Project to use ML and AI techniques to automatically chart songs to play in Clone Hero.

This project has been significantly updated to use a modern transformer-based architecture for drum transcription. Key features include:

• Audio Spectrogram Transformer (AST): Utilizes an 85M+ parameter model for improved accuracy.
• PyTorch Lightning: Employs a robust training pipeline with mixed precision support.
• Memory-Efficient Data Processing: Capable of handling large datasets.
• Multi-Environment Support: Configured for both local (M1-Max) and cloud (Google Colab) training.

For detailed information on model training, see model_training/README.md. For the refactoring plan and current progress, see TRANSFORMER_REFACTORING_PLAN.md.

Refer to model_training/README.md for data preparation and training commands.