graph with global features, and mydataset

[teltim]

Sorry, this is not an issue report, but I'd like to ask questions to an expert

In molecular generation on the tutorial, there is a split between representation learning and reinforcement learning,

1. In the representation learning phase, is it possible to train the graph with global features to improve the final generation capability? Do any of the models currently implemented in torch-drug use global features for training? Please let us know if any models use global information for training.

2. Is it possible to train using the zinc2m dataset for representation training and then have the molecules generated using my own dataset during the reinforcement learning phase? (My dataset is small, but I want to use molecular generation to find molecules with small target variables)