Write custom component output to json when executing rasa test nlu in CLI

Hi, I have trained a rasa nlu model with intents (using supervised_embeddings) combined with a custom component to extracts nouns. Using rasa shell nlu I get the intent and custom output returned in command line. However, using rasa test nlu i dont get the custom output saved with the successes and errors. Are there a way to get that output? Basically, I want to evaluate my custom component output through manually inspection, i.e. writing it to the error and success json files too.

I have tried to load my saved model into python but get the “The model version is too old to be loaded by this Rasa NLU instance. Either retrain or run with an older version. Model version: 0.0.0 Instance version: 0.13.02”

Reading a bit says Model version:0.0.0 means a faulty NLU model which is not the case given my rasa shell and rasa test outputs in CLI.

You need to add printing code from your custom component to corresponding method in rasa test

1 Like