大佬你好，训练的时候出现这个问题了，这是啥情况啊

[wangshicr7]

loss_semi_adv_value += loss_semi_adv.data.cpu().numpy()[0]/args.lambda_semi_adv
IndexError: too many indices for array

[hfslyc]

This repo was using Pytorch 0.2, which has a different behavior than recent version on single scalar. You could remove the [0] and it should work.

[gxwangupc]

loss_semi_adv_value += loss_semi_adv.data.cpu().numpy()[0]/args.lambda_semi_adv
IndexError: too many indices for array

hahhahahha