Open danijar opened 4 years ago
FWIW I love this suggestion. It's always bugged me as well, and I never thought through it enough to see this obvious-in-hindsight solution. I'd invite a PR, and otherwise will try to do this soon.
Is there some reason why we cannot simply
import tensorflow_probability.distributions
from tensorflow_probability.distributions import Normal
?
I usually first try this import that fails:
Then I try the second one that works
Is there a reason against including
distributions
in the parent package such that both imports work?