The number of input and output is inconsistent when using my own data for prediction

[rookiecoder-chen]

I try to use megnet algorithm for QSAR task, use 5000 data for training, 800 data for prediction, but only output 600 results, and my true label has also been modified. What's the reason?

preprocessor = MEGNetPreprocessor(max_atoms=-1, add_Hs=True, use_all_feature=False,
atom_list=['C', 'H', 'N', 'O'],include_unknown_atom=False, kekulize=False,
max_num_nbr=12, max_radius=8, expand_dim=100)
What do these three parameters （max_num_nbr,max_radius,expand_dim）mean? It doesn't seem to be explained in the notes. What's the impact?