scripts/upload_to_hf.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
upload_to_hf.py — Script de sauvegarde et publication des données, datasets et graphiques sur Hugging Face (Dataset Hub).

Ce script permet d'uploader les données brutes, les datasets de caractéristiques extraites,
les rapports d'évaluation, les graphiques générés (PNG/HTML) et le code du pipeline
sur le Hugging Face Dataset Hub. Il permet de reprendre l'entraînement ultérieurement
ou de partager les résultats de l'étude.

Usage:
  python scripts/upload_to_hf.py --repo_id "votre-username/nom-du-dataset" --token "HF_TOKEN"
"""

import os
import sys
import argparse
import yaml
from huggingface_hub import HfApi

def load_config(config_path="configs/config.yaml"):
    if os.path.exists(config_path):
        with open(config_path, "r", encoding="utf-8") as f:
            return yaml.safe_load(f)
    return {}

def main():
    parser = argparse.ArgumentParser(description="Upload datasets, plots, reports, and code to Hugging Face Datasets.")
    parser.add_argument("--repo_id", required=True, help="HF Dataset Repository ID (e.g., 'username/detecteur-ia-parlement-dataset')")
    parser.add_argument("--token", help="Hugging Face Write Token (or set HF_TOKEN env var)")
    parser.add_argument("--repo_type", default="dataset", choices=["dataset", "space"], help="HF Repository Type")
    parser.add_argument("--space_sdk", default="gradio", help="SDK if uploading as a Space (defaults to gradio)")
    parser.add_argument("--exclude_models", type=bool, default=True, help="Exclude binary model checkpoints (.pkl files in models/)")
    parser.add_argument("--config", default="configs/config.yaml", help="Path to config file")
    args = parser.parse_args()
    
    repo_id = args.repo_id.strip() if args.repo_id else ""
    token = args.token or os.environ.get("HF_TOKEN")
    if token:
        token = token.strip()
    else:
        print("Error: Hugging Face API token is required. Use --token or set the HF_TOKEN environment variable.")
        sys.exit(1)
        
    api = HfApi(token=token)
    
    # 1. Create repo if it doesn't exist
    print(f"Checking/Creating Hugging Face repository '{repo_id}' (Type: {args.repo_type})...")
    try:
        api.create_repo(
            repo_id=repo_id,
            repo_type=args.repo_type,
            space_sdk=args.space_sdk if args.repo_type == "space" else None,
            exist_ok=True
        )
        print("Repository is ready.")
    except Exception as e:
        print(f"Error creating/retrieving repository: {e}")
        sys.exit(1)
        
    # 2. Setup Upload Parameters
    workspace_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
    print(f"Uploading files from local workspace '{workspace_dir}' to HF repo...")
    
    # Define ignore list to filter files
    ignore_patterns = [
        "**/.git/**",
        "**/__pycache__/**",
        "**/*.pyc",
        "**/.ipynb_checkpoints/**",
        "**/.gradio/**",
        "models_upload_temp/**",
        "hf_model_upload_*/**"
    ]
    
    if args.exclude_models:
        print("Excluding heavy binary models (.pkl files in models/) to keep dataset repo clean...")
        ignore_patterns.append("models/**/*.pkl")
        ignore_patterns.append("models/*.pkl")
    else:
        print("Including binary models in the upload...")
        
    # 3. Add custom dataset card YAML frontmatter if uploading to dataset
    # We will upload the workspace as is. We can write a specific dataset card header if needed,
    # but the README.md is already excellent.
    
    try:
        api.upload_folder(
            folder_path=workspace_dir,
            repo_id=repo_id,
            repo_type=args.repo_type,
            ignore_patterns=ignore_patterns
        )
        print("\n🎉 Success! All data files, datasets, reports, plots, and scripts have been uploaded to Hugging Face.")
        
        if args.repo_type == "dataset":
            print(f"Your dataset repository is live at: https://huggingface.co/datasets/{repo_id}")
            print("\nTo resume training or run the pipeline elsewhere, you can clone this dataset repository:")
            print(f"  git clone https://huggingface.co/datasets/{repo_id}")
        elif args.repo_type == "space":
            print(f"Your application space is live at: https://huggingface.co/spaces/{repo_id}")
            
    except Exception as e:
        print(f"Error uploading files: {e}")
        sys.exit(1)

if __name__ == "__main__":
    main()