Training with rasa 1.0 (sample amount, story directory)

I have two questions:

  1. How to specify a story directory in stead of a story.md file with the new train (rasa train) command?

  2. For now, I added all my stories to one story file, and tried to train.

When I use my old version and the old training method I get this:

And when I use the new method:

Can someone explain why the amount of total samples is different when fitting the model? Cause the accuracy is much lower now…

Regarding my second issue, I found the reason.

Apparently the
max_number_of_augmented_trackers = augmentation_factor * 10 in rasa.core.training.generator gets activated with the new version, while for my old version, it didn’t

Hi @rbossie, if you put all of your data (stories, nlu data) within a data folder, that should work. e.g. you might have the structure

-- data
   |---stories
       |--- stories_1.md
       |--- stories_2.md
   |---nlu
       |--- nlu_1.md
       |--- nlu_2.md