I created trainable Transformed Distribution with using Tensorflow Probability.
transformed_distribution = tfd.TransformedDistribution(
distribution=base_distribution, bijector=stacked_ffjord)
After training I can sample from that:
generated_samples = transformed_distribution.sample(SAMPLE_SIZE).numpy()
I'd like to save this distribution (in the file?) in order to restore it later and be able to sample from it any time I want.
I know how to save Models but what about Distributions?