Enable more ruff UP rules (#40579)

* Import Sequence from collections.abc

Signed-off-by: cyy <cyyever@outlook.com>

* Apply ruff UP rules

Signed-off-by: cyy <cyyever@outlook.com>

---------

Signed-off-by: cyy <cyyever@outlook.com>

examples/pytorch/text-generation/run_generation.py

@@ -63,7 +63,7 @@ logging.basicConfig(
 )
 logger = logging.getLogger(__name__)
 
-MAX_LENGTH = int(10000)  # Hardcoded max length to avoid infinite loop
+MAX_LENGTH = 10000  # Hardcoded max length to avoid infinite loop
 
 MODEL_CLASSES = {
     "gpt2": (GPT2LMHeadModel, GPT2Tokenizer),

pyproject.toml

@@ -22,10 +22,15 @@ line-length = 119
 # SIM300: Yoda condition detected
 # SIM212: Checks for if expressions that check against a negated condition.
 # SIM905: Consider using a list literal instead of `str.split`
-ignore = ["C901", "E501", "E741", "F402", "F823", "SIM1", "SIM300", "SIM212", "SIM905"]
+# UP009: UTF-8 encoding declaration is unnecessary
+# UP015: Unnecessary mode argument
+# UP031: Use format specifiers instead of percent format
+# UP004: Class `XXX` inherits from `object`
+# UP028: Checks for for loops that can be replaced with yield from expressions
+ignore = ["C901", "E501", "E741", "F402", "F823", "SIM1", "SIM300", "SIM212", "SIM905", "UP009", "UP015", "UP031", "UP028", "UP004"]
 # RUF013: Checks for the use of implicit Optional
 #  in type annotations when the default parameter value is None.
-select = ["C", "E", "F", "I", "W", "RUF013", "UP006", "PERF102", "PLC1802", "PLC0208","SIM"]
+select = ["C", "E", "F", "I", "W", "RUF013", "PERF102", "PLC1802", "PLC0208", "SIM", "UP"]
 extend-safe-fixes = ["UP006"]
 
 # Ignore import violations in all `__init__.py` files.
@@ -33,6 +38,7 @@ extend-safe-fixes = ["UP006"]
 "__init__.py" = ["E402", "F401", "F403", "F811"]
 "src/transformers/file_utils.py" = ["F401"]
 "src/transformers/utils/dummy_*.py" = ["F401"]
+"examples/legacy/**/*.py" = ["UP"]
 
 [tool.ruff.lint.isort]
 lines-after-imports = 2

src/transformers/audio_utils.py

@@ -21,8 +21,9 @@ import importlib
 import io
 import os
 import warnings
+from collections.abc import Sequence
 from io import BytesIO
-from typing import Any, Optional, Sequence, Union
+from typing import Any, Optional, Union
 
 import numpy as np
 import requests

src/transformers/commands/serving.py

@@ -734,7 +734,7 @@ class ServeCommand(BaseTransformersCLICommand):
 
         uvicorn.run(app, host=self.args.host, port=self.args.port, log_level=self.args.log_level)
 
-    @functools.lru_cache(maxsize=None)
+    @functools.cache
     def get_gen_models(self) -> list[dict[str, any]]:
         """
         This is by no means a limit to which models may be instantiated with `transformers serve`: any chat-based

src/transformers/modeling_layers.py

@@ -95,7 +95,7 @@ class GradientCheckpointingLayer(nn.Module):
 
 
 @auto_docstring
-class GenericForSequenceClassification(object):
+class GenericForSequenceClassification:
     base_model_prefix = "model"
 
     def __init__(self, config):
@@ -170,7 +170,7 @@ class GenericForSequenceClassification(object):
 
 
 @auto_docstring
-class GenericForQuestionAnswering(object):
+class GenericForQuestionAnswering:
     base_model_prefix = "model"
 
     def __init__(self, config):
@@ -231,7 +231,7 @@ class GenericForQuestionAnswering(object):
 
 
 @auto_docstring
-class GenericForTokenClassification(object):
+class GenericForTokenClassification:
     base_model_prefix = "model"
 
     def __init__(self, config):

src/transformers/models/florence2/modular_florence2.py

@@ -994,7 +994,7 @@ class Florence2PostProcessor:
                     instances = self.parse_description_with_bboxes_from_text_and_spans(text, image_size=image_size)
                 parsed_dict["description_with_bboxes_or_polygons"] = instances
             else:
-                raise ValueError("task {} is not supported".format(task))
+                raise ValueError(f"task {task} is not supported")
 
         return parsed_dict
 

src/transformers/models/florence2/processing_florence2.py

@@ -795,7 +795,7 @@ class Florence2PostProcessor:
                     instances = self.parse_description_with_bboxes_from_text_and_spans(text, image_size=image_size)
                 parsed_dict["description_with_bboxes_or_polygons"] = instances
             else:
-                raise ValueError("task {} is not supported".format(task))
+                raise ValueError(f"task {task} is not supported")
 
         return parsed_dict
 

src/transformers/models/gemma3n/convert_gemma3n_weights.py

@@ -519,7 +519,7 @@ def convert_vision_weights(
     weights: np.ndarray,
 ) -> Iterable[tuple[str, np.ndarray]]:
     def generate_base_path(path: str, block_type: str) -> tuple[str, tuple[int, int]]:
-        re_str = r"{}(\d+)/".format(block_type)
+        re_str = rf"{block_type}(\d+)/"
         re_pattern = re.compile(re_str)
         match = re.search(re_pattern, path).group(1)
         idx = abs(int(match)) - 1

src/transformers/models/sam2/modeling_sam2.py

@@ -1487,9 +1487,7 @@ class Sam2Model(Sam2PreTrainedModel):
         if input_points is not None and input_boxes is not None:
             if input_points.shape[1] != input_boxes.shape[1]:
                 raise ValueError(
-                    "You should provide as many bounding boxes as input points per box. Got {} and {}.".format(
-                        input_points.shape[1], input_boxes.shape[1]
-                    )
+                    f"You should provide as many bounding boxes as input points per box. Got {input_points.shape[1]} and {input_boxes.shape[1]}."
                 )
 
         image_positional_embeddings = self.get_image_wide_positional_embeddings()

src/transformers/models/sam2/modular_sam2.py

@@ -1384,9 +1384,7 @@ class Sam2Model(SamModel):
         if input_points is not None and input_boxes is not None:
             if input_points.shape[1] != input_boxes.shape[1]:
                 raise ValueError(
-                    "You should provide as many bounding boxes as input points per box. Got {} and {}.".format(
-                        input_points.shape[1], input_boxes.shape[1]
-                    )
+                    f"You should provide as many bounding boxes as input points per box. Got {input_points.shape[1]} and {input_boxes.shape[1]}."
                 )
 
         image_positional_embeddings = self.get_image_wide_positional_embeddings()

src/transformers/models/sam2_video/modeling_sam2_video.py

@@ -21,8 +21,9 @@
 
 import math
 from collections import OrderedDict
+from collections.abc import Iterator
 from dataclasses import dataclass
-from typing import Any, Callable, Iterator, Optional, Union
+from typing import Any, Callable, Optional, Union
 
 import numpy as np
 import torch
@@ -988,7 +989,7 @@ class Sam2VideoMemoryFuserCXBlock(GradientCheckpointingLayer):
         )  # pointwise/1x1 convs, implemented with linear layers
         self.pointwise_conv2 = nn.Linear(config.memory_fuser_intermediate_dim, config.memory_fuser_embed_dim)
         self.scale = nn.Parameter(
-            config.memory_fuser_layer_scale_init_value * torch.ones((config.memory_fuser_embed_dim)),
+            config.memory_fuser_layer_scale_init_value * torch.ones(config.memory_fuser_embed_dim),
             requires_grad=True,
         )
 
@@ -1923,9 +1924,7 @@ class Sam2VideoModel(Sam2VideoPreTrainedModel):
         if input_points is not None and input_boxes is not None:
             if input_points.shape[1] != input_boxes.shape[1]:
                 raise ValueError(
-                    "You should provide as many bounding boxes as input points per box. Got {} and {}.".format(
-                        input_points.shape[1], input_boxes.shape[1]
-                    )
+                    f"You should provide as many bounding boxes as input points per box. Got {input_points.shape[1]} and {input_boxes.shape[1]}."
                 )
         elif input_points is not None:
             num_objects = input_points.shape[1]

src/transformers/models/sam2_video/modular_sam2_video.py

@@ -16,8 +16,9 @@
 
 import math
 from collections import OrderedDict
+from collections.abc import Iterator
 from dataclasses import dataclass
-from typing import Any, Callable, Iterator, Optional, Union
+from typing import Any, Callable, Optional, Union
 
 import numpy as np
 import torch
@@ -1326,7 +1327,7 @@ class Sam2VideoMemoryFuserCXBlock(GradientCheckpointingLayer):
         )  # pointwise/1x1 convs, implemented with linear layers
         self.pointwise_conv2 = nn.Linear(config.memory_fuser_intermediate_dim, config.memory_fuser_embed_dim)
         self.scale = nn.Parameter(
-            config.memory_fuser_layer_scale_init_value * torch.ones((config.memory_fuser_embed_dim)),
+            config.memory_fuser_layer_scale_init_value * torch.ones(config.memory_fuser_embed_dim),
             requires_grad=True,
         )
 
@@ -1634,9 +1635,7 @@ class Sam2VideoModel(Sam2Model):
         if input_points is not None and input_boxes is not None:
             if input_points.shape[1] != input_boxes.shape[1]:
                 raise ValueError(
-                    "You should provide as many bounding boxes as input points per box. Got {} and {}.".format(
-                        input_points.shape[1], input_boxes.shape[1]
-                    )
+                    f"You should provide as many bounding boxes as input points per box. Got {input_points.shape[1]} and {input_boxes.shape[1]}."
                 )
         elif input_points is not None:
             num_objects = input_points.shape[1]

src/transformers/pipelines/keypoint_matching.py

@@ -12,7 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Sequence, TypedDict, Union
+from collections.abc import Sequence
+from typing import Any, TypedDict, Union
 
 from typing_extensions import TypeAlias, overload
 
@@ -29,8 +30,16 @@ if is_vision_available():
 
 ImagePair: TypeAlias = Sequence[Union["Image.Image", str]]
 
-Keypoint = TypedDict("Keypoint", {"x": float, "y": float})
-Match = TypedDict("Match", {"keypoint_image_0": Keypoint, "keypoint_image_1": Keypoint, "score": float})
+
+class Keypoint(TypedDict):
+    x: float
+    y: float
+
+
+class Match(TypedDict):
+    keypoint_image_0: Keypoint
+    keypoint_image_1: Keypoint
+    score: float
 
 
 def validate_image_pairs(images: Any) -> Sequence[Sequence[ImagePair]]:

src/transformers/video_processing_utils.py

@@ -694,8 +694,8 @@ class BaseVideoProcessor(BaseImageProcessorFast):
                     _raise_exceptions_for_missing_entries=False,
                 )
                 resolved_video_processor_file = resolved_video_processor_files[0]
-            except EnvironmentError:
-                # Raise any environment error raise by `cached_file`. It will have a helpful error message adapted to
+            except OSError:
+                # Raise any OS error raise by `cached_file`. It will have a helpful error message adapted to
                 # the original exception.
                 raise
             except Exception: