src/deploy/huggingface/download_model.py

"""
Download exported model from Hugging Face Hub before API startup.

Set Space secret / env:
  HF_MODEL_REPO=your-username/deberta-complexity-lcp
  MODEL_PATH=/app/model
"""

from __future__ import annotations

import os
import sys
from pathlib import Path


def main() -> None:
    repo = os.environ.get("HF_MODEL_REPO", "").strip()
    target = Path(os.environ.get("MODEL_PATH", "/home/user/model"))

    if not repo:
        print("HF_MODEL_REPO not set — expecting model_weights.pt already on disk.")
        weights = target / "model_weights.pt"
        if weights.exists():
            print(f"Found {weights}")
        else:
            print(f"WARN: no weights at {weights}", file=sys.stderr)
        return

    weights = target / "model_weights.pt"
    if weights.exists():
        print(f"Model already present at {weights}")
        return

    print(f"Downloading {repo} → {target}")
    from huggingface_hub import snapshot_download

    token = os.environ.get("HF_TOKEN") or os.environ.get("HUGGING_FACE_HUB_TOKEN")

    target.mkdir(parents=True, exist_ok=True)
    snapshot_download(
        repo_id=repo,
        local_dir=str(target),
        token=token,
        allow_patterns=[
            "model_weights.pt",
            "config.json",
            "tokenizer/*",
            "*.json",
        ],
    )

    if not weights.exists():
        raise FileNotFoundError(
            f"Download finished but {weights} missing. "
            f"Upload model with: python deploy/huggingface/upload_model_to_hub.py"
        )
    print(f"Ready: {weights} ({weights.stat().st_size / 1e6:.1f} MB)")


if __name__ == "__main__":
    main()