Hello I tried to ask this question in the Forum, but somehow the Forum won't let me register saying something about timeout even if I refresh the page for registering. So instead I post this here as feature request.

What I want to achieve is that say for a 3x8x8 input for a dense layer I want the same 8x8 weights to be trained and repeated over the whole input. Is there a condition I can set for the training to repeat the weights multiple times so I train 8x8 weights over a Nx8x8 input instead of training Nx8x8 individual weights over the input?