Sometimes, you don't want the shape of your convolutional outputs to reduce in size. Other times, you wish to append zeroes to the inputs of your Conv1D layers. Padding - same/zero padding and causal ...