Fallback to CPU for dot implementation for long tensors

Open bhavya01 opened this issue 1 year ago • 1 comments

Fall back to cpu implementation for dot product if both the tensor are int64. #6700

Mar 09 '24 01:03 bhavya01

Is this the only operation missing support for long dtype? I assume this is TPU specific?

Mar 12 '24 00:03 miladm

Is this the only operation missing support for long dtype? I assume this is TPU specific?

On more testing, I found when one of the tensors is a long and other is any integral type, it still doesn't work. Change the implementation to reflect this. Yes, this is only for TPU.

Mar 13 '24 19:03 bhavya01