RASA Core Train Stuck

Rasa Open Source
1

Hi everyone,

Have anyone faced issue with rasa.core.train ? I tried to train core, and it’s not large stories, but i used extensive checkpoints there. When i train it on AWS ec2 t2.2xlarge instance with jupyter notebook and with this code:

from rasa.core import train
from rasa.core.agent import Agent
from rasa.core.policies.fallback import FallbackPolicy
from rasa.core.policies.keras_policy import KerasPolicy
from rasa.core.policies.mapping_policy import MappingPolicy
from rasa.core.policies.form_policy import FormPolicy
from rasa.core.policies.memoization import MemoizationPolicy
user_policies = [MemoizationPolicy(),MappingPolicy(),KerasPolicy(epochs=55, max_history=15), FormPolicy(),FallbackPolicy(nlu_threshold=0.65)]
a = Agent('data/domain.yml', user_policies)
story = await a.load_data('data/dialogue/', augmentation_factor=0)
a.train(story)
a.persist('models/dialogue/')

Screenshot%20from%202019-08-06%2012-22-49
Screenshot%20from%202019-08-06%2012-22-49988×102 24.8 KB
Is this code deprecated? Can someone help me with this? Thanks

2

It might be the case that something has gone wrong in your checkpoints, you can check the validity of your stories using rasa test.

3

Hi @imLew,

So i tried on my local machine, and it’s working, maybe it’s the AWS setup. I used t2.2xlarge but stuck, but my local machine only use core i7 with 16 gb rams, and working just find.

4

Yes, if the same code runs fine on your local machine then it has to be an issue with AWS.