stanfordmlgroup / ngboost

Natural Gradient Boosting for Probabilistic Prediction
Apache License 2.0
1.64k stars 215 forks source link

Adding pickle to test? #238

Closed MikeOMa closed 3 years ago

MikeOMa commented 3 years ago

is it worth adding:

serial = pickle.dumps(ngb)
ngb = pickle.loads(serial)
ngb.predict(X_test)

To test_distns.py for most of the tests? I know I use pickle all the time when I'm using ngboost/sklearn so ensuring that everything can pickle seems sensible to me.

Happy to do this if nobody else was already planning to and you think it is a good idea.

ryan-wolbeck commented 3 years ago

@MikeOMa I started going down this road a while ago here https://github.com/stanfordmlgroup/ngboost/compare/test-pickle but didn't get finished with it. If you have time your welcome to pick up where I left off. Generally think this is something that needs to get done

MikeOMa commented 3 years ago

I did see this branch after starting the issue, I was going to try to fix the Survival Pickling when I have a chance so I can add it in with that pull request. Probably won't be for a week or two though.