I have noticed this too in Rasa 2x version. So, i figure that the Tensorflow_embedding pipeline was introduced in RASA NLU 0.12, changed to “supervised_embeddings” pipeline in Rasa 1x version and later was deprecated in higher versions. The migration guides show an Embedding Intent Classifier , which is now replaced by DIET classifier. So, my question is the Rasa 's suggested config, derived from the original Tensorflow_embedding pipeline, since they share many similar components?
