I am currently uploading an Avocado model that is trained on both the training and the validation set, as well as that model's imputations of the training, validation and test sets. To be clear, these are imputed versions of these tracks, not the actual tracks. If your model uses Avocado in any manner, you should use these models and imputations to make your held out blind test set predictions. For example, if your model boosts off Avocado's predictions, you'll want to train your final model using the imputed versions of the training and validation tracks and then use the imputed test set tracks to make your final test set predictions. You should not use these files to make predictions on the validation set. Let me know if you have any questions! You can find them here: https://www.synapse.org/#!Synapse:syn20606511