Advice on extending to PyTorch

[jinglescode]

Any suggestions on how to use existing code to work on PyTorch? Because from the examples, evaluation.process depends on sklearn.pipeline, and thus it executes models from sklearn such as sklearn.linear_model import LogisticRegression. Any advice?

[ErikBjare]

I'm interested in this too, looks like one can use something like skorch: https://github.com/skorch-dev/skorch

[jinglescode]

I've never knew about skorch, and it looks good. Let me try training a simple PyTorch network with moabb's evaluation.process.