vutils | Utilities for VueJS | Frontend Framework library

Your model's output is not consistent with your criterion.

If you want to keep the model and change the criterion:

Use BCELoss instead of CrossEntropyLoss. Note: You will need to cast your labels to float before passing them in. Also consider removing the Sigmoid() from the model and using BCEWithLogitsLoss.

If you want to keep the criterion and change the model:

CrossEntropyLoss expects the shape (..., num_classes). So for your 2 class case (real & fake), you will have to predict 2 values for each image in the batch which means you will need to alter the output channels of the last layer in your model. It also expects the raw logits, so you should remove the Sigmoid().

The issue is actually with your model architecture. You are trying to place a conv2d layer just after a linear fully connected layer.The _create_layer_1 produces a 1d output. You are trying to feed this 1d output to a conv2d layer which expects a multidimensional input.

From your code the best thing I feel to make it work in a single go would be to remove "_create_layer_2" function completely from generator class and use _create_layer_1 function to define all your layers(so that all layers are fully connected layers). Also do this for your discriminator

If you still need to use conv2d. You should reshape the input to conv2d to a 2d tensor. Also you have to flatten the 2d tensor to 1d before your final linear layer. Or you could ditch the first linear nn.Linear layer and start with conv2d altogether.

To summarise As you are designing GANs you might have experience developing CNNs. The point is you dont simply mixup conv2d/conv layers with linear layers without using proper flatten/reshape.

Cheers

It could be because after the criterion for netG obtains an output that was detached from the parameters of netG thus the optimizer is not / can't be updating the parameters for netG.

Adding