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
Currently, a number of methods use torch.float32 as their default dtype if no explicit choice is provided. While this is often a reasonable choice, some applications require more (or less) precision, which the current default inhibits. Instead, the methods should query torch.get_default_dtype() such that the user can globally configure the desired level of precision.
The text was updated successfully, but these errors were encountered:
Currently, a number of methods use
torch.float32
as their defaultdtype
if no explicit choice is provided. While this is often a reasonable choice, some applications require more (or less) precision, which the current default inhibits. Instead, the methods should querytorch.get_default_dtype()
such that the user can globally configure the desired level of precision.The text was updated successfully, but these errors were encountered: