transformers 5.0.0rc3__py3-none-any.whl → 5.1.0__py3-none-any.whl

transformers/models/llava_next_video/modular_llava_next_video.py

@@ -31,8 +31,10 @@ from transformers.models.llava_next.modeling_llava_next import (
 from ...cache_utils import Cache
 from ...configuration_utils import PreTrainedConfig
 from ...modeling_flash_attention_utils import FlashAttentionKwargs
+from ...modeling_outputs import BaseModelOutputWithPooling
 from ...processing_utils import Unpack
-from ...utils import logging
+from ...utils import auto_docstring, logging, torch_compilable_check
+from ...utils.generic import check_model_inputs
 from ..auto import CONFIG_MAPPING, AutoConfig
 
 
@@ -80,6 +82,8 @@ class LlavaNextVideoConfig(PreTrainedConfig):
             Sequence length of one image embedding.
         video_seq_length (`int`, *optional*, defaults to 288):
             Sequence length of one video embedding.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
 
     Example:
 
@@ -122,6 +126,7 @@ class LlavaNextVideoConfig(PreTrainedConfig):
         spatial_pool_stride=2,
         image_seq_length=576,
         video_seq_length=288,
+        tie_word_embeddings=False,
         **kwargs,
     ):
         self.video_token_index = video_token_index
@@ -132,6 +137,7 @@ class LlavaNextVideoConfig(PreTrainedConfig):
         self.image_token_index = image_token_index
         self.projector_hidden_act = projector_hidden_act
         self.multimodal_projector_bias = multimodal_projector_bias
+        self.tie_word_embeddings = tie_word_embeddings
 
         if vision_feature_select_strategy not in ["default", "full"]:
             raise ValueError(
@@ -173,13 +179,13 @@ class LlavaNextVideoConfig(PreTrainedConfig):
 
         self.text_config = text_config
 
-        super().__init__(**kwargs)
+        # The default value is `False` but this config is used with many model types
+        # Attr `tie_word_embeddings` was saved in text config for those models, so we
+        # need an ugly workaround and forward-pass the attr from text config
+        if not tie_word_embeddings and self.text_config.tie_word_embeddings:
+            self.tie_word_embeddings = self.text_config.tie_word_embeddings
 
-        # Due to a mismatch at model addition-time, the `tie_word_embeddings` was saved in the text config, even
-        # though it concerns the main model, while it was set to False by default in the main model... So we hardcode a fix here
-        if not self.tie_word_embeddings and self.text_config.tie_word_embeddings:
-            self.tie_word_embeddings = True
-            self.text_config.tie_word_embeddings = False
+        super().__init__(**kwargs)
 
 
 class LlavaNextVideoModelOutputWithPast(LlavaNextModelOutputWithPast):
@@ -270,41 +276,32 @@ class LlavaNextVideoModel(LlavaNextModel):
         self.vision_resampler = LlavaNextVideoPooler(config)
         self.post_init()
 
+    @check_model_inputs(tie_last_hidden_states=False)
+    @auto_docstring(
+        custom_intro="Obtains image last hidden states from the vision tower and apply multimodal projection."
+    )
     def get_image_features(
         self,
         pixel_values: torch.FloatTensor,
         image_sizes: torch.Tensor,
         vision_feature_layer: int | list[int] | None = None,
         vision_feature_select_strategy: str | None = None,
-    ):
-        """
-        Obtains image last hidden states from the vision tower and apply multimodal projection.
-
-        Args:
-            pixel_values (`torch.FloatTensor]` of shape `(batch_size, num_patches, channels, height, width)`)
-               The tensors corresponding to the input images.
-            image_sizes (`torch.Tensor` of shape `(num_images, 2)`)
-                Actual image size of each images (H, W).
-            vision_feature_layer (`Union[int, list[int]]`, *optional*):
-                The index of the layer to select the vision feature. If multiple indices are provided,
-                the vision feature of the corresponding indices will be concatenated to form the
-                vision features.
-            vision_feature_select_strategy (`str`, *optional*):
-                The feature selection strategy used to select the vision feature from the vision backbone.
-                Can be one of `"default"` or `"full"`
-        Returns:
-            image_features (list[`torch.Tensor`]): List of image feature tensor, each contains all the visual feature of all patches
-            and are of shape `(num_patches, image_length, embed_dim)`).
+        output_hidden_states: bool | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple | BaseModelOutputWithPooling:
+        r"""
+        pixel_values (`torch.FloatTensor]` of shape `(batch_size, num_patches, channels, height, width)`)
+            The tensors corresponding to the input images.
+        image_sizes (`torch.Tensor` of shape `(num_images, 2)`)
+            Actual image size of each images (H, W).
+        vision_feature_layer (`Union[int, list[int]]`, *optional*):
+            The index of the layer to select the vision feature. If multiple indices are provided,
+            the vision feature of the corresponding indices will be concatenated to form the
+            vision features.
+        vision_feature_select_strategy (`str`, *optional*):
+            The feature selection strategy used to select the vision feature from the vision backbone.
+            Can be one of `"default"` or `"full"`
         """
-        vision_feature_layer = (
-            vision_feature_layer if vision_feature_layer is not None else self.config.vision_feature_layer
-        )
-        vision_feature_select_strategy = (
-            vision_feature_select_strategy
-            if vision_feature_select_strategy is not None
-            else self.config.vision_feature_select_strategy
-        )
-
         # ! infer image_num_patches from image_sizes
         image_num_patches = [
             image_size_to_num_patches(
@@ -322,13 +319,18 @@ class LlavaNextVideoModel(LlavaNextModel):
             # otherwise has to be stacked from list of (num_patches, num_channels, height, width)
             raise ValueError(f"pixel_values of shape {pixel_values.shape}, expect to be of 4 or 5 dimensions")
 
-        image_features = self.vision_tower(pixel_values, output_hidden_states=True)
+        image_outputs = self.vision_tower(
+            pixel_values,
+            output_hidden_states=True,  # Ignore arg on purpose
+            return_dict=True,
+            **kwargs,
+        )
         # If we have one vision feature layer, return the corresponding hidden states,
         # otherwise, select the hidden states of each feature layer and concatenate them
         if isinstance(vision_feature_layer, int):
-            selected_image_feature = image_features.hidden_states[vision_feature_layer]
+            selected_image_feature = image_outputs.hidden_states[vision_feature_layer]
         else:
-            hs_pool = [image_features.hidden_states[layer_idx] for layer_idx in vision_feature_layer]
+            hs_pool = [image_outputs.hidden_states[layer_idx] for layer_idx in vision_feature_layer]
             selected_image_feature = torch.cat(hs_pool, dim=-1)
 
         if vision_feature_select_strategy == "default":
@@ -342,50 +344,48 @@ class LlavaNextVideoModel(LlavaNextModel):
             vision_feature_select_strategy,
             image_newline=self.image_newline,
         )
-        return image_features
+        image_outputs.pooler_output = image_features
 
+        return image_outputs
+
+    @check_model_inputs(tie_last_hidden_states=False)
+    @auto_docstring(
+        custom_intro="Obtains video last hidden states from the vision tower and apply multimodal projection."
+    )
     def get_video_features(
         self,
         pixel_values: torch.FloatTensor,
         vision_feature_layer: int | list[int] | None = None,
         vision_feature_select_strategy: str | None = None,
-    ):
-        """
-        Obtains video last hidden states from the vision tower and apply multimodal projection.
-
-        Args:
-            pixel_values (`torch.FloatTensor]` of shape `(batch_size, num_frames, channels, height, width)`)
-               The tensors corresponding to the input video.
-            vision_feature_layer (`Union[int, list[int]]`, *optional;*):
-                The index of the layer to select the vision feature. If multiple indices are provided,
-                the vision feature of the corresponding indices will be concatenated to form the
-                vision features.
-            vision_feature_select_strategy (`str`, *optional*):
-                The feature selection strategy used to select the vision feature from the vision backbone.
-                Can be one of `"default"` or `"full"`
-        Returns:
-            video_features (list[`torch.Tensor`]): List of video feature tensor, each contains all the visual feature of all patches
-            and are of shape `(num_videos, video_length, embed_dim)`).
+        output_hidden_states: bool | None = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple | BaseModelOutputWithPooling:
+        r"""
+        pixel_values (`torch.FloatTensor]` of shape `(batch_size, num_frames, channels, height, width)`)
+            The tensors corresponding to the input video.
+        vision_feature_layer (`Union[int, list[int]]`, *optional;*):
+            The index of the layer to select the vision feature. If multiple indices are provided,
+            the vision feature of the corresponding indices will be concatenated to form the
+            vision features.
+        vision_feature_select_strategy (`str`, *optional*):
+            The feature selection strategy used to select the vision feature from the vision backbone.
+            Can be one of `"default"` or `"full"`
         """
-        vision_feature_layer = (
-            vision_feature_layer if vision_feature_layer is not None else self.config.vision_feature_layer
-        )
-        vision_feature_select_strategy = (
-            vision_feature_select_strategy
-            if vision_feature_select_strategy is not None
-            else self.config.vision_feature_select_strategy
-        )
-
         batch_size, frames, channels, height, width = pixel_values.shape
         pixel_values = pixel_values.reshape(batch_size * frames, channels, height, width)
-        video_features = self.vision_tower(pixel_values, output_hidden_states=True)
+        video_outputs = self.vision_tower(
+            pixel_values,
+            output_hidden_states=True,  # Ignore arg on purpose
+            return_dict=True,
+            **kwargs,
+        )
 
         # If we have one vision feature layer, return the corresponding hidden states,
         # otherwise, select the hidden states of each feature layer and concatenate them
         if isinstance(vision_feature_layer, int):
-            selected_video_features = video_features.hidden_states[vision_feature_layer]
+            selected_video_features = video_outputs.hidden_states[vision_feature_layer]
         else:
-            hs_pool = [video_features.hidden_states[layer_idx] for layer_idx in vision_feature_layer]
+            hs_pool = [video_outputs.hidden_states[layer_idx] for layer_idx in vision_feature_layer]
             selected_video_features = torch.cat(hs_pool, dim=-1)
 
         if vision_feature_select_strategy == "default":
@@ -395,7 +395,9 @@ class LlavaNextVideoModel(LlavaNextModel):
         video_features = self.vision_resampler(selected_video_features)
         video_features = self.multi_modal_projector(video_features)
         video_features = torch.split(video_features, frames, dim=0)
-        return video_features
+        video_outputs.pooler_output = video_features
+
+        return video_outputs
 
     def get_placeholder_mask(
         self,
@@ -423,20 +425,23 @@ class LlavaNextVideoModel(LlavaNextModel):
 
         n_image_tokens = special_image_mask.sum()
         special_image_mask = special_image_mask.unsqueeze(-1).expand_as(inputs_embeds).to(inputs_embeds.device)
-        if image_features is not None and inputs_embeds[special_image_mask].numel() != image_features.numel():
-            raise ValueError(
-                f"Image features and image tokens do not match: tokens: {n_image_tokens}, features {image_features.shape[0]}"
+        if image_features is not None:
+            torch_compilable_check(
+                inputs_embeds[special_image_mask].numel() == image_features.numel(),
+                f"Image features and image tokens do not match, tokens: {n_image_tokens}, features: {image_features.shape[0]}",
             )
 
         n_video_tokens = special_video_mask.sum()
         special_video_mask = special_video_mask.unsqueeze(-1).expand_as(inputs_embeds).to(inputs_embeds.device)
-        if video_features is not None and inputs_embeds[special_video_mask].numel() != video_features.numel():
-            raise ValueError(
-                f"Videos features and image tokens do not match: tokens: {n_video_tokens}, features {video_features.shape[0]}"
+        if video_features is not None:
+            torch_compilable_check(
+                inputs_embeds[special_video_mask].numel() == video_features.numel(),
+                f"Video features and video tokens do not match, tokens: {n_video_tokens}, features: {video_features.shape[0]}",
             )
-
         return special_image_mask, special_video_mask
 
+    @check_model_inputs(tie_last_hidden_states=False)
+    @auto_docstring
     def forward(
         self,
         input_ids: torch.LongTensor | None = None,
@@ -461,14 +466,6 @@ class LlavaNextVideoModel(LlavaNextModel):
             output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
         )
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
-        self.vision_feature_layer = (
-            vision_feature_layer if vision_feature_layer is not None else self.config.vision_feature_layer
-        )
-        self.vision_feature_select_strategy = (
-            vision_feature_select_strategy
-            if vision_feature_select_strategy is not None
-            else self.config.vision_feature_select_strategy
-        )
 
         if (input_ids is None) ^ (inputs_embeds is not None):
             raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
@@ -480,9 +477,10 @@ class LlavaNextVideoModel(LlavaNextModel):
             image_features = self.get_image_features(
                 pixel_values,
                 image_sizes,
-                vision_feature_layer=self.vision_feature_layer,
-                vision_feature_select_strategy=self.vision_feature_select_strategy,
-            )
+                vision_feature_layer=vision_feature_layer,
+                vision_feature_select_strategy=vision_feature_select_strategy,
+                return_dict=True,
+            ).pooler_output
             image_features = torch.cat(image_features, dim=0).to(inputs_embeds.device, inputs_embeds.dtype)
             special_image_mask, _ = self.get_placeholder_mask(
                 input_ids, inputs_embeds=inputs_embeds, image_features=image_features
@@ -492,9 +490,10 @@ class LlavaNextVideoModel(LlavaNextModel):
         if pixel_values_videos is not None:
             video_features = self.get_video_features(
                 pixel_values_videos,
-                vision_feature_layer=self.vision_feature_layer,
-                vision_feature_select_strategy=self.vision_feature_select_strategy,
-            )
+                vision_feature_layer=vision_feature_layer,
+                vision_feature_select_strategy=vision_feature_select_strategy,
+                return_dict=True,
+            ).pooler_output
             video_features = [feature.flatten(0, 1) for feature in video_features]
             video_feature_lens = [feature.size(0) for feature in video_features]
             video_features = torch.cat(video_features, dim=0)
@@ -530,18 +529,34 @@ class LlavaNextVideoModel(LlavaNextModel):
 
 
 class LlavaNextVideoForConditionalGeneration(LlavaNextForConditionalGeneration):
+    @auto_docstring
     def get_video_features(
         self,
         pixel_values: torch.FloatTensor,
         vision_feature_layer: int | list[int] | None = None,
         vision_feature_select_strategy: str | None = None,
-    ):
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple | BaseModelOutputWithPooling:
+        r"""
+        pixel_values (`torch.FloatTensor]` of shape `(batch_size, num_frames, channels, height, width)`)
+            The tensors corresponding to the input video.
+        vision_feature_layer (`Union[int, list[int]]`, *optional;*):
+            The index of the layer to select the vision feature. If multiple indices are provided,
+            the vision feature of the corresponding indices will be concatenated to form the
+            vision features.
+        vision_feature_select_strategy (`str`, *optional*):
+            The feature selection strategy used to select the vision feature from the vision backbone.
+            Can be one of `"default"` or `"full"`
+        """
         return self.model.get_video_features(
             pixel_values=pixel_values,
             vision_feature_layer=vision_feature_layer,
             vision_feature_select_strategy=vision_feature_select_strategy,
+            **kwargs,
         )
 
+    @check_model_inputs(tie_last_hidden_states=False)
+    @auto_docstring
     def forward(
         self,
         input_ids: torch.LongTensor | None = None,
@@ -573,7 +588,8 @@ class LlavaNextVideoForConditionalGeneration(LlavaNextForConditionalGeneration):
 
         ```python
         >>> from PIL import Image
-        >>> import requests
+        >>> import httpx
+        >>> from io import BytesIO
         >>> import av
         >>> from transformers import AutoProcessor, LlavaNextVideoForConditionalGeneration
 
@@ -613,7 +629,8 @@ class LlavaNextVideoForConditionalGeneration(LlavaNextForConditionalGeneration):
         >>> # load an image to generate from an image
         >>> prompt = "USER:<image>\nWhat is shown in this image? ASSISTANT:"
         >>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
-        >>> image = Image.open(requests.get(url, stream=True).raw)
+        >>> with httpx.stream("GET", url) as response:
+        ...     image = Image.open(BytesIO(response.read()))
         >>> inputs_image = processor(text=prompt, images=image, return_tensors="pt").to(model.device)
 
         >>> # Generate from video
@@ -631,14 +648,6 @@ class LlavaNextVideoForConditionalGeneration(LlavaNextForConditionalGeneration):
             output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
         )
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
-        vision_feature_layer = (
-            vision_feature_layer if vision_feature_layer is not None else self.config.vision_feature_layer
-        )
-        vision_feature_select_strategy = (
-            vision_feature_select_strategy
-            if vision_feature_select_strategy is not None
-            else self.config.vision_feature_select_strategy
-        )
 
         outputs = self.model(
             input_ids=input_ids,

transformers/models/llava_onevision/configuration_llava_onevision.py

@@ -59,6 +59,8 @@ class LlavaOnevisionConfig(PreTrainedConfig):
             of the form `(height, width)`.
         multimodal_projector_bias (`bool`, *optional*, defaults to `True`):
             Whether to use bias in the multimodal projector.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
 
     Example:
 
@@ -100,12 +102,14 @@ class LlavaOnevisionConfig(PreTrainedConfig):
         vision_aspect_ratio="anyres_max_9",
         image_grid_pinpoints=None,
         multimodal_projector_bias=True,
+        tie_word_embeddings=False,
         **kwargs,
     ):
         self.image_token_index = image_token_index
         self.video_token_index = video_token_index
         self.projector_hidden_act = projector_hidden_act
         self.multimodal_projector_bias = multimodal_projector_bias
+        self.tie_word_embeddings = tie_word_embeddings
 
         if vision_feature_select_strategy not in ["default", "full"]:
             raise ValueError(
@@ -184,13 +188,13 @@ class LlavaOnevisionConfig(PreTrainedConfig):
 
         self.text_config = text_config
 
-        super().__init__(**kwargs)
+        # The default value is `False` but this config is used with many model types
+        # Attr `tie_word_embeddings` was saved in text config for those models, so we
+        # need an ugly workaround and forward-pass the attr from text config
+        if not tie_word_embeddings and self.text_config.tie_word_embeddings:
+            self.tie_word_embeddings = self.text_config.tie_word_embeddings
 
-        # Due to a mismatch at model addition-time, the `tie_word_embeddings` was saved in the text config, even
-        # though it concerns the main model, while it was set to False by default in the main model... So we hardcode a fix here
-        if not self.tie_word_embeddings and self.text_config.tie_word_embeddings:
-            self.tie_word_embeddings = True
-            self.text_config.tie_word_embeddings = False
+        super().__init__(**kwargs)
 
 
 __all__ = ["LlavaOnevisionConfig"]

transformers/models/llava_onevision/image_processing_llava_onevision_fast.py

@@ -21,7 +21,7 @@
 from typing import Optional
 
 import torch
-from torchvision.transforms.v2 import functional as F
+import torchvision.transforms.v2.functional as tvF
 
 from ...image_processing_utils import BatchFeature, get_patch_output_size, select_best_resolution
 from ...image_processing_utils_fast import (
@@ -69,6 +69,7 @@ class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast):
     def preprocess(self, images: ImageInput, **kwargs: Unpack[LlavaOnevisionImageProcessorKwargs]) -> BatchFeature:
         if isinstance(images, (tuple, list)) and isinstance(images[0], (tuple, list)):
             # if the first element is a list, we assume that all elements are lists
+            images = [x for x in images if x]  # handle text-only case
             batch_num_images = [len(x) for x in images]
         elif isinstance(images, (tuple, list)):
             # treat this as a single-image case for backward compatibility
@@ -81,7 +82,7 @@ class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast):
         self,
         image: "torch.Tensor",
         target_resolution: tuple,
-        interpolation: "F.InterpolationMode",
+        interpolation: "tvF.InterpolationMode",
         input_data_format: ChannelDimension,
     ) -> "torch.Tensor":
         """
@@ -127,7 +128,7 @@ class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast):
         new_resolution = get_patch_output_size(image, target_resolution, input_data_format)
         padding = self._get_padding_size(new_resolution, target_resolution)
 
-        padded_image = F.pad(image, padding=padding)
+        padded_image = tvF.pad(image, padding=padding)
 
         return padded_image
 
@@ -137,7 +138,7 @@ class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast):
         grid_pinpoints,
         size: tuple,
         patch_size: int,
-        interpolation: "F.InterpolationMode",
+        interpolation: "tvF.InterpolationMode",
     ) -> list["torch.Tensor"]:
         """
         Process an image with variable resolutions by dividing it into patches.
@@ -169,7 +170,7 @@ class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast):
         )
         padded_image = self._pad_for_patching(resized_image, best_resolution, input_data_format=ChannelDimension.FIRST)
         patches = divide_to_patches(padded_image, patch_size=patch_size)
-        resized_original_image = F.resize(image, size=size, interpolation=interpolation)
+        resized_original_image = tvF.resize(image, size=size, interpolation=interpolation)
 
         image_patches = [resized_original_image] + patches
 
@@ -204,7 +205,7 @@ class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast):
         do_resize: bool,
         size: SizeDict,
         image_grid_pinpoints: list[list[int]],
-        interpolation: Optional["F.InterpolationMode"],
+        interpolation: Optional["tvF.InterpolationMode"],
         do_center_crop: bool,
         crop_size: SizeDict,
         do_rescale: bool,
@@ -320,7 +321,7 @@ class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast):
         paste_y_left = (max_dim - height) // 2
         paste_x_right = max_dim - width - paste_x_left
         paste_y_right = max_dim - height - paste_y_left
-        padded_images = F.pad(
+        padded_images = tvF.pad(
             images, padding=[paste_x_left, paste_y_left, paste_x_right, paste_y_right], fill=background_color
         )