Model evaluation metrics discrepancy between values showed graphically and values described in file .json

hello,

I found some different values between the downloaded ‘detailed model metrics’ json file and the values showed by the online report. Why this differences? What values should I consider as the final metrics of my model?

thank you in advance

Hi @chiaragironi

There are some differences when hardware is selected for the target deployment vs the actual deployment dependant on the emulated hardware that is used for testing, but you could have also taken shots between training sessions?

Can you share some more detail and screenshots of the timing of the testing and JSON? Thanks!

Best

Eoin