Stacking dense layer outputs before passing them through convolutional layers in PyTorch

Question

I\'m currently working with time-series data and have trained a simple dense network to perform a regression task taking in a single input sample at a time. I would now like