rsvalerioo/models

Convert models from source formats (PyTorch/ONNX) into optimized inference formats and publish them to Hugging Face Hub.

Structure

Each subdirectory under models/ is a self-contained conversion recipe:

models/
└── bge-base-en-v1.5-coreml/
    ├── convert.py         # Conversion script
    ├── config.yaml        # Source model, target format, HF repo
    └── README_TEMPLATE.md # Model card for HF

Adding a new model

1. Create models/<name>/config.yaml with source model and target format
2. Write models/<name>/convert.py to produce the converted artifacts

Converting

cd models/bge-base-en-v1.5-coreml
uv run python convert.py

Publishing to Hugging Face

Requires a HF_TOKEN environment variable with write access to the target repo.

HF_TOKEN=hf_... uv run python scripts/publish.py \
  --repo rsvalerio/bge-base-en-v1.5-coreml \
  --artifacts-dir models/bge-base-en-v1.5-coreml \
  --config models/bge-base-en-v1.5-coreml/config.yaml

Requirements

• uv
• HF_TOKEN with write access to target HF repos (for publishing)