Upcoming breaking changes #117
ziatdinovmax
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I'm in the process of streamlining GPax to broaden its applicability and make it easier to build applications on top of it. Unfortunately, this will come with a few breaking changes.
Breaking change 1:
There will be no need to provide a random number generator key in training and prediction since it will be handled internally. So instead of
you will be using
The same is true for prediction.
Breaking change 2:
The prediction will return directly posterior variance instead of samples. To get samples, you can use a
sample_from_posterior
method.So instead of
you will be using:
To get posterior samples, use
I do not expect any other major changes to API.
Please feel free to comment and let me know your thoughts.
Beta Was this translation helpful? Give feedback.
All reactions