Use pre-split data in UniMol Tools

[am2145]

Hello,

I was wondering if it is possible to use the unimol-tools interface to train on a pre-split dataset? Is there a way to pass in a specific split if I have one predetermined?

I can almost achieve this effect using the group option and providing the group columns as my train and test splits with 2 folds. However, this would have one split with my training and test sets as desired, and one in reverse. Is there a way to run predict for an individual model instead of averaging across all models?

Thank you!

[Naplessss]

Hi am2145, thanks for your suggestions. Currently a pre-split dataset is not support, even with group split. we will add this features ASAP.

[am2145]

Thanks for the quick response.

I'll wait for the implementation and use the in-built splits in the meantime.

Out of curiosity, is it possible to predict for individual models so that I can compare the performance of the models trained on the different folds?

Thank you again for your help. I'm looking forward to trying out UniMol more.