2023-05-27 08:32 AM
Hi, I have a problem validating my classifier on xcube ai (both validation on target and validation on desktop). It shows error scores but accuracy is n.a. persistently. I have tried setting "--classifier" option in settings, restarting cubeIDE, nothing works. It shows that option --classifier is used during compilation but still, n.a. is showed under acc upon finishing validation.
My model is a binary classifer model with sigmoid function as output, producing either 0 or 1.
Does anyone know how to fix this issue? Accuracy score is crutial for me to see, as I want to compare the accuracy score on test set on quantized model in my thesis.
Also, is there any way to show binary crossentropy (or any crossentropy error) among the default scores in validation? It would be useful to see.