Add ViTImageProcessorFast to tests (#31424)

* Add ViTImageProcessor to tests

* Correct data format

* Review comments

src/transformers/image_processing_utils.py

@@ -151,6 +151,11 @@ class BaseImageProcessor(ImageProcessingMixin):
             **kwargs,
         )
 
+    def to_dict(self):
+        encoder_dict = super().to_dict()
+        encoder_dict.pop("_valid_processor_keys", None)
+        return encoder_dict
+
 
 VALID_SIZE_DICT_KEYS = (
     {"height", "width"},

src/transformers/image_processing_utils_fast.py

@@ -61,3 +61,8 @@ class BaseImageProcessorFast(BaseImageProcessor):
     def get_transforms(self, **kwargs) -> "Compose":
         self._validate_params(**kwargs)
         return self._build_transforms(**kwargs)
+
+    def to_dict(self):
+        encoder_dict = super().to_dict()
+        encoder_dict.pop("_transform_params", None)
+        return encoder_dict

src/transformers/models/auto/image_processing_auto.py

@@ -399,7 +399,7 @@ class AutoImageProcessor:
             kwargs["token"] = use_auth_token
 
         config = kwargs.pop("config", None)
-        use_fast = kwargs.pop("use_fast", False)
+        use_fast = kwargs.pop("use_fast", None)
         trust_remote_code = kwargs.pop("trust_remote_code", None)
         kwargs["_from_auto"] = True
 
@@ -430,6 +430,7 @@ class AutoImageProcessor:
 
         if image_processor_class is not None:
             # Update class name to reflect the use_fast option. If class is not found, None is returned.
+            if use_fast is not None:
                 if use_fast and not image_processor_class.endswith("Fast"):
                     image_processor_class += "Fast"
                 elif not use_fast and image_processor_class.endswith("Fast"):

src/transformers/models/mask2former/image_processing_mask2former.py

@@ -772,7 +772,7 @@ class Mask2FormerImageProcessor(BaseImageProcessor):
             ignore_index,
             do_reduce_labels,
             return_tensors,
-            input_data_format=input_data_format,
+            input_data_format=data_format,
         )
         return encoded_inputs
 

src/transformers/models/maskformer/image_processing_maskformer.py

@@ -772,7 +772,7 @@ class MaskFormerImageProcessor(BaseImageProcessor):
             ignore_index,
             do_reduce_labels,
             return_tensors,
-            input_data_format=input_data_format,
+            input_data_format=data_format,
         )
         return encoded_inputs
 

src/transformers/models/oneformer/image_processing_oneformer.py

@@ -772,7 +772,7 @@ class OneFormerImageProcessor(BaseImageProcessor):
             ignore_index,
             do_reduce_labels,
             return_tensors,
-            input_data_format=input_data_format,
+            input_data_format=data_format,
         )
         return encoded_inputs
 

src/transformers/models/vit/image_processing_vit_fast.py

@@ -114,7 +114,6 @@ class ViTImageProcessorFast(BaseImageProcessorFast):
         self.rescale_factor = rescale_factor
         self.image_mean = image_mean if image_mean is not None else IMAGENET_STANDARD_MEAN
         self.image_std = image_std if image_std is not None else IMAGENET_STANDARD_STD
-        self._transform_settings = {}
 
     def _build_transforms(
         self,
@@ -285,5 +284,5 @@ class ViTImageProcessorFast(BaseImageProcessorFast):
         )
         transformed_images = [transforms(image) for image in images]
 
-        data = {"pixel_values": torch.vstack(transformed_images)}
+        data = {"pixel_values": torch.stack(transformed_images, dim=0)}
         return BatchFeature(data, tensor_type=return_tensors)

tests/models/bridgetower/test_image_processing_bridgetower.py

@@ -17,6 +17,8 @@
 import unittest
 from typing import Dict, List, Optional, Union
 
+import numpy as np
+
 from transformers.testing_utils import require_torch, require_vision
 from transformers.utils import is_vision_available
 
@@ -84,6 +86,8 @@ class BridgeTowerImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             scale = size / min(w, h)

tests/models/conditional_detr/test_image_processing_conditional_detr.py

@@ -18,6 +18,8 @@ import json
 import pathlib
 import unittest
 
+import numpy as np
+
 from transformers.testing_utils import require_torch, require_vision, slow
 from transformers.utils import is_torch_available, is_vision_available
 
@@ -87,6 +89,8 @@ class ConditionalDetrImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/deformable_detr/test_image_processing_deformable_detr.py

@@ -18,6 +18,8 @@ import json
 import pathlib
 import unittest
 
+import numpy as np
+
 from transformers.testing_utils import require_torch, require_vision, slow
 from transformers.utils import is_torch_available, is_vision_available
 
@@ -87,6 +89,8 @@ class DeformableDetrImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/detr/test_image_processing_detr.py

@@ -17,6 +17,8 @@ import json
 import pathlib
 import unittest
 
+import numpy as np
+
 from transformers.testing_utils import require_torch, require_vision, slow
 from transformers.utils import is_torch_available, is_vision_available
 
@@ -86,6 +88,8 @@ class DetrImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/glpn/test_image_processing_glpn.py

@@ -66,6 +66,8 @@ class GLPNImageProcessingTester(unittest.TestCase):
     def expected_output_image_shape(self, images):
         if isinstance(images[0], Image.Image):
             width, height = images[0].size
+        elif isinstance(images[0], np.ndarray):
+            height, width = images[0].shape[0], images[0].shape[1]
         else:
             height, width = images[0].shape[1], images[0].shape[2]
 

tests/models/grounding_dino/test_image_processing_grounding_dino.py

@@ -18,6 +18,8 @@ import json
 import pathlib
 import unittest
 
+import numpy as np
+
 from transformers.testing_utils import require_torch, require_vision, slow
 from transformers.utils import is_torch_available, is_vision_available
 
@@ -93,6 +95,8 @@ class GroundingDinoImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/idefics/test_image_processing_idefics.py

@@ -16,6 +16,8 @@
 
 import unittest
 
+import numpy as np
+
 from transformers.testing_utils import require_torch, require_torchvision, require_vision
 from transformers.utils import is_torch_available, is_torchvision_available, is_vision_available
 
@@ -75,6 +77,8 @@ class IdeficsImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             scale = size / min(w, h)

tests/models/idefics2/test_image_processing_idefics2.py

@@ -99,6 +99,8 @@ class Idefics2ImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
 
@@ -176,6 +178,10 @@ class Idefics2ImageProcessingTester(unittest.TestCase):
         if torchify:
             images_list = [[torch.from_numpy(image) for image in images] for images in images_list]
 
+        if numpify:
+            # Numpy images are typically in channels last format
+            images_list = [[image.transpose(1, 2, 0) for image in images] for images in images_list]
+
         return images_list
 
 
@@ -206,6 +212,7 @@ class Idefics2ImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
         self.assertTrue(hasattr(image_processing, "do_image_splitting"))
 
     def test_call_numpy(self):
+        for image_processing_class in self.image_processor_list:
             # Initialize image_processing
             image_processing = self.image_processing_class(**self.image_processor_dict)
             # create random numpy tensors
@@ -226,7 +233,39 @@ class Idefics2ImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
                 tuple(encoded_images.shape), (self.image_processor_tester.batch_size, *expected_output_image_shape)
             )
 
+    def test_call_numpy_4_channels(self):
+        for image_processing_class in self.image_processor_list:
+            # Initialize image_processing
+            image_processor_dict = self.image_processor_dict
+            image_processor_dict["image_mean"] = [0.5, 0.5, 0.5, 0.5]
+            image_processor_dict["image_std"] = [0.5, 0.5, 0.5, 0.5]
+            image_processing = self.image_processing_class(**image_processor_dict)
+            # create random numpy tensors
+            self.image_processor_tester.num_channels = 4
+            image_inputs = self.image_processor_tester.prepare_image_inputs(equal_resolution=False, numpify=True)
+
+            for sample_images in image_inputs:
+                for image in sample_images:
+                    self.assertIsInstance(image, np.ndarray)
+
+            # Test not batched input
+            encoded_images = image_processing(
+                image_inputs[0], input_data_format="channels_last", return_tensors="pt"
+            ).pixel_values
+            expected_output_image_shape = self.image_processor_tester.expected_output_image_shape([image_inputs[0]])
+            self.assertEqual(tuple(encoded_images.shape), (1, *expected_output_image_shape))
+
+            # Test batched
+            encoded_images = image_processing(
+                image_inputs, input_data_format="channels_last", return_tensors="pt"
+            ).pixel_values
+            expected_output_image_shape = self.image_processor_tester.expected_output_image_shape(image_inputs)
+            self.assertEqual(
+                tuple(encoded_images.shape), (self.image_processor_tester.batch_size, *expected_output_image_shape)
+            )
+
     def test_call_pil(self):
+        for image_processing_class in self.image_processor_list:
             # Initialize image_processing
             image_processing = self.image_processing_class(**self.image_processor_dict)
             # create random PIL images
@@ -248,6 +287,7 @@ class Idefics2ImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
             )
 
     def test_call_pytorch(self):
+        for image_processing_class in self.image_processor_list:
             # Initialize image_processing
             image_processing = self.image_processing_class(**self.image_processor_dict)
             # create random PyTorch tensors

tests/models/mask2former/test_image_processing_mask2former.py

@@ -98,6 +98,8 @@ class Mask2FormerImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/maskformer/test_image_processing_maskformer.py

@@ -98,6 +98,8 @@ class MaskFormerImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/oneformer/test_image_processing_oneformer.py

@@ -106,6 +106,8 @@ class OneFormerImageProcessorTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/oneformer/test_processor_oneformer.py

@@ -143,6 +143,8 @@ class OneFormerProcessorTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             if w < h:

tests/models/pix2struct/test_image_processing_pix2struct.py

@@ -232,7 +232,7 @@ class Pix2StructImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase)
         for max_patch in self.image_processor_tester.max_patches:
             # Test not batched input
             encoded_images = image_processor(
-                image_inputs[0], return_tensors="pt", max_patches=max_patch, input_data_format="channels_first"
+                image_inputs[0], return_tensors="pt", max_patches=max_patch, input_data_format="channels_last"
             ).flattened_patches
             self.assertEqual(
                 encoded_images.shape,
@@ -241,7 +241,7 @@ class Pix2StructImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase)
 
             # Test batched
             encoded_images = image_processor(
-                image_inputs, return_tensors="pt", max_patches=max_patch, input_data_format="channels_first"
+                image_inputs, return_tensors="pt", max_patches=max_patch, input_data_format="channels_last"
             ).flattened_patches
             self.assertEqual(
                 encoded_images.shape,

tests/models/swin2sr/test_image_processing_swin2sr.py

@@ -72,6 +72,8 @@ class Swin2SRImageProcessingTester(unittest.TestCase):
 
         if isinstance(img, Image.Image):
             input_width, input_height = img.size
+        elif isinstance(img, np.ndarray):
+            input_height, input_width = img.shape[-3:-1]
         else:
             input_height, input_width = img.shape[-2:]
 
@@ -160,7 +162,7 @@ class Swin2SRImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
 
         # Test not batched input
         encoded_images = image_processing(
-            image_inputs[0], return_tensors="pt", input_data_format="channels_first"
+            image_inputs[0], return_tensors="pt", input_data_format="channels_last"
         ).pixel_values
         expected_output_image_shape = self.image_processor_tester.expected_output_image_shape([image_inputs[0]])
         self.assertEqual(tuple(encoded_images.shape), (1, *expected_output_image_shape))

tests/models/video_llava/test_image_processing_video_llava.py

@@ -285,7 +285,7 @@ class VideoLlavaImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase)
         encoded_images = image_processor(
             image_inputs[0],
             return_tensors="pt",
-            input_data_format="channels_first",
+            input_data_format="channels_last",
             image_mean=0,
             image_std=1,
         ).pixel_values_images
@@ -296,7 +296,7 @@ class VideoLlavaImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase)
         encoded_images = image_processor(
             image_inputs,
             return_tensors="pt",
-            input_data_format="channels_first",
+            input_data_format="channels_last",
             image_mean=0,
             image_std=1,
         ).pixel_values_images

tests/models/vilt/test_image_processing_vilt.py

@@ -16,6 +16,8 @@
 
 import unittest
 
+import numpy as np
+
 from transformers.testing_utils import require_torch, require_vision
 from transformers.utils import is_vision_available
 
@@ -78,6 +80,8 @@ class ViltImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 w, h = image.size
+            elif isinstance(image, np.ndarray):
+                h, w = image.shape[0], image.shape[1]
             else:
                 h, w = image.shape[1], image.shape[2]
             scale = size / min(w, h)

tests/models/vit/test_image_processing_vit.py

@@ -17,7 +17,7 @@
 import unittest
 
 from transformers.testing_utils import require_torch, require_vision
-from transformers.utils import is_vision_available
+from transformers.utils import is_torchvision_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
 
@@ -25,6 +25,9 @@ from ...test_image_processing_common import ImageProcessingTestMixin, prepare_im
 if is_vision_available():
     from transformers import ViTImageProcessor
 
+if is_torchvision_available():
+    from transformers import ViTImageProcessorFast
+
 
 class ViTImageProcessingTester(unittest.TestCase):
     def __init__(
@@ -82,6 +85,7 @@ class ViTImageProcessingTester(unittest.TestCase):
 @require_vision
 class ViTImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
     image_processing_class = ViTImageProcessor if is_vision_available() else None
+    fast_image_processing_class = ViTImageProcessorFast if is_torchvision_available() else None
 
     def setUp(self):
         super().setUp()

tests/models/yolos/test_image_processing_yolos.py

@@ -18,6 +18,7 @@ import json
 import pathlib
 import unittest
 
+import numpy as np
 from parameterized import parameterized
 
 from transformers.testing_utils import require_torch, require_vision, slow
@@ -89,6 +90,8 @@ class YolosImageProcessingTester(unittest.TestCase):
             image = image_inputs[0]
             if isinstance(image, Image.Image):
                 width, height = image.size
+            elif isinstance(image, np.ndarray):
+                height, width = image.shape[0], image.shape[1]
             else:
                 height, width = image.shape[1], image.shape[2]
 

tests/test_image_processing_common.py

@@ -18,7 +18,9 @@ import json
 import os
 import pathlib
 import tempfile
+import time
 
+import numpy as np
 import requests
 
 from transformers import AutoImageProcessor, BatchFeature
@@ -28,7 +30,6 @@ from transformers.utils import is_torch_available, is_vision_available
 
 
 if is_torch_available():
-    import numpy as np
     import torch
 
 if is_vision_available():
@@ -72,6 +73,10 @@ def prepare_image_inputs(
     if torchify:
         image_inputs = [torch.from_numpy(image) for image in image_inputs]
 
+    if numpify:
+        # Numpy images are typically in channels last format
+        image_inputs = [image.transpose(1, 2, 0) for image in image_inputs]
+
     return image_inputs
 
 
@@ -167,33 +172,28 @@ class ImageProcessingTestMixin:
         encoding_slow = image_processor_slow(dummy_image, return_tensors="pt")
         encoding_fast = image_processor_fast(dummy_image, return_tensors="pt")
 
-        self.assertTrue(torch.allclose(encoding_slow.pixel_values, encoding_fast.pixel_values, atol=1e-3))
+        self.assertTrue(torch.allclose(encoding_slow.pixel_values, encoding_fast.pixel_values, atol=1e-2))
 
     @require_vision
     @require_torch
     def test_fast_is_faster_than_slow(self):
-        import time
-
-        def measure_time(self, image_processor, dummy_image):
-            start = time.time()
-            _ = image_processor(dummy_image, return_tensors="pt")
-            return time.time() - start
-
-        dummy_image = Image.open(
-            requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw
-        )
-
         if not self.test_slow_image_processor or not self.test_fast_image_processor:
             self.skipTest("Skipping speed test")
 
         if self.image_processing_class is None or self.fast_image_processing_class is None:
             self.skipTest("Skipping speed test as one of the image processors is not defined")
 
-        image_processor_slow = self.image_processing_class(**self.image_processor_dict)
-        image_processor_fast = self.fast_image_processing_class(**self.image_processor_dict)
+        def measure_time(image_processor, image):
+            start = time.time()
+            _ = image_processor(image, return_tensors="pt")
+            return time.time() - start
 
-        slow_time = self.measure_time(image_processor_slow, dummy_image)
-        fast_time = self.measure_time(image_processor_fast, dummy_image)
+        dummy_images = torch.randint(0, 255, (4, 3, 224, 224), dtype=torch.uint8)
+        image_processor_slow = self.image_processing_class(**self.image_processor_dict)
+        image_processor_fast = self.fast_image_processing_class()
+
+        fast_time = measure_time(image_processor_fast, dummy_images)
+        slow_time = measure_time(image_processor_slow, dummy_images)
 
         self.assertLessEqual(fast_time, slow_time)
 
@@ -238,6 +238,52 @@ class ImageProcessingTestMixin:
 
             self.assertEqual(image_processor_second.to_dict(), image_processor_first.to_dict())
 
+    def test_save_load_fast_slow(self):
+        "Test that we can load a fast image processor from a slow one and vice-versa."
+        if self.image_processing_class is None or self.fast_image_processing_class is None:
+            self.skipTest("Skipping slow/fast save/load test as one of the image processors is not defined")
+
+        image_processor_dict = self.image_processor_tester.prepare_image_processor_dict()
+        image_processor_slow_0 = self.image_processing_class(**image_processor_dict)
+
+        # Load fast image processor from slow one
+        with tempfile.TemporaryDirectory() as tmpdirname:
+            image_processor_slow_0.save_pretrained(tmpdirname)
+            image_processor_fast_0 = self.fast_image_processing_class.from_pretrained(tmpdirname)
+
+        image_processor_fast_1 = self.fast_image_processing_class(**image_processor_dict)
+
+        # Load slow image processor from fast one
+        with tempfile.TemporaryDirectory() as tmpdirname:
+            image_processor_fast_1.save_pretrained(tmpdirname)
+            image_processor_slow_1 = self.image_processing_class.from_pretrained(tmpdirname)
+
+        self.assertEqual(image_processor_slow_0.to_dict(), image_processor_slow_1.to_dict())
+        self.assertEqual(image_processor_fast_0.to_dict(), image_processor_fast_1.to_dict())
+
+    def test_save_load_fast_slow_auto(self):
+        "Test that we can load a fast image processor from a slow one and vice-versa using AutoImageProcessor."
+        if self.image_processing_class is None or self.fast_image_processing_class is None:
+            self.skipTest("Skipping slow/fast save/load test as one of the image processors is not defined")
+
+        image_processor_dict = self.image_processor_tester.prepare_image_processor_dict()
+        image_processor_slow_0 = self.image_processing_class(**image_processor_dict)
+
+        # Load fast image processor from slow one
+        with tempfile.TemporaryDirectory() as tmpdirname:
+            image_processor_slow_0.save_pretrained(tmpdirname)
+            image_processor_fast_0 = AutoImageProcessor.from_pretrained(tmpdirname, use_fast=True)
+
+        image_processor_fast_1 = self.fast_image_processing_class(**image_processor_dict)
+
+        # Load slow image processor from fast one
+        with tempfile.TemporaryDirectory() as tmpdirname:
+            image_processor_fast_1.save_pretrained(tmpdirname)
+            image_processor_slow_1 = AutoImageProcessor.from_pretrained(tmpdirname, use_fast=False)
+
+        self.assertEqual(image_processor_slow_0.to_dict(), image_processor_slow_1.to_dict())
+        self.assertEqual(image_processor_fast_0.to_dict(), image_processor_fast_1.to_dict())
+
     def test_init_without_params(self):
         for image_processing_class in self.image_processor_list:
             image_processor = image_processing_class()
@@ -358,7 +404,7 @@ class ImageProcessingTestMixin:
             encoded_images = image_processor(
                 image_inputs[0],
                 return_tensors="pt",
-                input_data_format="channels_first",
+                input_data_format="channels_last",
                 image_mean=0,
                 image_std=1,
             ).pixel_values
@@ -369,7 +415,7 @@ class ImageProcessingTestMixin:
             encoded_images = image_processor(
                 image_inputs,
                 return_tensors="pt",
-                input_data_format="channels_first",
+                input_data_format="channels_last",
                 image_mean=0,
                 image_std=1,
             ).pixel_values