uniformize processor kwargs of altclip, bridgetower, flava, instructb…

…lipvideo, llava_next, llava_next_video, siglip, video_llava, vilt

src/transformers/models/altclip/processing_altclip.py

@@ -16,10 +16,23 @@
 Image/Text processor class for AltCLIP
 """
 
+import sys
 import warnings
+from typing import List, Optional, Union
 
-from ...processing_utils import ProcessorMixin
-from ...tokenization_utils_base import BatchEncoding
+from ...image_utils import ImageInput
+from ...processing_utils import ProcessingKwargs, ProcessorMixin
+from ...tokenization_utils_base import BatchEncoding, PreTokenizedInput, TextInput
+
+
+if sys.version_info >= (3, 11):
+    from typing import Unpack
+else:
+    from typing_extensions import Unpack
+
+
+class AltCLIPProcessingKwargs(ProcessingKwargs, total=False):
+    _defaults = {}
 
 
 class AltCLIPProcessor(ProcessorMixin):
@@ -59,7 +72,12 @@ def __init__(self, image_processor=None, tokenizer=None, **kwargs):
 
         super().__init__(image_processor, tokenizer)
 
-    def __call__(self, text=None, images=None, return_tensors=None, **kwargs):
+    def __call__(
+        self,
+        text: Optional[Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]]] = None,
+        images: Optional[ImageInput] = None,
+        **kwargs: Unpack[AltCLIPProcessingKwargs],
+    ):
         """
         Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
         and `kwargs` arguments to XLMRobertaTokenizerFast's [`~XLMRobertaTokenizerFast.__call__`] if `text` is not
@@ -97,19 +115,27 @@ def __call__(self, text=None, images=None, return_tensors=None, **kwargs):
         if text is None and images is None:
             raise ValueError("You have to specify either text or images. Both cannot be none.")
 
+        output_kwargs = self._merge_kwargs(
+            AltCLIPProcessingKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
+
         if text is not None:
-            encoding = self.tokenizer(text, return_tensors=return_tensors, **kwargs)
+            encoding = self.tokenizer(text, **output_kwargs["text_kwargs"])
 
         if images is not None:
-            image_features = self.image_processor(images, return_tensors=return_tensors, **kwargs)
+            image_features = self.image_processor(images, **output_kwargs["images_kwargs"])
 
         if text is not None and images is not None:
             encoding["pixel_values"] = image_features.pixel_values
             return encoding
         elif text is not None:
             return encoding
         else:
-            return BatchEncoding(data=dict(**image_features), tensor_type=return_tensors)
+            return BatchEncoding(
+                data=dict(**image_features), tensor_type=output_kwargs["common_kwargs"]["return_tensors"]
+            )
 
     def batch_decode(self, *args, **kwargs):
         """

src/transformers/models/bridgetower/image_processing_bridgetower.py

@@ -115,8 +115,8 @@ def get_resize_output_image_size(
         new_width = scale * new_width
 
     new_height, new_width = int(new_height + 0.5), int(new_width + 0.5)
-    new_height = new_height // size_divisor * size_divisor
-    new_width = new_width // size_divisor * size_divisor
+    new_height = max(1, new_height // size_divisor) * size_divisor
+    new_width = max(1, new_width // size_divisor) * size_divisor
 
     return new_height, new_width
 

src/transformers/models/flava/processing_flava.py

@@ -16,13 +16,41 @@
 Image/Text processor class for FLAVA
 """
 
+import sys
 import warnings
 from typing import List, Optional, Union
 
 from ...image_utils import ImageInput
-from ...processing_utils import ProcessorMixin
-from ...tokenization_utils_base import BatchEncoding, PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
-from ...utils import TensorType
+from ...processing_utils import ImagesKwargs, ProcessingKwargs, ProcessorMixin
+from ...tokenization_utils_base import BatchEncoding, PreTokenizedInput, TextInput
+
+
+if sys.version_info >= (3, 11):
+    from typing import Unpack
+else:
+    from typing_extensions import Unpack
+
+
+class FlavaImagesKwargs(ImagesKwargs, total=False):
+    return_image_mask: Optional[bool]
+    return_codebook_pixels: Optional[bool]
+
+
+class FlavaProcessorKwargs(ProcessingKwargs, total=False):
+    images_kwargs: FlavaImagesKwargs
+    _defaults = {
+        "text_kwargs": {
+            "add_special_tokens": True,
+            "padding": False,
+            "truncation": False,
+            "stride": 0,
+            "return_overflowing_tokens": False,
+            "return_special_tokens_mask": False,
+            "return_offsets_mapping": False,
+            "return_length": False,
+            "verbose": True,
+        },
+    }
 
 
 class FlavaProcessor(ProcessorMixin):
@@ -64,23 +92,7 @@ def __call__(
         self,
         images: Optional[ImageInput] = None,
         text: Optional[Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]]] = None,
-        add_special_tokens: bool = True,
-        padding: Union[bool, str, PaddingStrategy] = False,
-        truncation: Union[bool, str, TruncationStrategy] = False,
-        max_length: Optional[int] = None,
-        stride: int = 0,
-        pad_to_multiple_of: Optional[int] = None,
-        return_image_mask: Optional[bool] = None,
-        return_codebook_pixels: Optional[bool] = None,
-        return_token_type_ids: Optional[bool] = None,
-        return_attention_mask: Optional[bool] = None,
-        return_overflowing_tokens: bool = False,
-        return_special_tokens_mask: bool = False,
-        return_offsets_mapping: bool = False,
-        return_length: bool = False,
-        verbose: bool = True,
-        return_tensors: Optional[Union[str, TensorType]] = None,
-        **kwargs,
+        **kwargs: Unpack[FlavaProcessorKwargs],
     ):
         """
         This method uses [`FlavaImageProcessor.__call__`] method to prepare image(s) for the model, and
@@ -92,41 +104,26 @@ def __call__(
         if text is None and images is None:
             raise ValueError("You have to specify either text or images. Both cannot be none.")
 
+        output_kwargs = self._merge_kwargs(
+            FlavaProcessorKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
+
         if text is not None:
-            encoding = self.tokenizer(
-                text=text,
-                add_special_tokens=add_special_tokens,
-                padding=padding,
-                truncation=truncation,
-                max_length=max_length,
-                stride=stride,
-                pad_to_multiple_of=pad_to_multiple_of,
-                return_token_type_ids=return_token_type_ids,
-                return_attention_mask=return_attention_mask,
-                return_overflowing_tokens=return_overflowing_tokens,
-                return_special_tokens_mask=return_special_tokens_mask,
-                return_offsets_mapping=return_offsets_mapping,
-                return_length=return_length,
-                verbose=verbose,
-                return_tensors=return_tensors,
-                **kwargs,
-            )
+            encoding = self.tokenizer(text=text, **output_kwargs["text_kwargs"])
         if images is not None:
-            image_features = self.image_processor(
-                images,
-                return_image_mask=return_image_mask,
-                return_codebook_pixels=return_codebook_pixels,
-                return_tensors=return_tensors,
-                **kwargs,
-            )
+            image_features = self.image_processor(images, **output_kwargs["images_kwargs"])
 
         if text is not None and images is not None:
             encoding.update(image_features)
             return encoding
         elif text is not None:
             return encoding
         else:
-            return BatchEncoding(data=dict(**image_features), tensor_type=return_tensors)
+            return BatchEncoding(
+                data=dict(**image_features), tensor_type=output_kwargs["common_kwargs"]["return_tensors"]
+            )
 
     def batch_decode(self, *args, **kwargs):
         """

src/transformers/models/instructblipvideo/processing_instructblipvideo.py

@@ -17,26 +17,48 @@
 """
 
 import os
+import sys
 from typing import List, Optional, Union
 
 from ...image_processing_utils import BatchFeature
 from ...image_utils import VideoInput
-from ...processing_utils import ProcessorMixin
+from ...processing_utils import ProcessingKwargs, ProcessorMixin
 from ...tokenization_utils_base import (
     AddedToken,
     BatchEncoding,
-    PaddingStrategy,
     PreTokenizedInput,
     TextInput,
-    TruncationStrategy,
 )
-from ...utils import TensorType, logging
+from ...utils import logging
 from ..auto import AutoTokenizer
 
 
+if sys.version_info >= (3, 11):
+    from typing import Unpack
+else:
+    from typing_extensions import Unpack
+
+
 logger = logging.get_logger(__name__)
 
 
+class InstructBlipVideoProcessorKwargs(ProcessingKwargs, total=False):
+    _defaults = {
+        "text_kwargs": {
+            "add_special_tokens": True,
+            "padding": False,
+            "truncation": None,
+            "stride": 0,
+            "return_overflowing_tokens": False,
+            "return_special_tokens_mask": False,
+            "return_offsets_mapping": False,
+            "return_token_type_ids": False,
+            "return_length": False,
+            "verbose": True,
+        },
+    }
+
+
 class InstructBlipVideoProcessor(ProcessorMixin):
     r"""
     Constructs an InstructBLIPVideo processor which wraps a InstructBLIP image processor and a LLaMa/T5 tokenizer into a single
@@ -71,30 +93,24 @@ def __init__(self, image_processor, tokenizer, qformer_tokenizer=None, num_query
 
     def __call__(
         self,
-        images: VideoInput = None,
-        text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]] = None,
-        add_special_tokens: bool = True,
-        padding: Union[bool, str, PaddingStrategy] = False,
-        truncation: Union[bool, str, TruncationStrategy] = None,
-        max_length: Optional[int] = None,
-        stride: int = 0,
-        pad_to_multiple_of: Optional[int] = None,
-        return_attention_mask: Optional[bool] = None,
-        return_overflowing_tokens: bool = False,
-        return_special_tokens_mask: bool = False,
-        return_offsets_mapping: bool = False,
-        return_token_type_ids: bool = False,
-        return_length: bool = False,
-        verbose: bool = True,
-        return_tensors: Optional[Union[str, TensorType]] = None,
-        **kwargs,
+        images: Optional[VideoInput] = None,
+        text: Optional[Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]]] = None,
+        audio=None,
+        videos=None,
+        **kwargs: Unpack[InstructBlipVideoProcessorKwargs],
     ) -> BatchFeature:
         """
         This method uses [`InstructBlipVideoImageProcessor.__call__`] method to prepare image(s) or video(s) for the model, and
         [`BertTokenizerFast.__call__`] to prepare text for the model.
 
         Please refer to the docstring of the above two methods for more information.
         """
+        output_kwargs = self._merge_kwargs(
+            InstructBlipVideoProcessorKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
+
         encoding = BatchFeature()
 
         if text is not None:
@@ -105,21 +121,10 @@ def __call__(
 
             _text_encoding = self.tokenizer(
                 text=text,
-                add_special_tokens=add_special_tokens,
-                padding=padding,
-                truncation=truncation,
-                max_length=max_length,
-                stride=stride,
-                pad_to_multiple_of=pad_to_multiple_of,
-                return_attention_mask=return_attention_mask,
-                return_overflowing_tokens=return_overflowing_tokens,
-                return_special_tokens_mask=return_special_tokens_mask,
-                return_offsets_mapping=return_offsets_mapping,
-                return_token_type_ids=return_token_type_ids,
-                return_length=return_length,
-                verbose=verbose,
-                return_tensors=None,  # required to concatenate below
-                **kwargs,
+                **{
+                    **output_kwargs["text_kwargs"],
+                    "return_tensors": None,  # required to concatenate below
+                },
             )
 
             # if we know how many query tokens, expand text inside processor. We need this hacky manipulation
@@ -145,31 +150,16 @@ def __call__(
                     )
 
             # cast to desired return tensors type after concatenating
-            text_encoding = BatchEncoding(text_encoding, tensor_type=return_tensors)
-            encoding.update(text_encoding)
-            qformer_text_encoding = self.qformer_tokenizer(
-                text=text,
-                add_special_tokens=add_special_tokens,
-                padding=padding,
-                truncation=truncation,
-                max_length=max_length,
-                stride=stride,
-                pad_to_multiple_of=pad_to_multiple_of,
-                return_attention_mask=return_attention_mask,
-                return_overflowing_tokens=return_overflowing_tokens,
-                return_special_tokens_mask=return_special_tokens_mask,
-                return_offsets_mapping=return_offsets_mapping,
-                return_token_type_ids=return_token_type_ids,
-                return_length=return_length,
-                verbose=verbose,
-                return_tensors=return_tensors,
-                **kwargs,
+            text_encoding = BatchEncoding(
+                text_encoding, tensor_type=output_kwargs["common_kwargs"].get("return_tensors")
             )
+            encoding.update(text_encoding)
+            qformer_text_encoding = self.qformer_tokenizer(text=text, **output_kwargs["text_kwargs"])
             encoding["qformer_input_ids"] = qformer_text_encoding.pop("input_ids")
             encoding["qformer_attention_mask"] = qformer_text_encoding.pop("attention_mask")
 
         if images is not None:
-            image_encoding = self.image_processor(images, return_tensors=return_tensors)
+            image_encoding = self.image_processor(images, **output_kwargs["images_kwargs"])
             encoding.update(image_encoding)
 
         return encoding

src/transformers/models/llava_next/image_processing_llava_next.py

@@ -720,7 +720,11 @@ def preprocess(
             image_patches = self.get_image_patches(
                 image,
                 image_grid_pinpoints,
-                size=(size["shortest_edge"], size["shortest_edge"]),
+                size=(
+                    (size["shortest_edge"], size["shortest_edge"])
+                    if "shortest_edge" in size
+                    else (size["height"], size["width"])
+                ),
                 patch_size=crop_size["height"],
                 resample=resample,
                 data_format=input_data_format,