Is rasa remove duplicates training utterances while training?
It does, but only recently (>= 1.5.0): https://github.com/RasaHQ/rasa/pull/4414
Having multiple training examples for the same utterance was an odd way to add weights to the training data. Is there a possibility to add weights in a formal way? @erohmensing
Could just think of other odd ways, which are specific to certain classifiers.