We uploaded a priliminary version of our code. One may try to train our model with the uploaded code. In this version, only shared tokens and detail tokens are implemented (where intermediate tokens can be easily implemented).
Command for training:
python train.py hiv {save_path} 5000 {number_of_molecules}