dataset_lib/encode.py

import logging
import os

import pandas as pd
import torch
from torch.utils.data import DataLoader
from tqdm import tqdm

import dataset_lib.multimodal as multimodal
from dataset_lib.config import Config
from dataset_lib.config import Constants as c
from dataset_lib.datasets import get_dataset

logger = logging.getLogger(__name__)


@torch.no_grad()
def encode(config: Config, device=c.DEVICE, workdir=c.WORKDIR):
    logger.info(
        f"Encoding dataset {config.data.dataset.lower()} with"
        f" backbone = {config.data.backbone}"
    )

    datasets = get_dataset(config.data.dataset)
    encode_image = multimodal.get_image_encoder(config, device=device)

    for op, dataset in datasets.items():
        data = {"embedding": [], "label": []}
        for image, label in tqdm(dataset, desc=f"Encoding {op}"):
            embedding = encode_image(image).float()
            embedding /= torch.linalg.norm(embedding, dim=-1, keepdim=True)
            embedding = embedding.cpu().numpy()

            data["embedding"].extend(embedding)
            data["label"].append(label)

        df = pd.DataFrame(data)

        data_path = os.path.join(
            f"{config.data.dataset.lower()}_{op}_{config.backbone_name()}.parquet"
        )
        df.to_parquet(data_path, index=False)