Training with 96x96 input

[rezendegabriel]

Hello,

I am trying to adapt the code to run 96x96 inputs, more specifically the BigGAN.py script, by adding arch[96] into the functions def G_arch and def D_arch. However, I get the following error when compiling D_input = torch.cat([img for img in [G_z, x] if img is not None], 0) line in the def G_D function:

RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 0. Got 96 and 64 in dimension 2 at /pytorch/aten/src/THC/generic/THCTensorMath.cu:71

What I could investigate so far are the G_z and x tensors. They have different dimensions, torch.Size([50, 3, 64, 64]) and torch.Size([50, 3, 96, 96]), respectively. Therefore, concatenation cannot be done. I think that G_z tensor should also be in the form of x tensor. I would appreciate any help.

arch[96] = {'in_channels':  [int(ch * item) for item in [16, 16, 8, 4]],
            'out_channels': [int(ch * item) for item in [16,  8, 4, 2]],
            'upsample': [True] * 4,
            'resolution': [12, 24, 48, 96],
            'attention': {2**i + 2**(i-1): (2**i + 2**(i-1) in [int(item) for item in attention.split('_')]) for i in range(3, 7)}}
.
.
.
arch[96] = {'in_channels':  [3] + [int(ch*item) for item in [1, 2, 4, 8]],
            'out_channels': [int(item * ch) for item in [1, 2, 4, 8, 16]],
            'downsample': [True] * 4 + [False],
            'resolution': [48, 24, 12, 6, 6],
            'attention': {2**i + 2**(i-1): 2**i + 2**(i-1) in [int(item) for item in attention.split('_')] for i in range(2, 7)}}

[rezendegabriel]

I solved the problem by changing the bottom_width variable's default to 6 in the Generator class.