I am trying to convert a fairly large model with an EfficientDet backbone to FP16. Conversion succeeds, but when running inference I get NaN output values for certain inputs. I managed to reproduce ...
EMBED <iframe src="https://archive.org/embed/microsoft-office-2021-ltsc-16.0.15726.20188" width="560" height="384" frameborder="0" webkitallowfullscreen="true ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback