Rasa Training Killed unexpectedly

Rasa Open Source Feedback on Rasa Open Source
1

Hello,

When running rasa train the process goes well until a certain point and then it stops unexpectedly. These are the last lines of the outuput (I ran this in the debug mode):

Processed trackers: 100%|█████████████████████████████| 9927/9927 [00:13<00:00, 743.52it/s, # action=129719]
2022-09-30 15:13:09 DEBUG    rasa.core.featurizers.tracker_featurizers  - Created 129719 action examples.
/home/tlabruna/anaconda3/rasaProva/lib/python3.9/site-packages/rasa/utils/tensorflow/model_data_utils.py:388: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
  np.array(values), number_of_dimensions=4
2022-09-30 15:20:11 DEBUG    rasa.core.policies.ted_policy  - Entity recognition cannot be performed, set 'entity_recognition' config parameter to 'False'.
/home/tlabruna/anaconda3/rasaProva/lib/python3.9/site-packages/rasa/utils/tensorflow/model_data.py:529: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
  np.array(
2022-09-30 15:20:45 DEBUG    rasa.utils.tensorflow.data_generator  - The provided batch size is a list, this data generator will use a linear increasing batch size.
Killed

Thank you, Tiziano

2

Does rasa data validate report any errors?

3

Hi @stephens , thank you for the answer.

That command returns a lot of warnings, but no errors. The warnings are of this type:

/home/tlabruna/anaconda3/rasaProva/lib/python3.9/site-packages/rasa/shared/utils/io.py:98: UserWarning: The 
example 'Can I get a taxi to take me between the two ?   I need to leave the Cafe Jello attraction by 21:00 
.   I need the contact Number and car type ?' was found labeled with multiple different intents in the train
ing data. Each annotated message should only appear with one intent. You should fix that conflict The exampl
e is labeled with: Taxi-Inform, Taxi-Request.

or this one:

/home/tlabruna/anaconda3/rasaProva/lib/python3.9/site-packages/rasa/shared/utils/io.py:98: UserWarning: The utterance 'utter_Hotel-Ref' is not used in any story or rule.

Do you think this can be the problem? How can I fix it?

4

I suspect it’s something about your training data. Can you share the repo?

5

Hi @stephens, sure, this is the repo: tLabruna/rasaMultiWOZ (github.com)

6

You have too many utterance examples. Your intents have a lot of confusion. You need to manually review and clean this up.

Regarding your stories, the bot: feature is experimental. I don’t know anyone using this for a production bot.