relik/retriever/indexers/faiss.py

import contextlib
import logging
import math
import os
from dataclasses import dataclass
from typing import Callable, List, Optional, Union

import numpy
import torch
from pytorch_modules import RetrievedSample
from torch.utils.data import DataLoader
from tqdm import tqdm

from relik.common.log import get_logger
from relik.common.utils import is_package_available
from relik.retriever.common.model_inputs import ModelInputs
from relik.retriever.data.base.datasets import BaseDataset
from relik.retriever.data.labels import Labels
from relik.retriever.indexers.base import BaseDocumentIndex
from relik.retriever.pytorch_modules import PRECISION_MAP
from relik.retriever.pytorch_modules.model import GoldenRetriever

if is_package_available("faiss"):
    import faiss
    import faiss.contrib.torch_utils

logger = get_logger(__name__, level=logging.INFO)


@dataclass
class FaissOutput:
    indices: Union[torch.Tensor, numpy.ndarray]
    distances: Union[torch.Tensor, numpy.ndarray]


class FaissDocumentIndex(BaseDocumentIndex):
    DOCUMENTS_FILE_NAME = "documents.json"
    EMBEDDINGS_FILE_NAME = "embeddings.pt"
    INDEX_FILE_NAME = "index.faiss"

    def __init__(
        self,
        documents: Union[List[str], Labels],
        embeddings: Optional[Union[torch.Tensor, numpy.ndarray]] = None,
        index=None,
        index_type: str = "Flat",
        metric: int = faiss.METRIC_INNER_PRODUCT,
        normalize: bool = False,
        device: str = "cpu",
        name_or_dir: Optional[Union[str, os.PathLike]] = None,
        *args,
        **kwargs,
    ) -> None:
        super().__init__(documents, embeddings, name_or_dir)

        if embeddings is not None and documents is not None:
            logger.info("Both documents and embeddings are provided.")
            if documents.get_label_size() != embeddings.shape[0]:
                raise ValueError(
                    "The number of documents and embeddings must be the same."
                )

        # device to store the embeddings
        self.device = device

        # params
        self.index_type = index_type
        self.metric = metric
        self.normalize = normalize

        if index is not None:
            self.embeddings = index
            if self.device == "cuda":
                # use a single GPU
                faiss_resource = faiss.StandardGpuResources()
                self.embeddings = faiss.index_cpu_to_gpu(
                    faiss_resource, 0, self.embeddings
                )
        else:
            if embeddings is not None:
                # build the faiss index
                logger.info("Building the index from the embeddings.")
                self.embeddings = self._build_faiss_index(
                    embeddings=embeddings,
                    index_type=index_type,
                    normalize=normalize,
                    metric=metric,
                )

    def _build_faiss_index(
        self,
        embeddings: Optional[Union[torch.Tensor, numpy.ndarray]],
        index_type: str,
        normalize: bool,
        metric: int,
    ):
        # build the faiss index
        self.normalize = (
            normalize
            and metric == faiss.METRIC_INNER_PRODUCT
            and not isinstance(embeddings, torch.Tensor)
        )
        if self.normalize:
            index_type = f"L2norm,{index_type}"
        faiss_vector_size = embeddings.shape[1]
        if self.device == "cpu":
            index_type = index_type.replace("x,", "x_HNSW32,")
        index_type = index_type.replace(
            "x", str(math.ceil(math.sqrt(faiss_vector_size)) * 4)
        )
        self.embeddings = faiss.index_factory(faiss_vector_size, index_type, metric)

        # convert to GPU
        if self.device == "cuda":
            # use a single GPU
            faiss_resource = faiss.StandardGpuResources()
            self.embeddings = faiss.index_cpu_to_gpu(faiss_resource, 0, self.embeddings)
        else:
            # move to CPU if embeddings is a torch.Tensor
            embeddings = (
                embeddings.cpu() if isinstance(embeddings, torch.Tensor) else embeddings
            )

        # convert to float32 if embeddings is a torch.Tensor and is float16
        if isinstance(embeddings, torch.Tensor) and embeddings.dtype == torch.float16:
            embeddings = embeddings.float()

        self.embeddings.add(embeddings)

        # save parameters for saving/loading
        self.index_type = index_type
        self.metric = metric

        # clear the embeddings to free up memory
        embeddings = None

        return self.embeddings

    @torch.no_grad()
    @torch.inference_mode()
    def index(
        self,
        retriever: GoldenRetriever,
        documents: Optional[List[str]] = None,
        batch_size: int = 32,
        num_workers: int = 4,
        max_length: Optional[int] = None,
        collate_fn: Optional[Callable] = None,
        encoder_precision: Optional[Union[str, int]] = None,
        compute_on_cpu: bool = False,
        force_reindex: bool = False,
        *args,
        **kwargs,
    ) -> "FaissDocumentIndex":
        """
        Index the documents using the encoder.

        Args:
            retriever (:obj:`torch.nn.Module`):
                The encoder to be used for indexing.
            documents (:obj:`List[str]`, `optional`, defaults to None):
                The documents to be indexed.
            batch_size (:obj:`int`, `optional`, defaults to 32):
                The batch size to be used for indexing.
            num_workers (:obj:`int`, `optional`, defaults to 4):
                The number of workers to be used for indexing.
            max_length (:obj:`int`, `optional`, defaults to None):
                The maximum length of the input to the encoder.
            collate_fn (:obj:`Callable`, `optional`, defaults to None):
                The collate function to be used for batching.
            encoder_precision (:obj:`Union[str, int]`, `optional`, defaults to None):
                The precision to be used for the encoder.
            compute_on_cpu (:obj:`bool`, `optional`, defaults to False):
                Whether to compute the embeddings on CPU.
            force_reindex (:obj:`bool`, `optional`, defaults to False):
                Whether to force reindexing.

        Returns:
            :obj:`InMemoryIndexer`: The indexer object.
        """

        if self.embeddings is not None and not force_reindex:
            logger.log(
                "Embeddings are already present and `force_reindex` is `False`. Skipping indexing."
            )
            if documents is None:
                return self

        # release the memory
        if collate_fn is None:
            tokenizer = retriever.passage_tokenizer

            def collate_fn(x):
                return ModelInputs(
                    tokenizer(
                        x,
                        padding=True,
                        return_tensors="pt",
                        truncation=True,
                        max_length=max_length or tokenizer.model_max_length,
                    )
                )

        if force_reindex:
            if documents is not None:
                self.documents.add_labels(documents)
            data = [k for k in self.documents.get_labels()]

        else:
            if documents is not None:
                data = [k for k in Labels(documents).get_labels()]
            else:
                return self

        dataloader = DataLoader(
            BaseDataset(name="passage", data=data),
            batch_size=batch_size,
            shuffle=False,
            num_workers=num_workers,
            pin_memory=False,
            collate_fn=collate_fn,
        )

        encoder = retriever.passage_encoder

        # Create empty lists to store the passage embeddings and passage index
        passage_embeddings: List[torch.Tensor] = []

        encoder_device = "cpu" if compute_on_cpu else self.device

        # fucking autocast only wants pure strings like 'cpu' or 'cuda'
        # we need to convert the model device to that
        device_type_for_autocast = str(encoder_device).split(":")[0]
        # autocast doesn't work with CPU and stuff different from bfloat16
        autocast_pssg_mngr = (
            contextlib.nullcontext()
            if device_type_for_autocast == "cpu"
            else (
                torch.autocast(
                    device_type=device_type_for_autocast,
                    dtype=PRECISION_MAP[encoder_precision],
                )
            )
        )
        with autocast_pssg_mngr:
            # Iterate through each batch in the dataloader
            for batch in tqdm(dataloader, desc="Indexing"):
                # Move the batch to the device
                batch: ModelInputs = batch.to(encoder_device)
                # Compute the passage embeddings
                passage_outs = encoder(**batch)
                # Append the passage embeddings to the list
                if self.device == "cpu":
                    passage_embeddings.extend([c.detach().cpu() for c in passage_outs])
                else:
                    passage_embeddings.extend([c for c in passage_outs])

        # move the passage embeddings to the CPU if not already done
        passage_embeddings = [c.detach().cpu() for c in passage_embeddings]
        # stack it
        passage_embeddings: torch.Tensor = torch.stack(passage_embeddings, dim=0)
        # convert to float32 for faiss
        passage_embeddings.to(PRECISION_MAP["float32"])

        # index the embeddings
        self.embeddings = self._build_faiss_index(
            embeddings=passage_embeddings,
            index_type=self.index_type,
            normalize=self.normalize,
            metric=self.metric,
        )
        # free up memory from the unused variable
        del passage_embeddings

        return self

    @torch.no_grad()
    @torch.inference_mode()
    def search(self, query: torch.Tensor, k: int = 1) -> list[list[RetrievedSample]]:
        k = min(k, self.embeddings.ntotal)

        if self.normalize:
            faiss.normalize_L2(query)
        if isinstance(query, torch.Tensor) and self.device == "cpu":
            query = query.detach().cpu()
        # Retrieve the indices of the top k passage embeddings
        retriever_out = self.embeddings.search(query, k)

        # get int values (second element of the tuple)
        batch_top_k: List[List[int]] = retriever_out[1].detach().cpu().tolist()
        # get float values (first element of the tuple)
        batch_scores: List[List[float]] = retriever_out[0].detach().cpu().tolist()
        # Retrieve the passages corresponding to the indices
        batch_passages = [
            [self.documents.get_label_from_index(i) for i in indices]
            for indices in batch_top_k
        ]
        # build the output object
        batch_retrieved_samples = [
            [
                RetrievedSample(label=passage, index=index, score=score)
                for passage, index, score in zip(passages, indices, scores)
            ]
            for passages, indices, scores in zip(
                batch_passages, batch_top_k, batch_scores
            )
        ]
        return batch_retrieved_samples

    # def save(self, saving_dir: Union[str, os.PathLike]):
    #     """
    #     Save the indexer to the disk.

    #     Args:
    #         saving_dir (:obj:`Union[str, os.PathLike]`):
    #             The directory where the indexer will be saved.
    #     """
    #     saving_dir = Path(saving_dir)
    #     # save the passage embeddings
    #     index_path = saving_dir / self.INDEX_FILE_NAME
    #     logger.info(f"Saving passage embeddings to {index_path}")
    #     faiss.write_index(self.embeddings, str(index_path))
    #     # save the passage index
    #     documents_path = saving_dir / self.DOCUMENTS_FILE_NAME
    #     logger.info(f"Saving passage index to {documents_path}")
    #     self.documents.save(documents_path)

    # @classmethod
    # def load(
    #     cls,
    #     loading_dir: Union[str, os.PathLike],
    #     device: str = "cpu",
    #     document_file_name: Optional[str] = None,
    #     embedding_file_name: Optional[str] = None,
    #     index_file_name: Optional[str] = None,
    #     **kwargs,
    # ) -> "FaissDocumentIndex":
    #     loading_dir = Path(loading_dir)

    #     document_file_name = document_file_name or cls.DOCUMENTS_FILE_NAME
    #     embedding_file_name = embedding_file_name or cls.EMBEDDINGS_FILE_NAME
    #     index_file_name = index_file_name or cls.INDEX_FILE_NAME

    #     # load the documents
    #     documents_path = loading_dir / document_file_name

    #     if not documents_path.exists():
    #         raise ValueError(f"Document file `{documents_path}` does not exist.")
    #     logger.info(f"Loading documents from {documents_path}")
    #     documents = Labels.from_file(documents_path)

    #     index = None
    #     embeddings = None
    #     # try to load the index directly
    #     index_path = loading_dir / index_file_name
    #     if not index_path.exists():
    #         # try to load the embeddings
    #         embedding_path = loading_dir / embedding_file_name
    #         # run some checks
    #         if embedding_path.exists():
    #             logger.info(f"Loading embeddings from {embedding_path}")
    #             embeddings = torch.load(embedding_path, map_location="cpu")
    #         logger.warning(
    #             f"Index file `{index_path}` and embedding file `{embedding_path}` do not exist."
    #         )
    #     else:
    #         logger.info(f"Loading index from {index_path}")
    #         index = faiss.read_index(str(embedding_path))

    #     return cls(
    #         documents=documents,
    #         embeddings=embeddings,
    #         index=index,
    #         device=device,
    #         **kwargs,
    #     )

    def get_embeddings_from_index(
        self, index: int
    ) -> Union[torch.Tensor, numpy.ndarray]:
        """
        Get the document vector from the index.

        Args:
            index (`int`):
                The index of the document.

        Returns:
            `torch.Tensor`: The document vector.
        """
        if self.embeddings is None:
            raise ValueError(
                "The documents must be indexed before they can be retrieved."
            )
        if index >= self.embeddings.ntotal:
            raise ValueError(
                f"The index {index} is out of bounds. The maximum index is {self.embeddings.ntotal}."
            )
        return self.embeddings.reconstruct(index)