Keras says dimensions aren't equal, but I'm sure they are. Works when I set batch size to 1

Question

So I\'m trying to get a convolutional variational autoencoder to work on MNIST. It was working fine with I didn\'t use convolution layers, and had everything with Dense laye