scripts/download_assets.py

#!/usr/bin/env python3
import argparse
from pathlib import Path

from huggingface_hub import snapshot_download


def main():
    parser = argparse.ArgumentParser(description="Download PixDLM model assets and DRSeg metadata.")
    parser.add_argument("--model-repo", default="WhynotHug/PixDLM")
    parser.add_argument("--dataset-repo", default="WhynotHug/DRSeg")
    parser.add_argument("--output-dir", default=".")
    parser.add_argument("--with-data", action="store_true", help="Download full DRSeg data files when hosted in the dataset repo.")
    args = parser.parse_args()

    root = Path(args.output_dir).resolve()
    checkpoint_dir = root / "pretrained" / "pixdlm-7b"
    data_dir = root / "data"
    checkpoint_dir.mkdir(parents=True, exist_ok=True)
    data_dir.mkdir(parents=True, exist_ok=True)

    snapshot_download(
        repo_id=args.model_repo,
        repo_type="model",
        local_dir=checkpoint_dir,
        allow_patterns=[
            "added_tokens.json",
            "config.json",
            "generation_config.json",
            "model*.safetensors",
            "model.safetensors.index.json",
            "pytorch_model*.bin",
            "pytorch_model.bin.index.json",
            "special_tokens_map.json",
            "tokenizer.json",
            "tokenizer.model",
            "tokenizer_config.json",
        ],
    )

    dataset_ignore = None if args.with_data else ["*.jpg", "*.png", "*.zip", "*.tar", "*.tar.gz"]
    snapshot_download(
        repo_id=args.dataset_repo,
        repo_type="dataset",
        local_dir=data_dir / "DRSeg",
        ignore_patterns=dataset_ignore,
    )

    print(f"PixDLM model snapshot: {checkpoint_dir}")
    print(f"DRSeg files: {data_dir / 'DRSeg'}")


if __name__ == "__main__":
    main()