src/upload_to_hub.py

#!/usr/bin/env python3
"""Upload Bamboo-1 model to Hugging Face Hub."""

import click
from pathlib import Path
from huggingface_hub import HfApi, create_repo


MODEL_CARD = '''---
language:
- vi
library_name: bamboo1
tags:
- dependency-parsing
- vietnamese
- nlp
license: apache-2.0
datasets:
- undertheseanlp/UDD-1
---

# Bamboo-1: Vietnamese Dependency Parser

A Vietnamese dependency parser using XLM-RoBERTa + Biaffine architecture, trained on the UDD-1 dataset.

## Installation

```bash
pip install bamboo1
# or
pip install git+https://github.com/undertheseanlp/bamboo-1.git
```

## Usage

```python
from bamboo1 import load, parse

# Quick parse (auto-downloads model from HF Hub)
sent = parse("Tôi yêu Việt Nam")
for token in sent:
    head = sent.get_head(token)
    print(f"{token.form} -> {head.form if head else 'ROOT'} ({token.deprel})")

# Or load explicitly
parser = load("undertheseanlp/bamboo-1")
sent = parser.parse("Hà Nội là thủ đô của Việt Nam")

# Batch parsing
sentences = parser.parse_batch([
    "Tôi yêu Việt Nam",
    "Em gái tôi học tiếng Anh",
])

# Export to CoNLL-U format
print(sent.to_conllu())
```

## Model Details

- **Architecture**: XLM-RoBERTa (xlm-roberta-base) + Biaffine attention
- **Training data**: UDD-1 (Vietnamese Universal Dependencies)
- **Framework**: PyTorch

## Performance

| Metric | Score |
|--------|-------|
| UAS    | TBD   |
| LAS    | TBD   |

## Citation

```bibtex
@misc{bamboo1,
  title={Bamboo-1: Vietnamese Dependency Parser},
  author={Underthesea NLP},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/undertheseanlp/bamboo-1}
}
```
'''


@click.command()
@click.option('--model-dir', '-m', default='models/bamboo-1', help='Local model directory')
@click.option('--repo-id', '-r', default='undertheseanlp/bamboo-1-model', help='HF repo ID')
@click.option('--private', is_flag=True, help='Make repo private')
def upload(model_dir: str, repo_id: str, private: bool):
    """Upload Bamboo-1 model to Hugging Face Hub."""
    model_path = Path(model_dir)
    model_file = model_path / 'model.pt'

    if not model_file.exists():
        raise click.ClickException(f"Model file not found: {model_file}")

    click.echo(f"Uploading {model_file} to {repo_id}...")

    api = HfApi()

    # Create repo if it doesn't exist
    try:
        create_repo(repo_id, repo_type="model", private=private, exist_ok=True)
        click.echo(f"Repository {repo_id} ready")
    except Exception as e:
        click.echo(f"Note: {e}")

    # Upload model file
    click.echo("Uploading model.pt...")
    api.upload_file(
        path_or_fileobj=str(model_file),
        path_in_repo="model.pt",
        repo_id=repo_id,
        repo_type="model",
    )

    # Upload README
    click.echo("Uploading README.md...")
    api.upload_file(
        path_or_fileobj=MODEL_CARD.encode(),
        path_in_repo="README.md",
        repo_id=repo_id,
        repo_type="model",
    )

    click.echo(f"\nModel uploaded successfully!")
    click.echo(f"View at: https://huggingface.co/{repo_id}")


if __name__ == '__main__':
    upload()