I have a dataset with batch input of shape (10,6) where 6 is the number of features. I want to have a kernel is size (3×6) which can slide only in the downward direction. So I can convolve along the rows which is my time stamp data. In this case, is it better to use a 1DCNN with num_features=6? And use a kernel size of 3?

I have tried implementing 1DCNN, but I am not sure how the data is arranged for multiple features. My output of 1DCNN does not contain my feature length but I want that axis to be consistent.