SCRAPL DDSP

Scattering with Random Paths as Loss for Differentiable Digital Signal Processing

---

Instructions for Reproducibility

1. Clone this repository and open its directory.
2. Install the requirements:
   conda env create --file=conda_env_gpu.yml
   or
   pip install uv
uv pip install -r requirements_gpu.txt
   For posterity, requirements_all_gpu.txt and requirements_all_cpu.txt are also provided.
3. The source code can be explored in the experiments/, scrapl/, and eval_808/ directories.
4. All experiment config files can be found in the configs/ directory.
5. The dataset for the Roland TR-808 sound matching task can be found here.
6. Create an out directory (mkdir out).
7. All experiments can be run by modifying scripts/train.py and the corresponding configs/.../train_ ... .yml config file and then running python scripts/train.py.
   Make sure your PYTHONPATH has been set correctly by running commands like:
   export PYTHONPATH=$PYTHONPATH:BASE_DIR/scrapl/,
   export PYTHONPATH=$PYTHONPATH:BASE_DIR/scrapl/kymatio/,
   and export PYTHONPATH=$PYTHONPATH:BASE_DIR/scrapl/scrapl/.
8. The source code is currently not documented, but don't hesitate to open an issue if you have any questions or comments.
9. A pip installable Python package of SCRAPL for the JTFS is coming soon.