You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
#The distributions.utils.broadcast_all() is used in all distribution constructors to clean up parameters. This is expensive. It would be nice to make this cheaper by reimplementing in C/C++.
Adam suggested using torch._C._infer_size() under the hood but we would need to add more support for scalars.
It would also be good to implement Distribution._validate_log_prob_arg() in C/C++.
The text was updated successfully, but these errors were encountered:
#The
distributions.utils.broadcast_all()
is used in all distribution constructors to clean up parameters. This is expensive. It would be nice to make this cheaper by reimplementing in C/C++.Adam suggested using
torch._C._infer_size()
under the hood but we would need to add more support for scalars.It would also be good to implement
Distribution._validate_log_prob_arg()
in C/C++.The text was updated successfully, but these errors were encountered: