A question about loss_G.backward(Mone)

[EliotChenKJ]

First, i have to thank for your sharing of this work, it does help me understanding the wgan-gp better, but here is a little question:

• what does the Mone in the loss_G.backward() mean?

l've just got to know wgan-gp recently, so hope the author can help me with this question.

[jalola]

Hi, Thanks for asking, basically mone is minus one tensor. So in the loss, this is the same to: gen_cost = -gen_cost gen_cost.backward()

You can check this issue: https://github.com/jalola/improved-wgan-pytorch/issues/1

[EliotChenKJ]

Ok, now i understand its meaning. Thank for ur answer~