joncarter1/wav2sleep ? reverse-engineered prompt

Build me a Python project for sleep stage classification from physiological signals, based on the wav2sleep idea.

I want it to train and run inference on sleep data from EDF files, with support for different signal sets like ECG, PPG, respiratory signals, and EOG. The model should work even when not all modalities are present, and it should be able to handle both a cardio respiratory version and an EOG version. Please include scripts for preprocessing raw sleep study files into a faster format for training, plus a training workflow and a prediction workflow that can run on a folder of files and save the results somewhere useful.

It would be great if the project can load models from Hugging Face too, and if there’s a simple Python API for loading a model and predicting on a folder. If you need to, look up current docs online for any missing details, but keep the setup easy to run locally and cleanly organized.