Jian Shi

I will take look.

I think there will be two steps of optimisation here: 1. generate random parameters under `batchsize==1`. 2. apply the transformation efficiently with `reshape(1, N*C, H, W)`, etc. The general idea...

Is there any references for such operation?

> @shijianjian yes the formula is correct. > Adding the option to pass a background color would be pretty handy, since it is often used. > Making me wonder if...

Hi @zzyunzhi, any updates on this?

> As overall idea, what’s the main difference between v1 and v2 ? `V1` forces the input as `forward(input, label)` which is pretty strict to use. `v2` will follow the...

> @shijianjian any update here ? Did not notice there are bugs there... I will finish this weekend.

@edgarriba Finally, the annoying bug has been fixed. XD. Safe to merge I suppose.

Sure thing. I would prepare one.

This is a nn module wrapper to include all the modules that can be used as neural network layers. This is a more or less AOP idea to collect and...