WIP: Retrieval Augmented Diffusion Models #1846

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed

isamu-isozaki wants to merge 96 commits into huggingface:main from isamu-isozaki:rdm_retrieval

src/diffusers/__init__.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -123,6 +123,7 @@ @@
             VersatileDiffusionPipeline,
             VersatileDiffusionTextToImagePipeline,
             VQDiffusionPipeline,
+            RDMPipeline,
         )
     try:
@@ Expand Down @@

src/diffusers/models/retriever.py

-Original file line number
+Diff line change
@@ -0,0 +1,234 @@
+    """
+    Idea for structure
+     Retriever aggregates an Index class and a RetrieverConfig class
+     The Index class aggregates a Dataset and RetrieverConfig class
+     from_pretrained in the retriever's class, it takes in a huggingface path to a dataset, optional path to an index file+config file in huggingface if there is one
+     If an index file is provided, add that index to the dataset.
+     If the dataset doesn't have the column embedding or a corresponding index file, in the Index class, the index is computed based on the clip model defined in the config. Then add that to the index of the dataset. This is done in the Index class
+     In retrieve we just call the retrieve method in the Index class that gets knn based on the faiss embedding.
+     In the save_pretrained method, save index using save_faiss_index. Save this dataset along with config.
+     The call method will just call retrieve.
+     I'll also have a way to pass the clip model and its components via default arguments.
+     Test save_pretrained and from_pretrained methods on new dataset.
+    """
+    from transformers import CLIPModel, CLIPFeatureExtractor, CLIPTokenizer, PretrainedConfig
+    from datasets import load_dataset, Image, load_dataset_builder, load_from_disk, Dataset
+    import torch
+    from typing import Callable, List, Optional, Union
+    import numpy as np
+    from ..utils import deprecate, logging
+    from transformers.models.rag.retrieval_rag import LegacyIndex, CustomHFIndex, CanonicalHFIndex
+    logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+    from diffusers.pipelines.rdm.pipeline_rdm import preprocess_images, normalize_images
+    import os
+    from torch.nn import Module
+    class IndexConfig(PretrainedConfig):
+        def __init__(
+            self,
+            clip_name_or_path="openai/clip-vit-large-patch14",
+            dataset_name="Isamu136/oxford_pets_with_l14_emb",
+            image_column="image",
+            index_name="embeddings",
+            index_path=None,
+            dataset_save_path=None,
+            dataset_set="train",
+            **kwargs,
+        ):
+            super().__init__(**kwargs)
+            self.clip_name_or_path = clip_name_or_path
+            self.dataset_name = dataset_name
+            self.image_column = image_column
+            self.index_name = index_name
+            self.index_path = index_path
+            self.dataset_save_path = dataset_save_path
+            self.dataset_set = dataset_set
+    class Index:
+        """
+        Each index for a retrieval model is specific to the clip model used and the dataset used.
+        """
+        def __init__(self, config: IndexConfig, dataset: Dataset):
+            self.config = config
+            self.dataset = dataset
+            self.index_initialized = False
+            self.index_name = config.index_name
+            self.index_path = config.index_path
+            self.init_index()
+        def set_index_name(self, index_name: str):
+            self.index_name = index_name
+        def init_index(self):
+            if not self.index_initialized:
+                if self.index_path and self.index_name:
+                    try:
+                        self.dataset.load_faiss_index(self.index_name, self.index_path)
+                        self.index_initialized = True
+                    except:
+                        logger.info("Index not initialized")
+                if self.index_name in self.dataset.features:
+                    self.dataset.add_faiss_index(column=self.index_name)
+                    self.index_initialized = True
+        def build_index(
+            self,
+            model=None,
+            feature_extractor: CLIPFeatureExtractor = None,
+            torch_dtype=torch.float32,
+        ):
+            if not self.index_initialized:
+                model = model or CLIPModel.from_pretrained(self.config.clip_name_or_path).to(dtype=torch_dtype)
+                feature_extractor = feature_extractor or CLIPFeatureExtractor.from_pretrained(
+                    self.config.clip_name_or_path
+                )
+                self.dataset = get_dataset_with_emb_from_model(
+                    self.dataset,
+                    model,
+                    feature_extractor,
+                    device=model.device,
+                    image_column=self.config.image_column,
+                    index_name=self.config.index_name,
+                )
+                self.init_index()
+        def retrieve_imgs(self, vec, k: int = 20):
+            vec = np.array(vec).astype(np.float32)
+            return self.dataset.get_nearest_examples(self.index_name, vec, k=k)
+        def retrieve_indices(self, vec, k: int = 20):
+            vec = np.array(vec).astype(np.float32)
+            return self.dataset.search(self.index_name, vec, k=k)
+    class Retriever:
+        def __init__(
+            self,
+            config: IndexConfig,
+            index: Index = None,
+            dataset: Dataset = None,
+            model=None,
+            feature_extractor: CLIPFeatureExtractor = None,
+        ):
+            self.config = config
+            self.index = index or self._build_index(config, dataset, model=model, feature_extractor=feature_extractor)
+        @classmethod
+        def from_pretrained(
+            cls,
+            retriever_name_or_path: str,
+            index: Index = None,
+            dataset: Dataset = None,
+            model=None,
+            feature_extractor: CLIPFeatureExtractor = None,
+            **kwargs,
+        ):
+            config = kwargs.pop("config", None) or IndexConfig.from_pretrained(retriever_name_or_path, **kwargs)
+            return cls(config, index=index, dataset=dataset, model=model, feature_extractor=feature_extractor)
+        @staticmethod
+        def _build_index(
+            config: IndexConfig, dataset: Dataset = None, model=None, feature_extractor: CLIPFeatureExtractor = None
+        ):
+            dataset = dataset or load_dataset(config.dataset_name)
+            dataset = dataset[config.dataset_set]
+            index = Index(config, dataset)
+            index.build_index(model=model, feature_extractor=feature_extractor)
+            return index
+        def save_pretrained(self, save_directory):
+            os.makedirs(save_directory, exist_ok=True)
+            if self.config.index_path is None:
+                index_path = os.path.join(save_directory, "hf_dataset_index.faiss")
+                self.index.dataset.get_index(self.config.index_name).save(index_path)
+                self.config.index_path = index_path
+            if self.config.dataset_save_path is None:
+                dataset_save_path = os.path.join(save_directory, "hf_dataset")
+                # datasets don't support save_to_disk with indexes right now
+                faiss_index = self.index.dataset._indexes.pop(self.config.index_name)
+                self.index.dataset.save_to_disk(dataset_save_path)
+                self.index.dataset._indexes[self.config.index_name] = faiss_index
+                self.config.dataset_save_path = dataset_save_path
+            self.config.save_pretrained(save_directory)
+        def init_retrieval(self):
+            logger.info("initializing retrieval")
+            self.index.init_index()
+        def retrieve_imgs(self, embeddings: np.ndarray, k: int):
+            return self.index.retrieve_imgs(embeddings, k)
+        def retrieve_indices(self, embeddings: np.ndarray, k: int):
+            return self.index.retrieve_indices(embeddings, k)
+        def __call__(
+            self,
+            embeddings,
+            k: int = 20,
+        ):
+            return self.index.retrieve_imgs(embeddings, k)
+    def map_txt_to_clip_feature(clip_model, tokenizer, prompt):
+        text_inputs = tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=tokenizer.model_max_length,
+            return_tensors="pt",
+        )
+        text_input_ids = text_inputs.input_ids
+        if text_input_ids.shape[-1] > tokenizer.model_max_length:
+            removed_text = tokenizer.batch_decode(text_input_ids[:, tokenizer.model_max_length :])
+            logger.warning(
+                "The following part of your input was truncated because CLIP can only handle sequences up to"
+                f" {tokenizer.model_max_length} tokens: {removed_text}"
+            )
+            text_input_ids = text_input_ids[:, : tokenizer.model_max_length]
+        text_embeddings = clip_model.get_text_features(text_input_ids.to(clip_model.device))
+        text_embeddings = text_embeddings / torch.linalg.norm(text_embeddings, dim=-1, keepdim=True)
+        text_embeddings = text_embeddings[:, None, :]
+        return text_embeddings[0][0].cpu().detach().numpy()
+    def map_img_to_model_feature(model, feature_extractor, imgs):
+        for i, image in enumerate(imgs):
+            if not image.mode == "RGB":
+                imgs[i] = image.convert("RGB")
+        imgs = normalize_images(imgs)
+        retrieved_images = preprocess_images(imgs, feature_extractor).to(model.device)
+        image_embeddings = model(retrieved_images)
+        image_embeddings = image_embeddings / torch.linalg.norm(image_embeddings, dim=-1, keepdim=True)
+        image_embeddings = image_embeddings[None, ...]
+        return image_embeddings
+    def get_dataset_with_emb_from_model(dataset, model, feature_extractor, image_column="image", index_name="embeddings"):
+        return dataset.map(
+            lambda example: {
+                index_name: map_img_to_model_feature(model, feature_extractor, [example[image_column]], model.device)
+                .cpu()
+                .detach()
+                .numpy()[0][0]
+            }
+        )
+    def get_dataset_with_emb_from_clip_model(
+        dataset, clip_model, feature_extractor, image_column="image", index_name="embeddings"
+    ):
+        return dataset.map(
+            lambda example: {
+                index_name: map_img_to_model_feature(
+                    clip_model.get_image_features, feature_extractor, [example[image_column]], clip_model.device
+                )
+                .cpu()
+                .detach()
+                .numpy()[0][0]
+            }
+        )

src/diffusers/pipelines/__init__.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -62,6 +62,7 @@ @@
             VersatileDiffusionTextToImagePipeline,
         )
         from .vq_diffusion import VQDiffusionPipeline
+        from .rdm import RDMPipeline
     try:
         if not is_onnx_available():
@@ Expand Down @@

src/diffusers/pipelines/rdm/__init__.py

-Original file line number
+Diff line change
@@ -0,0 +1,5 @@
+    from ...utils import is_torch_available, is_transformers_available
+    if is_transformers_available() and is_torch_available():
+        from .pipeline_rdm import RDMPipeline

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

WIP: Retrieval Augmented Diffusion Models #1846

Uh oh!

Diff view

Diff view

There are no files selected for viewing

Uh oh!