Fix transformers update compatibility

hyvideo/text_encoder/processing_llava.py

@@ -20,10 +20,63 @@ from typing import List, Union
 
 from transformers.feature_extraction_utils import BatchFeature
 from transformers.image_utils import ImageInput, get_image_size, to_numpy_array
-from transformers.processing_utils import ProcessingKwargs, ProcessorMixin, Unpack, _validate_images_text_input_order
+from transformers.processing_utils import ProcessingKwargs, ProcessorMixin, Unpack
 from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
 from transformers.utils import logging
 
+def _validate_images_text_input_order(images, text):
+    """
+    For backward compatibility: reverse the order of `images` and `text` inputs if they are swapped.
+    This method should only be called for processors where `images` and `text` have been swapped for uniformization purposes.
+    Note that this method assumes that two `None` inputs are valid inputs. If this is not the case, it should be handled
+    in the processor's `__call__` method before calling this method.
+    """
+
+    def _is_valid_images_input_for_processor(imgs):
+        # If we have an list of images, make sure every image is valid
+        if isinstance(imgs, (list, tuple)):
+            for img in imgs:
+                if not _is_valid_images_input_for_processor(img):
+                    return False
+        # If not a list or tuple, we have been given a single image or batched tensor of images
+        elif imgs is None:
+            return False
+        return True
+
+    def _is_valid_text_input_for_processor(t):
+        if isinstance(t, str):
+            # Strings are fine
+            return True
+        elif isinstance(t, (list, tuple)):
+            # List are fine as long as they are...
+            if len(t) == 0:
+                # ... not empty
+                return False
+            for t_s in t:
+                return _is_valid_text_input_for_processor(t_s)
+        return False
+
+    def _is_valid(input, validator):
+        return validator(input) or input is None
+
+    images_is_valid = _is_valid(images, _is_valid_images_input_for_processor)
+    images_is_text = _is_valid_text_input_for_processor(images)
+
+    text_is_valid = _is_valid(text, _is_valid_text_input_for_processor)
+    text_is_images = _is_valid_images_input_for_processor(text)
+    # Handle cases where both inputs are valid
+    if images_is_valid and text_is_valid:
+        return images, text
+
+    # Handle cases where inputs need to and can be swapped
+    if (images is None and text_is_images) or (text is None and images_is_text) or (images_is_text and text_is_images):
+        logger.warning_once(
+            "You may have used the wrong order for inputs. `images` should be passed before `text`. "
+            "The `images` and `text` inputs will be swapped. This behavior will be deprecated in transformers v4.47."
+        )
+        return text, images
+
+    raise ValueError("Invalid input type. Check that `images` and/or `text` are valid inputs.")
 
 logger = logging.get_logger(__name__)