half precision returning nan

Open ririya opened this issue 3 years ago • 0 comments

I'm trying to run inference on the trained models with half precision. The output for the detection model is returning all nan's. Anyone tried to do this before?

Aug 30 '22 21:08 ririya