Yard1 / ray-skorch

Distributed skorch on Ray Train
Apache License 2.0
57 stars 3 forks source link

Implement `predict_proba`, add `ddp_kwargs`, improve checkpointing #11

Closed Yard1 closed 2 years ago

Yard1 commented 2 years ago

Implements predict_proba, returning a ray.data.Dataset. Refactors generation of training functions. Adds ddp_kwargs argument. Improves checkpointing and adds an example.