improved-diffusion F16 training error

F16 training error

Open 3bobo opened this issue 1 year ago • 2 comments

when set fp16=True, there is an error about _unflatten_dense_tensors:

TypeError: unflatten_dense_tensors(): argument 'tensors' (position 2) must be tuple of Tensors, not generator

Is this due to the version issue with the torch？

ENV: torch-2.0.1 cuda-11.8

Jun 11 '23 08:06 3bobo

I am having the same trouble. Did anyone resolve it?

Nov 13 '23 07:11 nipunagarwala

@3bobo @nipunagarwala please see my PR #129 for the fix.

Cheers and happy training

Mar 11 '24 22:03 NoahSchiro