Hello, I am interested in using dropout during inference time as a way to assess model confidence for a tabular model I have trained. It appears that the latest versions of fastai have learn.predict_with_mc_dropout and learn.pred_batch(dropout=True) to accomplish this for the scenario of either (1) making a prediction on a single item or (2) a single batch of items. However, I would like to be able to perform this operation on a varying sized data set during inference time. I have considered the idea of creating a data bunch with the inference data, and setting the batch size equal to the inference data size and using pred_batch(dropout=True) as a fast hack. I think this should return a prediction for all items in the inference data set with dropout if I set the validation set size to 0 when creating the databunch. However, I am wondering if there is a more elegant solution. get_preds appears to be the function best suited for this type of problem, but It appears that the current get_preds function does not have a dropout enabling option. Also, it appears that the validate function called by get_preds in basic_train.py hard codes model.eval(). Does anyone have any ideas on how I could perform get_preds with dropout enabled?