Train with lower precision

[mhusseinsh]

Hello,
I would like to ask if there is a way to train the UNIT model with a lower precision? just to to give it a try if I can result in reducing training/inference time

Which part do I need to edit in the code?

[mingyuliutw]

The updated UNIT implementation is available at

https://github.com/NVlabs/imaginaire

We have low precision training.

Support should be better there.