This library would form a permanent home for reusable components for deep probabilistic programming. The library would form and harness a community of users and contributors by focusing initially on complete infra and documentation for how to use and create components.
One awkwardness of the previous API is that the user has to ensure they instantiate a Bijector using a keyword argument for params. I.e. there will be an error if the user writes:
Motivation
One awkwardness of the previous API is that the user has to ensure they instantiate a Bijector using a keyword argument for
params. I.e. there will be an error if the user writes:bijs.AffineAutoregressive(params.DenseAutoregressive())rather than
bijs.AffineAutoregressive(params=params.DenseAutoregressive()).Changes proposed
I propose to fix this and reduce potential user errors by making
paramsthe first and non-keyword argument of Bijector.There are a few analogous changes around the code to support this.