What should I use for the parameter “resume” in train_ssd.py?

I trained a model for 30 epochs, and now I have all 30 files which names are the epoch and the validation loss. Are these files the kind of value I have to use in parameter “resume”? For example —resume=epoch29-loss-3.9472601746 ?

Thanks in advance

Hi @100375195, yes you would pick the one with the lowest loss. Include the full filename, e.g. --resume mb1-ssd-Epoch-19-Loss-0.7719583538174629.pth

You can also try --pretrained-ssd mb1-ssd-Epoch-19-Loss-0.7719583538174629.pth instead if that works better for you.

Thanks very much.

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.