ggchen1997 / gdw-nips2021 Goto Github PK

I am reading your great work, but I am confused about the vnet in your code. Could you explain the function of vnet in your code？

I guess v_lambda[index, targets] = (torch.sum(v_lambda * p.data, dim=1) - v_lambda[index, targets] * p.data[index, targets]) / (1 - p.data[index, targets]) is the code for eq.4 in the paper.
But why v_lambda is the output of vnet in v_lambda = vnet(cost_v.data) ??

I'm looking forward to your reply. Thank you so much.

(If you can speak Chinese, could you reply me in Chinese? My English is not very good.)