Varitational Autoencoder

I am trying to build a Variational Autoencoder from Scratch and stuck at a problem with dimensions.

Below Architecture:

Variational Autoencoder Class:


I am not sure why decoder is expecting (64, 64, 3, 3) ?
Help me out :slight_smile:

Hy @mayurat22 Can you post the code of your training section also.

Can you try reshaping with batch size = 1? Like
x = x.reshape((1,64,7,7))

1 Like


It worked. Thank you.

1 Like