Hello everyone, I hope you're doing well. I have two quick questions about the rules for the ROAD Barbados Historic Handwriting Challenge:

1. Are hyperparameter optimization libraries such as Optuna allowed? I understand the rules restrict AutoML tools — I'd like to confirm whether standard hyperparameter search (learning rate, dropout, etc.) on our own models falls under that restriction or not.
2. During cross-validation we identified ~20 training rows with corrupted labels (e.g., image–label mismatches, an empty image). Is it allowed to exclude these rows from training? To be clear: no test data is touched in any way, no labels are modified, and the original Train.csv stays unchanged — we would only skip these rows when training.

Thanks a lot!