Is it possible to sample from a trained model?

[UnixJunkie]

Hello,

From reading the code, I guess that for a trained model, a given
input sequence would produce a single output SMILES and that
would be reproducible (i.e. always the same).

Is my understanding correct?

Users would probably be interested in creating 10s or hundreds of SMILES
for a given input sequence given to a trained model.

Regards,
F.