Question about data format for prediciton

[tomh014]

Question about data format for prediciton

[tomh014]

Update for anyone interested:
The repurpose function in CompoundPred can use the model to predict without any labels, and gives the exact same result as the code above with "fake" labels using the predict function, meaning predict doesn't use the labels and both functions do the same thing, although repurpose takes in SMILES and encodes while predict takes in encodings. For large datasets, it seems preferable to use predict since encoding can be time-consuming, but this is a little unclear in the documentation.