There is nowhere in the documentation about how the amp implementation works with the complex tensors added in the recent major pytorch update. Does anyone here know by any chance?
The only thing that I’ve been able to find up to this point is some type error messages on github issues.