Uniform kwargs for processors of audio-text models

src/transformers/models/altclip/processing_altclip.py

@@ -16,10 +16,24 @@
 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 ...feature_extraction_utils import BatchFeature
+from ...image_utils import ImageInput
+from ...processing_utils import ProcessingKwargs, ProcessorMixin
+from ...tokenization_utils_base import PreTokenizedInput, TextInput
+
+
+if sys.version_info >= (3, 11):
+    from typing import Unpack
+else:
+    from typing_extensions import Unpack
+
+
+class AltCLIPProcessorKwargs(ProcessingKwargs, total=False):
+    _defaults = {}
 
 
 class AltCLIPProcessor(ProcessorMixin):
@@ -59,7 +73,14 @@ 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,
+        audio=None,
+        videos=None,
+        **kwargs: Unpack[AltCLIPProcessorKwargs],
+    ) -> BatchFeature:
         """
         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
@@ -68,24 +89,16 @@ def __call__(self, text=None, images=None, return_tensors=None, **kwargs):
         of the above two methods for more information.
 
         Args:
-            text (`str`, `List[str]`, `List[List[str]]`):
+            text (`TextInput`, `PreTokenizedInput`, `List[TextInput]`, `List[PreTokenizedInput]`):
                 The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
                 (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
                 `is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
-            images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
+            images (`ImageInput`, *optional*):
                 The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
                 tensor. Both channels-first and channels-last formats are supported.
 
-            return_tensors (`str` or [`~utils.TensorType`], *optional*):
-                If set, will return tensors of a particular framework. Acceptable values are:
-
-                - `'tf'`: Return TensorFlow `tf.constant` objects.
-                - `'pt'`: Return PyTorch `torch.Tensor` objects.
-                - `'np'`: Return NumPy `np.ndarray` objects.
-                - `'jax'`: Return JAX `jnp.ndarray` objects.
-
         Returns:
-            [`BatchEncoding`]: A [`BatchEncoding`] with the following fields:
+            [`BatchFeature`]: A [`BatchFeature`] with the following fields:
 
             - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`.
             - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
@@ -97,19 +110,20 @@ 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.")
 
-        if text is not None:
-            encoding = self.tokenizer(text, return_tensors=return_tensors, **kwargs)
+        output_kwargs = self._merge_kwargs(
+            AltCLIPProcessorKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
 
+        data = {}
+        if text is not None:
+            text_features = self.tokenizer(text, **output_kwargs["text_kwargs"])
+            data.update(text_features)
         if images is not None:
-            image_features = self.image_processor(images, return_tensors=return_tensors, **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)
+            image_features = self.image_processor(images, **output_kwargs["images_kwargs"])
+            data.update(image_features)
+        return BatchFeature(data=data, tensor_type=output_kwargs["common_kwargs"].get("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
 
@@ -238,9 +238,7 @@ def resize(
                 The channel dimension format of the input image. If not provided, it will be inferred.
         """
         size = get_size_dict(size, default_to_square=False)
-        if "shortest_edge" not in size:
-            raise ValueError(f"The `size` dictionary must contain the key `shortest_edge`. Got {size.keys()}")
-        shorter = size["shortest_edge"]
+        shorter = size["shortest_edge"] if "shortest_edge" in size else min(size["height"], size["width"])
         longer = int(1333 / 800 * shorter)
         output_size = get_resize_output_image_size(
             image, shorter=shorter, longer=longer, size_divisor=size_divisor, input_data_format=input_data_format

src/transformers/models/chameleon/image_processing_chameleon.py

@@ -44,7 +44,8 @@
     import PIL
 
 
-def make_batched_images(images) -> List[List[ImageInput]]:
+# Copied from transformers.models.llava_next.image_processing_llava_next.make_batched_images
+def make_batched_images(images) -> List[ImageInput]:
     """
     Accepts images in list or nested list format, and makes a list of images for preprocessing.
 

src/transformers/models/chameleon/processing_chameleon.py

@@ -16,13 +16,36 @@
 Processor class for Chameleon.
 """
 
+import sys
 from typing import List, Optional, Union
 
 from ...feature_extraction_utils import BatchFeature
 from ...image_utils import ImageInput
-from ...processing_utils import ProcessorMixin
-from ...tokenization_utils_base import PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
-from ...utils import TensorType
+from ...processing_utils import ProcessingKwargs, ProcessorMixin, TextKwargs
+from ...tokenization_utils_base import PreTokenizedInput, TextInput
+
+
+if sys.version_info >= (3, 11):
+    from typing import Unpack
+else:
+    from typing_extensions import Unpack
+
+
+class ChameleonTextKwargs(TextKwargs, total=False):
+    return_for_text_completion: bool
+
+
+class ChameleonProcessorKwargs(ProcessingKwargs, total=False):
+    text_kwargs: ChameleonTextKwargs
+    _defaults = {
+        "text_kwargs": {
+            "padding": False,
+            "return_for_text_completion": False,
+        },
+        "common_kwargs": {
+            "return_tensors": "pt",
+        },
+    }
 
 
 class ChameleonProcessor(ProcessorMixin):
@@ -57,13 +80,9 @@ def __init__(self, image_processor, tokenizer, image_seq_length: int = 1024, ima
 
     def __call__(
         self,
-        text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]] = None,
-        images: ImageInput = None,
-        padding: Union[bool, str, PaddingStrategy] = False,
-        truncation: Union[bool, str, TruncationStrategy] = None,
-        max_length: int = None,
-        return_tensors: Optional[Union[str, TensorType]] = TensorType.PYTORCH,
-        return_for_text_completion: bool = False,
+        text: Optional[Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]]] = None,
+        images: Optional[ImageInput] = None,
+        **kwargs: Unpack[ChameleonProcessorKwargs],
     ) -> BatchFeature:
         """
         Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
@@ -73,33 +92,13 @@ def __call__(
         of the above two methods for more information.
 
         Args:
-            text (`str`, `List[str]`, `List[List[str]]`):
+            text (`TextInput`, `PreTokenizedInput`, `List[TextInput]`, `List[PreTokenizedInput]`, *optional*):
                 The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
                 (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
                 `is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
-            images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
+            images (`ImageInput`, *optional*):
                 The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
                 tensor. Both channels-first and channels-last formats are supported.
-            padding (`bool`, `str` or [`~utils.PaddingStrategy`], *optional*, defaults to `False`):
-                Select a strategy to pad the returned sequences (according to the model's padding side and padding
-                index) among:
-                - `True` or `'longest'`: Pad to the longest sequence in the batch (or no padding if only a single
-                  sequence if provided).
-                - `'max_length'`: Pad to a maximum length specified with the argument `max_length` or to the maximum
-                  acceptable input length for the model if that argument is not provided.
-                - `False` or `'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of different
-                  lengths).
-            max_length (`int`, *optional*):
-                Maximum length of the returned list and optionally padding length (see above).
-            truncation (`bool`, *optional*):
-                Activates truncation to cut input sequences longer than `max_length` to `max_length`.
-            return_tensors (`str` or [`~utils.TensorType`], *optional*):
-                If set, will return tensors of a particular framework. Acceptable values are:
-
-                - `'tf'`: Return TensorFlow `tf.constant` objects.
-                - `'pt'`: Return PyTorch `torch.Tensor` objects.
-                - `'np'`: Return NumPy `np.ndarray` objects.
-                - `'jax'`: Return JAX `jnp.ndarray` objects.
 
         Returns:
             [`BatchFeature`]: A [`BatchFeature`] with the following fields:
@@ -114,6 +113,15 @@ def __call__(
             text = [text]
         elif not isinstance(text, list) and not isinstance(text[0], str):
             raise TypeError("Invalid input text. Please provide a string, or a list of strings")
+        if text is None and images is None:
+            raise ValueError("You must provide either text or images as prompt")
+
+        output_kwargs = self._merge_kwargs(
+            ChameleonProcessorKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
+        return_for_text_completion = output_kwargs["text_kwargs"].pop("return_for_text_completion", False)
 
         # Replace the image token with the expanded image token sequence
         prompt_strings = []
@@ -124,19 +132,10 @@ def __call__(
                 sample += self.tokenizer.sep_token  # special Chameleon treatment to add sep for chat mode
             prompt_strings.append(sample)
 
-        data = self.tokenizer(
-            prompt_strings,
-            return_tensors=return_tensors,
-            padding=padding,
-            truncation=truncation,
-            max_length=max_length,
-        )
-
+        features = self.tokenizer(prompt_strings, **output_kwargs["text_kwargs"])
         if images is not None:
-            pixel_values = self.image_processor(images, return_tensors=return_tensors)["pixel_values"]
-            data["pixel_values"] = pixel_values
-
-        return BatchFeature(data=data, tensor_type=return_tensors)
+            features["pixel_values"] = self.image_processor(images, **output_kwargs["images_kwargs"])["pixel_values"]
+        return features
 
     # Copied from transformers.models.clip.processing_clip.CLIPProcessor.batch_decode with CLIP->Llama
     def batch_decode(self, *args, **kwargs):

src/transformers/models/clap/processing_clap.py

@@ -16,8 +16,22 @@
 Audio/Text processor class for CLAP
 """
 
-from ...processing_utils import ProcessorMixin
-from ...tokenization_utils_base import BatchEncoding
+import sys
+from typing import List, Optional, Union
+
+from ...feature_extraction_utils import BatchFeature
+from ...processing_utils import ProcessingKwargs, ProcessorMixin
+from ...tokenization_utils_base import AudioInput, PreTokenizedInput, TextInput
+
+
+if sys.version_info >= (3, 11):
+    from typing import Unpack
+else:
+    from typing_extensions import Unpack
+
+
+class ClapProcessorKwargs(ProcessingKwargs, total=False):
+    _defaults = {}
 
 
 class ClapProcessor(ProcessorMixin):
@@ -40,7 +54,14 @@ class ClapProcessor(ProcessorMixin):
     def __init__(self, feature_extractor, tokenizer):
         super().__init__(feature_extractor, tokenizer)
 
-    def __call__(self, text=None, audios=None, return_tensors=None, **kwargs):
+    def __call__(
+        self,
+        text: Optional[Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]]] = None,
+        audios: Optional[AudioInput] = None,
+        images=None,
+        videos=None,
+        **kwargs: Unpack[ClapProcessorKwargs],
+    ) -> BatchFeature:
         """
         Main method to prepare for the model one or several sequences(s) and audio(s). This method forwards the `text`
         and `kwargs` arguments to RobertaTokenizerFast's [`~RobertaTokenizerFast.__call__`] if `text` is not `None` to
@@ -49,52 +70,42 @@ def __call__(self, text=None, audios=None, return_tensors=None, **kwargs):
         doctsring of the above two methods for more information.
 
         Args:
-            text (`str`, `List[str]`, `List[List[str]]`):
+            text (`TextInput`, `PreTokenizedInput`, `List[TextInput]`, `List[PreTokenizedInput]`, *optional*):
                 The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
                 (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
                 `is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
-            audios (`np.ndarray`, `torch.Tensor`, `List[np.ndarray]`, `List[torch.Tensor]`):
+            audio (`AudioInput`, *optional*):
                 The audio or batch of audios to be prepared. Each audio can be NumPy array or PyTorch tensor. In case
                 of a NumPy array/PyTorch tensor, each audio should be of shape (C, T), where C is a number of channels,
                 and T the sample length of the audio.
 
-            return_tensors (`str` or [`~utils.TensorType`], *optional*):
-                If set, will return tensors of a particular framework. Acceptable values are:
-
-                - `'tf'`: Return TensorFlow `tf.constant` objects.
-                - `'pt'`: Return PyTorch `torch.Tensor` objects.
-                - `'np'`: Return NumPy `np.ndarray` objects.
-                - `'jax'`: Return JAX `jnp.ndarray` objects.
-
         Returns:
-            [`BatchEncoding`]: A [`BatchEncoding`] with the following fields:
+            [`BatchFeature`]: A [`BatchFeature`] with the following fields:
 
             - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`.
             - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
               `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not
               `None`).
             - **audio_features** -- Audio features to be fed to a model. Returned when `audios` is not `None`.
         """
-        sampling_rate = kwargs.pop("sampling_rate", None)
 
         if text is None and audios is None:
             raise ValueError("You have to specify either text or audios. Both cannot be none.")
 
-        if text is not None:
-            encoding = self.tokenizer(text, return_tensors=return_tensors, **kwargs)
+        output_kwargs = self._merge_kwargs(
+            ClapProcessorKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
 
+        data = {}
+        if text is not None:
+            text_features = self.tokenizer(text, **output_kwargs["text_kwargs"])
+            data.update(text_features)
         if audios is not None:
-            audio_features = self.feature_extractor(
-                audios, sampling_rate=sampling_rate, return_tensors=return_tensors, **kwargs
-            )
-
-        if text is not None and audios is not None:
-            encoding.update(audio_features)
-            return encoding
-        elif text is not None:
-            return encoding
-        else:
-            return BatchEncoding(data=dict(**audio_features), tensor_type=return_tensors)
+            audio_features = self.feature_extractor(audios, **output_kwargs["audio_kwargs"])
+            data.update(audio_features)
+        return BatchFeature(data=data, tensor_type=output_kwargs["common_kwargs"].get("return_tensors"))
 
     def batch_decode(self, *args, **kwargs):
         """