Incremental training using supervised_embeddings

Hi All , I want to train my bot with RASA NLU . But i want to do it in a incremental way. As i have trained it for 25 intents already, when i train it next for another intent , i want it to train it for only one intent now, not all. is this possible?

@rohitiec, no, this is not possible. Supervised embeddings creates word embeddings based on all of your data. Therefore adding new data would change the word embeddings of the old data, so it must all be re-trained to correctly reflect the relationships between the words found in each of the intents.

1 Like