Continuous bernoulli distribution (take 2) (#34619)
Summary:
We recently had a NeurIPS paper (https://arxiv.org/abs/1907.06845 and https://papers.nips.cc/paper/9484-the-continuous-bernoulli-fixing-a-pervasive-error-in-variational-autoencoders) where we introduce a new [0,1]-supported distribution: the continuous Bernoulli. This pull request implements this distribution in pytorch.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/34619
Differential Revision: D20403123
Pulled By: ngimel
fbshipit-source-id: d807c7d0d372c6daf6cb6ef09df178bc7491abb2
