I have a model with 320 intents. I was running the nlu testing and tried to examine the confusion matrix which is generated from the testing, but it is so packed that it’s unreadable (even when zooming in to the max).
It there a way to handle it somehow?
Thanks an stay healthy
Hi @Aviad, Thanks for pointing this out! We are aware of the problem but solving it is currently not a priority.
You could try adjusting the code e.g. to have several smaller plots with parts of the matrix. The function that creates the matrix plot is this one. If you come up with a nice general solution, you are very welcome to create a pull request so we can include the fix in a future release of Rasa Open Source.