half precision returning nan

frotms / PaddleOCR2Pytorch

PaddleOCR inference in PyTorch. Converted from [PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR)

Apache License 2.0

809 stars 159 forks source link

half precision returning nan #54

Open ririya opened 1 year ago

ririya commented 1 year ago

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?