Basic Python interface class to handle pickle objects in hdf5 files, including PyTorch tensors, enabling also compression.
This is achieved by converting Pickle object as byte stream into NumPy byte arrays. By default the /picked prefix is appended to pickle object to distinguish them in the base hdf5 file.
Just run pickled_hdf5.py for a minimal usage example.