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
We noticed that einsum is not added to the supported ops list for low precision policy in autocast, is there a reason for that? Does this op have some issues in the support?
The text was updated successfully, but these errors were encountered:
In order to support einsum, would it require more extensive changes or just a change to torch_xla/csrc/autocast_mode.cpp that adds KERNEL_XLA(einsum, lower_precision_fp)?
We noticed that einsum is not added to the supported ops list for low precision policy in autocast, is there a reason for that? Does this op have some issues in the support?
The text was updated successfully, but these errors were encountered: