Pixel Time Series

I have a time series of pixels (from satellite data). They are represented in an HDF5 file and I can read them using pandas such that each row corresponds to a pixel and each column to a time step. So each cell contains an array of values where each value corresponds to the value of one of the satellite bands. It is the first time, I am working with time series data in Pytorch, so I am not sure how to feed them into a (sequential model). When calling the DataLoader, what should one sample represent? One specific time step of all pixels or just a time step of one pixel?