RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 1. Got 128 and 32 in dimension 2 at /opt/conda/conda-bld/pytorch_1549624239588/work/aten/src/TH/generic/THTensorMoreMath.cpp:1307

[hinash88]

Hi there,

Thank you for sharing the code. I am new to learn GAN and when I run this code I face this following issue:

Traceback (most recent call last): File "main.py", line 341, in <module> G_log, D_log = train(epoch_max, batch_size) File "main.py", line 316, in train D_loss = train_D(real_images, real_labels) File "main.py", line 218, in train_D real_outputs = D.forward(real_images,real_labels_fill) File "main.py", line 137, in forward x = torch.cat((x, y), dim=1) RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 1. Got 128 and 32 in dimension 2 at /opt/conda/conda-bld/pytorch_1549624239588/work/aten/src/TH/generic/THTensorMoreMath.cpp:1307

My batch size is 16 and i am running the code for 2 classes with 698 and 500 images respectively. My input images are 256x256 and I did not change any other parameter. These are the size of my x and y respectively where the problem occurs (main.py line 137)

torch.Size([16, 64, 128, 128]) torch.Size([16, 64, 32, 32])

Another problem is this code gives error when more than 2 classes are arranged in data folder.

Any kind of help is much appreciated!!!

[hinash88]

Got it solved. The input image must be 64x64 in size. Replace this code:

data_transform = transforms.Compose([transforms.ToTensor()])

with the following:
img_size = 64

data_transform = transforms.Compose([transforms.Resize(img_size), transforms.CenterCrop(img_size), transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5)),])

But still could not figure out how to deal with more than 2 classes :(