ask for help
22wei22 opened this issue · comments
22wei22 commented
when I train the generator of step 1. I find that the discriminator is powerful .the g_loss will be 1 and the d_loss will be 0. How could I do
ricvolpi commented
Did you observe this behavior running Step 0 and Step 1 out of the box, or did you modify something?
22wei22 commented
Yeah I implement it in pytorch. Now I can not converge it. In step 0 ,the accuracy of pretrained model is 0.87. Is it right?
ricvolpi commented
No, the accuracy of the model trained at Step 0 should be much lower (check the paper - Table 2, first row).