# Copyright 2024 The HuggingFace Inc. team. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. from typing import Optional, Union from transformers.models.paligemma.processing_paligemma import PaliGemmaProcessor from ...feature_extraction_utils import BatchFeature from ...image_utils import ImageInput, make_flat_list_of_images from ...processing_utils import ProcessingKwargs, ProcessorMixin, Unpack from ...tokenization_utils_base import PreTokenizedInput, TextInput from ...utils import auto_docstring, is_torch_available, logging if is_torch_available(): import torch logger = logging.get_logger(__name__) class ColPaliProcessorKwargs(ProcessingKwargs, total=False): _defaults = { "text_kwargs": { "padding": "longest", "return_mm_token_type_ids": False, "return_text_replacement_offsets": False, }, "images_kwargs": { "data_format": "channels_first", "do_convert_rgb": True, }, "common_kwargs": {"return_tensors": "pt"}, } class ColPaliProcessor(PaliGemmaProcessor): valid_processor_kwargs = ColPaliProcessorKwargs def __init__( self, image_processor=None, tokenizer=None, chat_template=None, visual_prompt_prefix: str = "Describe the image.", query_prefix: str = "Question: ", ): r""" visual_prompt_prefix (`str`, *optional*, defaults to `"Describe the image."`): A string that gets tokenized and prepended to the image tokens. query_prefix (`str`, *optional*, defaults to `"Question: "`): A prefix to be used for the query. """ self.visual_prompt_prefix = visual_prompt_prefix self.query_prefix = query_prefix super().__init__(image_processor=image_processor, tokenizer=tokenizer, chat_template=chat_template) @property def query_augmentation_token(self) -> str: """ Return the query augmentation token. Query augmentation buffers are used as reasoning buffers during inference. """ return self.tokenizer.pad_token @auto_docstring def __call__( self, images: ImageInput | None = None, text: TextInput | PreTokenizedInput | list[TextInput] | list[PreTokenizedInput] = None, **kwargs: Unpack[ColPaliProcessorKwargs], ) -> BatchFeature: if text is not None and images is not None: raise ValueError("Only one of text or images can be processed at a time") kwargs["return_token_type_ids"] = True output_kwargs = self._merge_kwargs( self.valid_processor_kwargs, tokenizer_init_kwargs=self.tokenizer.init_kwargs, **kwargs, ) suffix = output_kwargs["text_kwargs"].pop("suffix", self.query_augmentation_token * 10) if text is not None: text = [f"{self.tokenizer.bos_token}{self.query_prefix}{sample}{suffix}\n" for sample in text] if output_kwargs["text_kwargs"].get("max_length", None) is not None: output_kwargs["text_kwargs"]["max_length"] += self.image_seq_length model_inputs = ProcessorMixin.__call__(images=images, text=text, **output_kwargs) if images is not None: model_inputs["labels"] = model_inputs["input_ids"].masked_fill(model_inputs["token_type_ids"] == 0, -100) return model_inputs def validate_inputs( self, images: ImageInput | None = None, text: TextInput | PreTokenizedInput | list[TextInput] | list[PreTokenizedInput] | None = None, **kwargs: Unpack[ProcessingKwargs], ): ProcessorMixin.validate_inputs(images=images, text=text) if text is None and images is None: raise ValueError("Either text or images must be provided") def prepare_inputs_layout(self, images=None, text=None, **kwargs): images, text, *_ = ProcessorMixin.prepare_inputs_layout(images=images, text=text, **kwargs) if images is not None: images = make_flat_list_of_images(images) text = [ f"{self.image_token}{self.tokenizer.bos_token}{self.visual_prompt_prefix}\n" for _ in range(len(images)) ] return images, text, None, None def replace_image_token(self, image_inputs: dict, image_idx: int, **kwargs) -> str: return self.image_token * self.image_seq_length @auto_docstring( custom_intro="This method forwards the `images` and `kwargs` arguments to ColPaliProcessor's [`~ColPaliProcessor.__call__`]." ) def process_images( self, images: ImageInput | None = None, **kwargs, ) -> BatchFeature: return self.__call__(images=images, **kwargs) @auto_docstring( custom_intro="This method forwards the `text` and `kwargs` arguments to ColPaliProcessor's [`~ColPaliProcessor.__call__`]." ) def process_queries( self, text: TextInput | list[TextInput], **kwargs, ) -> BatchFeature: return self.__call__(text=text, **kwargs) def score_retrieval( self, query_embeddings: Union["torch.Tensor", list["torch.Tensor"]], passage_embeddings: Union["torch.Tensor", list["torch.Tensor"]], batch_size: int = 128, output_dtype: Optional["torch.dtype"] = None, output_device: Union["torch.device", str] = "cpu", ) -> "torch.Tensor": """ Compute the late-interaction/MaxSim score (ColBERT-like) for the given multi-vector query embeddings (`qs`) and passage embeddings (`ps`). For ColPali, a passage is the image of a document page. Because the embedding tensors are multi-vector and can thus have different shapes, they should be fed as: (1) a list of tensors, where the i-th tensor is of shape (sequence_length_i, embedding_dim) (2) a single tensor of shape (n_passages, max_sequence_length, embedding_dim) -> usually obtained by padding the list of tensors. Args: query_embeddings (`Union[torch.Tensor, list[torch.Tensor]`): Query embeddings. passage_embeddings (`Union[torch.Tensor, list[torch.Tensor]`): Passage embeddings. batch_size (`int`, *optional*, defaults to 128): Batch size for computing scores. output_dtype (`torch.dtype`, *optional*, defaults to `torch.float32`): The dtype of the output tensor. If `None`, the dtype of the input embeddings is used. output_device (`torch.device` or `str`, *optional*, defaults to "cpu"): The device of the output tensor. Returns: `torch.Tensor`: A tensor of shape `(n_queries, n_passages)` containing the scores. The score tensor is saved on the "cpu" device. """ if len(query_embeddings) == 0: raise ValueError("No queries provided") if len(passage_embeddings) == 0: raise ValueError("No passages provided") if query_embeddings[0].device != passage_embeddings[0].device: raise ValueError("Queries and passages must be on the same device") if query_embeddings[0].dtype != passage_embeddings[0].dtype: raise ValueError("Queries and passages must have the same dtype") if output_dtype is None: output_dtype = query_embeddings[0].dtype scores: list[torch.Tensor] = [] for i in range(0, len(query_embeddings), batch_size): batch_scores: list[torch.Tensor] = [] batch_queries = torch.nn.utils.rnn.pad_sequence( query_embeddings[i : i + batch_size], batch_first=True, padding_value=0 ) for j in range(0, len(passage_embeddings), batch_size): batch_passages = torch.nn.utils.rnn.pad_sequence( passage_embeddings[j : j + batch_size], batch_first=True, padding_value=0 ) batch_scores.append( torch.einsum("bnd,csd->bcns", batch_queries, batch_passages).max(dim=3)[0].sum(dim=2) ) scores.append(torch.cat(batch_scores, dim=1).to(output_dtype).to(output_device)) return torch.cat(scores, dim=0) __all__ = [ "ColPaliProcessor", ]