LSTM implementation

I am new with deep learning especially LSTM, I have time-series data for 62 days of 63 station and each station has divided into 8 parts in each day. it means I have 62 * 63 * 8 columns and the period of data is weekly for example every Sunday the station number 4 in part 1 of the day has a similar act. my question is how can set the LSTM parameters like input_size, seq_length, batch_size, n_layers,…
I should note that I don’t have all stations in all days for example in some days I have data from 58 stations and I fill my table with zero for the rest of the stations
thanks for your quick response!