Background: (Back before transform and itransform we used train() for batches of records and update() for single records. Now we expect train to receive a batch and let OnlineLearningWrapper.transform() handle single records.)
So,
Is there any difference any more?
Do we need change LocalTrainer to call train instead?
Do we need to add update() back into stateful model or do we need to clean-up GensimSimilarity?