],
-    stage_names: List[str],
-) -> Tuple[List[str], List[int]]:
+    out_features: Optional[list[str]],
+    out_indices: Optional[Union[list[int], tuple[int]]],
+    stage_names: list[str],
+) -> tuple[list[str], list[int]]:
     """
     Get the `out_features` and `out_indices` so that they are aligned.
 
@@ -198,7 +198,7 @@ def out_features(self):
         return self._out_features
 
     @out_features.setter
-    def out_features(self, out_features: List[str]):
+    def out_features(self, out_features: list[str]):
         """
         Set the out_features attribute. This will also update the out_indices attribute to match the new out_features.
         """
@@ -211,7 +211,7 @@ def out_indices(self):
         return self._out_indices
 
     @out_indices.setter
-    def out_indices(self, out_indices: Union[Tuple[int], List[int]]):
+    def out_indices(self, out_indices: Union[tuple[int], list[int]]):
         """
         Set the out_indices attribute. This will also update the out_features attribute to match the new out_indices.
         """
@@ -264,7 +264,7 @@ def out_features(self):
         return self._out_features
 
     @out_features.setter
-    def out_features(self, out_features: List[str]):
+    def out_features(self, out_features: list[str]):
         """
         Set the out_features attribute. This will also update the out_indices attribute to match the new out_features.
         """
@@ -277,7 +277,7 @@ def out_indices(self):
         return self._out_indices
 
     @out_indices.setter
-    def out_indices(self, out_indices: Union[Tuple[int], List[int]]):
+    def out_indices(self, out_indices: Union[tuple[int], list[int]]):
         """
         Set the out_indices attribute. This will also update the out_features attribute to match the new out_indices.
         """

src/transformers/tokenization_utils_base.py

@@ -19,7 +19,7 @@
 from contextlib import contextmanager
 from datetime import datetime
 from functools import lru_cache
-from typing import Any, Callable, Dict, List, Optional, Tuple, Union, get_args, get_origin, get_type_hints
+from typing import Any, Callable, Optional, Union, get_args, get_origin, get_type_hints
 
 from packaging import version
 
@@ -71,7 +71,7 @@ class DocstringParsingException(Exception):
     pass
 
 
-def _get_json_schema_type(param_type: str) -> Dict[str, str]:
+def _get_json_schema_type(param_type: str) -> dict[str, str]:
     type_mapping = {
         int: {"type": "integer"},
         float: {"type": "number"},
@@ -87,7 +87,7 @@ def _get_json_schema_type(param_type: str) -> Dict[str, str]:
     return type_mapping.get(param_type, {"type": "object"})
 
 
-def _parse_type_hint(hint: str) -> Dict:
+def _parse_type_hint(hint: str) -> dict:
     origin = get_origin(hint)
     args = get_args(hint)
 
@@ -152,7 +152,7 @@ def _parse_type_hint(hint: str) -> Dict:
     raise TypeHintParsingException("Couldn't parse this type hint, likely due to a custom class or object: ", hint)
 
 
-def _convert_type_hints_to_json_schema(func: Callable) -> Dict:
+def _convert_type_hints_to_json_schema(func: Callable) -> dict:
     type_hints = get_type_hints(func)
     signature = inspect.signature(func)
     required = []
@@ -173,7 +173,7 @@ def _convert_type_hints_to_json_schema(func: Callable) -> Dict:
     return schema
 
 
-def parse_google_format_docstring(docstring: str) -> Tuple[Optional[str], Optional[Dict], Optional[str]]:
+def parse_google_format_docstring(docstring: str) -> tuple[Optional[str], Optional[dict], Optional[str]]:
     """
     Parses a Google-style docstring to extract the function description,
     argument descriptions, and return description.
@@ -206,7 +206,7 @@ def parse_google_format_docstring(docstring: str) -> Tuple[Optional[str], Option
     return description, args_dict, returns
 
 
-def get_json_schema(func: Callable) -> Dict:
+def get_json_schema(func: Callable) -> dict:
     """
     This function generates a JSON schema for a given function, based on its docstring and type hints. This is
     mostly used for passing lists of tools to a chat template. The JSON schema contains the name and description of
@@ -398,7 +398,7 @@ def is_active(self) -> bool:
             return self._rendered_blocks or self._generation_indices
 
         @contextmanager
-        def activate_tracker(self, rendered_blocks: List[int], generation_indices: List[int]):
+        def activate_tracker(self, rendered_blocks: list[int], generation_indices: list[int]):
             try:
                 if self.is_active():
                     raise ValueError("AssistantTracker should not be reused before closed")

src/transformers/utils/__init__.py

@@ -1,4 +1,3 @@
-# coding=utf-8
 # Copyright 2021 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -24,7 +23,7 @@
 import random
 import sys
 import warnings
-from typing import Any, Callable, Dict, List, Literal, Optional, Tuple, Type, Union
+from typing import Any, Callable, Literal, Optional, Union
 
 import torch
 import torch.utils._pytree as pytree
@@ -78,9 +77,9 @@
 
 
 def _generate_supported_model_class_names(
-    model_name: Type[PretrainedConfig],
-    supported_tasks: Optional[Union[str, List[str]]] = None,
-) -> List[str]:
+    model_name: type[PretrainedConfig],
+    supported_tasks: Optional[Union[str, list[str]]] = None,
+) -> list[str]:
     task_mapping = {
         "default": MODEL_MAPPING_NAMES,
         "pretraining": MODEL_FOR_PRETRAINING_MAPPING_NAMES,
@@ -588,7 +587,7 @@ def to_concrete(t):
     return operator.getitem(a, b)
 
 
-_MANUAL_META_OVERRIDES: Dict[Callable, Callable] = {
+_MANUAL_META_OVERRIDES: dict[Callable, Callable] = {
     torch.nn.Embedding: torch_nn_embedding,
     torch.nn.functional.embedding: torch_nn_functional_embedding,
     torch.nn.LayerNorm: torch_nn_layernorm,
@@ -714,7 +713,7 @@ class HFCacheProxy(HFProxy):
     Proxy that represents an instance of `transformers.cache_utils.Cache`.
     """
 
-    def install_orig_cache_cls(self, orig_cache_cls: Type[Cache]):
+    def install_orig_cache_cls(self, orig_cache_cls: type[Cache]):
         self._orig_cache_cls = orig_cache_cls
 
     @property
@@ -768,8 +767,8 @@ class HFProxyableClassMeta(type):
     def __new__(
         cls,
         name: str,
-        bases: Tuple[Type, ...],
-        attrs: Dict[str, Any],
+        bases: tuple[type, ...],
+        attrs: dict[str, Any],
         proxy_factory_fn: Optional[Callable[[Node], Proxy]] = None,
     ):
         cls = super().__new__(cls, name, bases, attrs)
@@ -792,7 +791,7 @@ def __new__(
         return cls
 
 
-def gen_constructor_wrapper(target: Callable) -> Tuple[Callable, Callable]:
+def gen_constructor_wrapper(target: Callable) -> tuple[Callable, Callable]:
     """
     Wraps `target` to be proxyable. Used for tensor creators like `torch.ones`, `torch.arange` and so on.
     """
@@ -811,7 +810,7 @@ def _proxies_to_metas(v):
     return v
 
 
-def create_cache_proxy_factory_fn(orig_cache_cls: Type[Cache]) -> Callable[[Node], HFCacheProxy]:
+def create_cache_proxy_factory_fn(orig_cache_cls: type[Cache]) -> Callable[[Node], HFCacheProxy]:
     def cache_proxy_factory_fn(n: Node) -> HFCacheProxy:
         global _CURRENT_TRACER
         if not isinstance(_CURRENT_TRACER, HFTracer):
@@ -847,7 +846,7 @@ def cache_proxy_factory_fn(n: Node) -> HFCacheProxy:
 )
 
 
-def _generate_random_int(low: int = 10, high: int = 20, forbidden_values: Optional[List[int]] = None):
+def _generate_random_int(low: int = 10, high: int = 20, forbidden_values: Optional[list[int]] = None):
     if forbidden_values is None:
         forbidden_values = []
     value = random.randint(low, high)
@@ -897,8 +896,8 @@ def __init__(self, autowrap_modules=(math,), autowrap_functions=()):
             )
 
     def _generate_dummy_input(
-        self, model: "PreTrainedModel", input_name: str, shape: List[int], input_names: List[str]
-    ) -> Dict[str, torch.Tensor]:
+        self, model: "PreTrainedModel", input_name: str, shape: list[int], input_names: list[str]
+    ) -> dict[str, torch.Tensor]:
         """Generates dummy input for model inference recording."""
         # Retrieving the model class, either from the "class_for_deserialization" attribute if the model was restored
         # from pickle, or from the "__class__" attribute in the general case.
@@ -1179,7 +1178,7 @@ def maybe_get_proxy_for_attr(attr_val, collection_to_search, parameter_proxy_cac
             return attr_val
 
     # Needed for PyTorch 1.13+
-    def getattr(self, attr: str, attr_val: Any, parameter_proxy_cache: Dict[str, Any]):
+    def getattr(self, attr: str, attr_val: Any, parameter_proxy_cache: dict[str, Any]):
         return self._module_getattr(attr, attr_val, parameter_proxy_cache)
 
     def call_module(self, m, forward, args, kwargs):
@@ -1231,8 +1230,8 @@ def patch_for_tracing(self, root: Union[torch.nn.Module, Callable[..., Any]]):
     def trace(
         self,
         root: Union[torch.nn.Module, Callable[..., Any]],
-        concrete_args: Optional[Dict[str, Any]] = None,
-        dummy_inputs: Optional[Dict[str, Any]] = None,
+        concrete_args: Optional[dict[str, Any]] = None,
+        dummy_inputs: Optional[dict[str, Any]] = None,
         complete_concrete_args_with_inputs_not_in_dummy_inputs: bool = True,
     ) -> Graph:
         """
@@ -1420,7 +1419,7 @@ def keys(self, obj: "Proxy") -> Any:
         return attribute
 
 
-def get_concrete_args(model: nn.Module, input_names: List[str]):
+def get_concrete_args(model: nn.Module, input_names: list[str]):
     sig = inspect.signature(model.forward)
 
     if not (set(input_names) <= set(sig.parameters.keys())):
@@ -1448,9 +1447,9 @@ def check_if_model_is_supported(model: "PreTrainedModel"):
 
 def symbolic_trace(
     model: "PreTrainedModel",
-    input_names: Optional[List[str]] = None,
+    input_names: Optional[list[str]] = None,
     disable_check: bool = False,
-    tracer_cls: Type[HFTracer] = HFTracer,
+    tracer_cls: type[HFTracer] = HFTracer,
 ) -> GraphModule:
     """
     Performs symbolic tracing on the model.