I am getting following assertion error for discriminator, I have not changed any of the parameters and while calling the ConvTranspose2d or batchnorm I am just passing the parameters directly received in function call, not sure why I am getting this error.
The assertion error is because of the wrong shape of the hidden_output. If you have implemented the correct layers and passed the correct parameters then it wouldn’t be happening. From your query, I get to know that you are using the ConvTranspose2d layer in the Discriminator block which is wrong. The discriminator is a normal CNN involving Conv2d layers where you have image inputs. We are not trying to generate an image from noise using ConvTranspose2d like in the case of the generator.
Have a great day.
May I get some help with another assertion error please, if lines 11 and 12 are commented seems like theres no problem.
Welcome to the community
Hope you are doing well. Sorry for the late response.
Your hidden output seems fine, i.e. but the final output is wrong. There are two blocks in the test cells → test_hidden_block and test_final_block. Both call the make_gen_block of the Generator but the difference is that test_final_block sets the final_layer parameter as
True . This in turn goes to the else part of make_gen_block. Thus, you might be doing some mistakes in the
else part of make_gen_block. Please, check if you have used the correct layer type and activation function in the