PyTorch-CIFAR-10-autoencoder
PyTorch-CIFAR-10-autoencoder copied to clipboard
chenjie
Pretrained model does not match the model definition
The pre-trained model weights could not be loaded to the current model definition. I changed the model to
self.encoder = nn.Sequential(
nn.Conv2d(3, 32, 4, stride=2, padding=1), # [batch, 12, 16, 16]
nn.ReLU(),
nn.Conv2d(32, 48, 3, stride=2, padding=1), # [batch, 24, 8, 8]
nn.ReLU(),
)
self.decoder = nn.Sequential(
nn.ConvTranspose2d(48, 32, 3, stride=2, padding=1), # [batch, 24, 8, 8]
nn.ReLU(),
nn.ConvTranspose2d(32, 3, 4, stride=2, padding=1), # [batch, 12, 16, 16]
nn.ReLU(),
)
Now the code works, but the reconstruction is quite bad.
Do you have a different pre-trained model that can be used? Thanks!
I have the same question too!It seems that the author didn't provides the right weights?
