Error in dlfeval for Variational Autoencoders

I am trying to implement the VAE on my data. I found the following link:

https://www.mathworks.com/help/deeplearning/examples/train-a-variational-autoencoder-vae-to-generate-images.html#GeneratingHanddrawnDigitsUsingAVariationalAutoencoderVAEExample-13

In this example the number of training images from the MNIST dataset is 60,000 images for training and 10,000 images for testing, the imageSize = [28 28 1];

My dataset is 8522 images for training, and 2135 images for testing, the imageSize = [227 227 3];

My 1st question is, does this code work only for gray scale images? if yes, is there a way to use VAE for RGB images?

2nd question is: How can I solve this error from implementing [infGrad, genGrad] = dlfeval(@modelGradients, encoderNet, decoderNet, XBatch); do I need to change the encoderNet, or decoderNet for my data?

error description:

Error using -

Matrix dimensions must agree.

Error in deep.internal.recording.operations.MinusBroadcastOp/forward (line 32)

x = x - y;

Error in - (line 39)

xdata = xdata - ydata;

Error in ELBOloss (line 2)

squares = 0.5*(xPred-x).^2;

Error in modelGradients (line 4)

loss = ELBOloss(x, xPred, zMean, zLogvar);

Error in deep.internal.dlfeval (line 18)

[varargout{1:nout}] = fun(x{:});

Error in dlfeval (line 40)

[varargout{1:nout}] = deep.internal.dlfeval(fun,varargin{:});

Error in Test_VAE_mydata (line 67)

[infGrad, genGrad] = dlfeval(...

[Nick]

Have you adjusted the image pre-processor and the layers to match your new data format ?