Loss: nan

[cisan86]

hello, I run your code with the default lr=0.02 but get Loss:nan , my batchsize=8, env macos10.15, python3.6,

[ivanwhaf]

maybe your learning rate is too big, better lower than 0.01

[cisan86]

thx, i try 0.01 and get the right loss. but the output file -- pth file is more than 900mb, is that anything wrong?

[ivanwhaf]

probably nothing wrong, see models/yolov1.py for more model-structure details