I have a issue similar to this one (Training nlu without restarting), but as I said in the post, using the same name is an acceptable solution.
My problem is that if train the new model with the same, and if someone send a message to the NLU when the new model is being saved, the NLU will stop working for an obscur reason. I could understand that it can’t access the old or the new model, but I would prefer a correct error rather than the NLU being blocked and needing to be restarted.
Anyone have a solution for this ? I’m open to everything that works to train the NLU without restarting the server.