rasa_core.policies.ensemble.InvalidPolicyConfig: You didn't define any policies. Please define them under 'policies:' in your policy configuration file.

[Arghya999]

Rasa Core version:0.13.0

Python version: Python 3.6.8
Operating system (windows, osx, ...): Ubuntu 16.04 LTS

Issue: rasa_nlu and rasa_core training error :-
INFO:rasa_nlu.training_data.loading:Training data format of nlu.md is md
INFO:rasa_nlu.training_data.training_data:Training data stats:
- intent examples: 188 (15 distinct intents)
- Found intents: 'about_OPP', 'city_name', 'OPP_Phones_Price', 'goodbye', 'CustomerCare_info', 'provide_founder', 'brand_info', 'mood_deny', 'greet', 'model_name_lookout', 'Official_Website', 'CEO_info', 'mood_affirm', 'Service_Center_OR_Factory', 'Opp_store_location'
- entity examples: 99 (7 distinct entities)
- found entities: 'workshop_location', 'shop_location', 'web_address', 'model', 'information', 'customer_location', 'PERSON'

DEBUG:rasa_nlu.training_data.training_data:Validating training data...
Traceback (most recent call last):
File "bot_enhancement_updated.py", line 53, in
trainer = Trainer(config.load('config_mitie_sklearn.yml'))
File "/home/arghya/anaconda3/envs/myenv/lib/python3.6/site-packages/rasa_core/config.py", line 22, in load
return PolicyEnsemble.from_dict(config_data)
File "/home/arghya/anaconda3/envs/myenv/lib/python3.6/site-packages/rasa_core/policies/ensemble.py", line 222, in from_dict
raise InvalidPolicyConfig("You didn't define any policies. "
rasa_core.policies.ensemble.InvalidPolicyConfig: You didn't define any policies. Please define them under 'policies:' in your policy configuration file.

Content of policy.yml
'''yaml
policies:

• name: KerasPolicy
  epochs: 50
  max_history: 6
• name: AugmentedMemoizationPolicy
  max_history: 6
• name: TwoStageFallbackPolicy
  nlu_threshold: 0.8
  core_threshold: 0.3
• name: FormPolicy
  '''
  Content of nlu_congif.yml(name of the file :- )

'''yaml
language: "en"

pipeline:

• name: "nlp_mitie"
  model: "/home/arghya/mitie_models/MITIE-models/english/total_word_feature_extractor.dat"
• name: "tokenizer_mitie"
• name: "ner_mitie"
• name: "ner_synonyms"
• name: "intent_entity_featurizer_regex"
• name: "intent_featurizer_mitie"
• name: "intent_classifier_sklearn"
  '''

bot logic for rasa_nlu and rasa_core training :-

-- coding: utf-8 --

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from __future__ import unicode_literals

rasa nlu imports

import logging
import warnings
from rasa_nlu.training_data import load_data
from rasa_nlu.config import RasaNLUModelConfig
from rasa_nlu.model import Interpreter, Trainer
from rasa_nlu import config
from rasa_nlu.training_data.loading import load_data_from_endpoint
from rasa_nlu.utils import EndpointConfig, read_endpoints

rasa core imports

import rasa_core
from rasa_core import config, cli
from rasa_core import utils
from rasa_core.agent import Agent
from rasa_core.run import AvailableEndpoints
from rasa_core.policies import FallbackPolicy, KerasPolicy, MemoizationPolicy
from rasa_core.interpreter import NaturalLanguageInterpreter
from rasa_core.domain import Domain
from rasa_core.utils import EndpointConfig
from rasa_core.run import serve_application
from rasa_core.tracker_store import RedisTrackerStore
from rasa_core.tracker_store import MongoTrackerStore

logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger(__name__)
warnings.filterwarnings('ignore')

##training nlu

if the probability of the intent/action is too low , we will strop predicting and use a default action.

global_fallback = FallbackPolicy(fallback_action_name='utter_unclear',
core_threshold=0.3,
nlu_threshold=0.8)

training_data = load_data('nlu.md')

trainer to educate our pipeline

trainer = Trainer(config.load('config_mitie_sklearn.yml'))

training the nlu model

interpreter = trainer.train(training_data)

storing the model for future use

global_model_directory = trainer.persist('./models/nlu',
fixed_model_name='current')

training nlg -------------------->

domain = Domain.load('domain.yml')

global_db_mg = MongoTrackerStore(
domain,
host='mongodb://localhost:27017',
db='rasa',
username=None,
password=None,
collection='conversations',
event_broker=None,
)

def train_dialogue(
domain='domain.yml',
model_path='models/dialogue',
training_data_file='stories.md',
endpoints=AvailableEndpoints(),
policy_config=None,
):

Train a dialogue model

nlu_interpreter = RasaNLUInterpreter('./models/nlu/default/')

policies = config.load('policy.yml')
nlu_interpreter = global_model_directory
endpoints = '/home/arghya/rasa_test/data/place_finder/endpoints.yml'
agent = Agent(
    domain=domain,
    policies=policies,
    interpreter=nlu_interpreter,
    generator=None,
    tracker_store=global_db_mg,
    action_endpoint=endpoints,
    fingerprint=None,
    )
data = agent.load_data(training_data_file)
agent.train(data, epochs=400, batch_size=50, validation_split=0.2)
agent.persist(model_path)
return agent

if __name__ == '__main__':
agent = train_dialogue()
rasa_core.run.serve_application(agent, channel='cmdline')

extract list of events and print its names

events = global_db_mg.retrieve('default').as_dialogue().events
for event in events:
    print(event.as_story_string())

Content of domain file (if used & relevant):

The issue is with the training logic since policy.yml is available and configured and I am able to run rasa_nlu and rasa_core training using

1)python3 -m rasa_nlu.train -c config_mitie_sklearn.yml --fixed_model_name current --data nlu.md -o models --project nlu --verbose
2)python3 -m rasa_core.train -d domain.yml -s stories.md -c policy.yml --debug -o models/dialogue.

I have just upated the rasa_core version to 0.13.0 and not able to figure out how to pass the policy.yml for training.
Expert guidance is much appreciated.

[akelad]

Thanks for raising this issue, @erohmensing will get back to you about it soon.

[erohmensing]

Hi! What seems to be happening here is that when you call

trainer = Trainer(config.load('config_mitie_sklearn.yml'))

it's pulling config.load() from core, not nlu. This is because you have from rasa_nlu import config but this is overwritten by from rasa_core import config, cli. I think the easiest fix here would be to rename one of them, for example from rasa_nlu import config as nlu_config (and then use this name when you call the load function)

[Arghya999]

Hi @erohmensing , corrected the logical flaw.Thanks for the help.