search

gasteigerjo / dimenet

DimeNet and DimeNet++ models, as proposed in "Directional Message Passing for Molecular Graphs" (ICLR 2020) and "Fast and Uncertainty-Aware Directional Message Passing for Non-Equilibrium Molecules" (NeurIPS-W 2020)
https://www.daml.in.tum.de/dimenet
Other
292 stars 61 forks source link

Training time #12

Closed xnuohz closed 3 years ago

xnuohz commented 3 years ago

Thank you for the great work. So, how long did you train a single task model? And what early stopping value did you choose?

gasteigerjo commented 3 years ago

As specified in the config, we trained for 3M steps on QM9, with no early stopping. The only early stopping-like mechanism we use is choosing the model parameters with the best performance on the validation set.