Is it normal that total samples changes

Rasa Open Source
1

Hi without changing anything and running the training I have different values for the trackers and samples is it normal?

Once

2018-11-21 10:53:17 DEBUG    rasa_core.training.generator  - Finished phase (30343 training samples found).
2018-11-21 10:53:17 DEBUG    rasa_core.training.generator  - Found 30343 training trackers.
...
2018-11-21 10:53:43 DEBUG    rasa_core.featurizers  - Created 37360 action examples.

Another time

2018-11-21 10:54:51 DEBUG    rasa_core.training.generator  - Finished phase (29527 training samples found).
2018-11-21 10:54:51 DEBUG    rasa_core.training.generator  - Found 29527 training trackers.
...
2018-11-21 10:55:19 DEBUG    rasa_core.featurizers  - Created 39987 action examples.

rasa-core==0.12.2 rasa-core-sdk==0.12.1 rasa-nlu==0.13.8

2

Probably is it normal, since I didn’t have any answer

3

Hey @Neofita Yes, that’s normal, because Rasa Core by default samples and augments the data you have in your training example, that’s why in every augmentation round you should get different number of training examples and trackers :slight_smile:

4

Hi @Juste, thank you for the answer, but I really do not understand since I run this command multiple times

python3 -m rasa_core.train -d domain.yml -o models/dialogue/ -vv -s data/stories -c config.yaml

on every run I do

rm models/dialogue -Rf

and this what I got (Memoried X unique examples) where X is

40025

41271

39308

40553

And I do not modify anything else if I understand correctly, if i run it multiple time without erasing this value is going to increase, ore the information that training mode generate are stored somewhere else. Thank you