Reading database - pytorch
As explain at #317 (closed), the actual strategy to read the input data base is each batch are querying what it needs. It can be a drawback when the pieces of data to be read are small (small batch size).
We have to investigate how to load more that one batch at a time in order to open the database less often.