Unify is_torchvision_v2_available with is_torchvision_available

src/transformers/image_processing_utils_fast.py

@@ -46,7 +46,6 @@
     auto_docstring,
     is_torch_available,
     is_torchvision_available,
-    is_torchvision_v2_available,
     is_vision_available,
     logging,
 )
@@ -60,14 +59,13 @@
     import torch
 
 if is_torchvision_available():
+    from torchvision.transforms.v2 import functional as F
+
     from .image_utils import pil_torch_interpolation_mapping
+
 else:
     pil_torch_interpolation_mapping = None
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-elif is_torchvision_available():
-    from torchvision.transforms import functional as F
 
 logger = logging.get_logger(__name__)
 

src/transformers/image_utils.py

@@ -28,7 +28,6 @@
     is_torch_available,
     is_torch_tensor,
     is_torchvision_available,
-    is_torchvision_v2_available,
     is_vision_available,
     logging,
     requires_backends,
@@ -54,9 +53,7 @@
         from torchvision.transforms import InterpolationMode
 
         pil_torch_interpolation_mapping = {
-            PILImageResampling.NEAREST: InterpolationMode.NEAREST_EXACT
-            if is_torchvision_v2_available()
-            else InterpolationMode.NEAREST,
+            PILImageResampling.NEAREST: InterpolationMode.NEAREST_EXACT,
             PILImageResampling.BOX: InterpolationMode.BOX,
             PILImageResampling.BILINEAR: InterpolationMode.BILINEAR,
             PILImageResampling.HAMMING: InterpolationMode.HAMMING,

src/transformers/models/beit/image_processing_beit_fast.py

@@ -17,6 +17,7 @@
 from typing import Optional, Union
 
 import torch
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils import BatchFeature
 from ...image_processing_utils_fast import (
@@ -38,16 +39,9 @@
 from ...utils import (
     TensorType,
     auto_docstring,
-    is_torchvision_v2_available,
 )
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
-
 class BeitFastImageProcessorKwargs(DefaultFastImageProcessorKwargs):
     r"""
     do_reduce_labels (`bool`, *optional*, defaults to `self.do_reduce_labels`):

src/transformers/models/bridgetower/image_processing_bridgetower_fast.py

@@ -18,6 +18,7 @@
 from typing import Optional, Union
 
 import torch
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils_fast import (
     BaseImageProcessorFast,
@@ -31,13 +32,7 @@
     reorder_images,
 )
 from ...image_utils import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD, PILImageResampling
-from ...utils import auto_docstring, is_torchvision_v2_available
-
-
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
+from ...utils import auto_docstring
 
 
 def make_pixel_mask(

src/transformers/models/chameleon/image_processing_chameleon_fast.py

@@ -19,17 +19,13 @@
 import numpy as np
 import PIL
 import torch
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils_fast import BaseImageProcessorFast
 from ...image_utils import ImageInput, PILImageResampling, SizeDict
-from ...utils import auto_docstring, is_torchvision_v2_available, logging
+from ...utils import auto_docstring, logging
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
 logger = logging.get_logger(__name__)
 
 

src/transformers/models/cohere2_vision/image_processing_cohere2_vision_fast.py

@@ -24,6 +24,7 @@
 
 import numpy as np
 import torch
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils import BatchFeature
 from ...image_processing_utils_fast import (
@@ -34,13 +35,7 @@
 )
 from ...image_utils import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD, ImageInput, PILImageResampling, SizeDict
 from ...processing_utils import Unpack
-from ...utils import TensorType, auto_docstring, is_torchvision_v2_available
-
-
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
+from ...utils import TensorType, auto_docstring
 
 
 class Cohere2VisionFastImageProcessorKwargs(DefaultFastImageProcessorKwargs):

src/transformers/models/conditional_detr/image_processing_conditional_detr_fast.py

@@ -10,6 +10,7 @@
 import torch
 from torch import nn
 from torchvision.io import read_image
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils import BatchFeature, get_size_dict
 from ...image_processing_utils_fast import (
@@ -33,7 +34,7 @@
     validate_annotations,
 )
 from ...processing_utils import Unpack
-from ...utils import TensorType, auto_docstring, is_torchvision_v2_available, logging
+from ...utils import TensorType, auto_docstring, logging
 from ...utils.import_utils import requires
 from .image_processing_conditional_detr import (
     compute_segments,
@@ -43,12 +44,6 @@
 )
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
-
 logger = logging.get_logger(__name__)
 
 
@@ -433,13 +428,7 @@ def resize_annotation(
             resample (`InterpolationMode`, defaults to `F.InterpolationMode.NEAREST_EXACT`):
                 The resampling filter to use when resizing the masks.
         """
-        interpolation = (
-            interpolation
-            if interpolation is not None
-            else F.InterpolationMode.NEAREST_EXACT
-            if is_torchvision_v2_available()
-            else F.InterpolationMode.NEAREST
-        )
+        interpolation = interpolation if interpolation is not None else F.InterpolationMode.NEAREST_EXACT
         ratio_height, ratio_width = [target / orig for target, orig in zip(target_size, orig_size)]
 
         new_annotation = {}

src/transformers/models/convnext/image_processing_convnext_fast.py

@@ -17,6 +17,7 @@
 from typing import Optional, Union
 
 import torch
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils import BatchFeature
 from ...image_processing_utils_fast import (
@@ -37,16 +38,9 @@
 from ...utils import (
     TensorType,
     auto_docstring,
-    is_torchvision_v2_available,
 )
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
-
 class ConvNextFastImageProcessorKwargs(DefaultFastImageProcessorKwargs):
     """
     crop_pct (`float`, *optional*):

src/transformers/models/deepseek_vl/image_processing_deepseek_vl.py

@@ -38,12 +38,7 @@
     valid_images,
     validate_preprocess_arguments,
 )
-from ...utils import (
-    TensorType,
-    filter_out_non_signature_kwargs,
-    is_vision_available,
-    logging,
-)
+from ...utils import TensorType, filter_out_non_signature_kwargs, is_vision_available, logging
 
 
 if is_vision_available():

src/transformers/models/deepseek_vl/modeling_deepseek_vl.py

@@ -29,11 +29,7 @@
 from ...modeling_outputs import ModelOutput
 from ...modeling_utils import PreTrainedModel
 from ...processing_utils import Unpack
-from ...utils import (
-    TransformersKwargs,
-    auto_docstring,
-    can_return_tuple,
-)
+from ...utils import TransformersKwargs, auto_docstring, can_return_tuple
 from ..auto import AutoModel
 from .configuration_deepseek_vl import DeepseekVLConfig
 

src/transformers/models/deformable_detr/image_processing_deformable_detr_fast.py

@@ -9,6 +9,7 @@
 
 import torch
 from torchvision.io import read_image
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils import BatchFeature, get_size_dict
 from ...image_processing_utils_fast import (
@@ -32,17 +33,11 @@
     validate_annotations,
 )
 from ...processing_utils import Unpack
-from ...utils import TensorType, auto_docstring, is_torchvision_v2_available, logging
+from ...utils import TensorType, auto_docstring, logging
 from ...utils.import_utils import requires
 from .image_processing_deformable_detr import get_size_with_aspect_ratio
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
-
 logger = logging.get_logger(__name__)
 
 
@@ -427,13 +422,7 @@ def resize_annotation(
             resample (`InterpolationMode`, defaults to `F.InterpolationMode.NEAREST_EXACT`):
                 The resampling filter to use when resizing the masks.
         """
-        interpolation = (
-            interpolation
-            if interpolation is not None
-            else F.InterpolationMode.NEAREST_EXACT
-            if is_torchvision_v2_available()
-            else F.InterpolationMode.NEAREST
-        )
+        interpolation = interpolation if interpolation is not None else F.InterpolationMode.NEAREST_EXACT
         ratio_height, ratio_width = [target / orig for target, orig in zip(target_size, orig_size)]
 
         new_annotation = {}

src/transformers/models/depth_pro/image_processing_depth_pro_fast.py

@@ -30,7 +30,6 @@
 from ...utils import (
     TensorType,
     auto_docstring,
-    is_torchvision_v2_available,
     logging,
     requires_backends,
 )
@@ -41,10 +40,7 @@
     from .modeling_depth_pro import DepthProDepthEstimatorOutput
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
+from torchvision.transforms.v2 import functional as F
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/detr/image_processing_detr_fast.py

@@ -23,6 +23,7 @@
 import torch
 from torch import nn
 from torchvision.io import read_image
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils import BatchFeature, get_size_dict
 from ...image_processing_utils_fast import (
@@ -49,7 +50,6 @@
 from ...utils import (
     TensorType,
     auto_docstring,
-    is_torchvision_v2_available,
     logging,
 )
 from ...utils.import_utils import requires
@@ -61,12 +61,6 @@
 )
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
-
 logger = logging.get_logger(__name__)
 
 SUPPORTED_ANNOTATION_FORMATS = (AnnotationFormat.COCO_DETECTION, AnnotationFormat.COCO_PANOPTIC)
@@ -450,13 +444,7 @@ def resize_annotation(
             resample (`InterpolationMode`, defaults to `F.InterpolationMode.NEAREST_EXACT`):
                 The resampling filter to use when resizing the masks.
         """
-        interpolation = (
-            interpolation
-            if interpolation is not None
-            else F.InterpolationMode.NEAREST_EXACT
-            if is_torchvision_v2_available()
-            else F.InterpolationMode.NEAREST
-        )
+        interpolation = interpolation if interpolation is not None else F.InterpolationMode.NEAREST_EXACT
         ratio_height, ratio_width = [target / orig for target, orig in zip(target_size, orig_size)]
 
         new_annotation = {}

src/transformers/models/dinov3_vit/image_processing_dinov3_vit_fast.py

@@ -17,24 +17,19 @@
 from typing import Optional, Union
 
 import torch
+from torchvision.transforms.v2 import functional as F
 
 from transformers.image_processing_base import BatchFeature
 from transformers.image_processing_utils_fast import BaseImageProcessorFast, group_images_by_shape, reorder_images
 from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD, PILImageResampling, SizeDict
 from transformers.utils import (
     TensorType,
     auto_docstring,
-    is_torchvision_v2_available,
     logging,
 )
 from transformers.utils.import_utils import requires
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
 logger = logging.get_logger(__name__)
 
 

src/transformers/models/donut/image_processing_donut_fast.py

@@ -17,6 +17,7 @@
 from typing import Optional, Union
 
 import torch
+from torchvision.transforms.v2 import functional as F
 
 from ...image_processing_utils_fast import BaseImageProcessorFast, BatchFeature, DefaultFastImageProcessorKwargs
 from ...image_transforms import group_images_by_shape, reorder_images
@@ -25,16 +26,10 @@
 from ...utils import (
     TensorType,
     auto_docstring,
-    is_torchvision_v2_available,
     logging,
 )
 
 
-if is_torchvision_v2_available():
-    from torchvision.transforms.v2 import functional as F
-else:
-    from torchvision.transforms import functional as F
-
 logger = logging.get_logger(__name__)