Closed skodapetr closed 3 years ago
Automatic hyper parameter optimization wasn't fully implemented when the models were published.
Nevertheless, there are some continually updated tutorials on training and optimization here: https://github.com/rdk/p2rank/tree/develop/misc/tutorials
In order to utilize p2rank to it's fullest potential t hyper-parameter training is required. Would it be possible to make available a tutorial/log that was used to train published models? It may also include notes on evaluation etc .. ie. how to decide how good the parameters are.
Such a document can be used as a starting point for other learning and applications.