Unable to train CountVectorizer for message attribute response

While training the bot with rasa train nlu I am getting the following error: Unable to train CountVectorizer for message attribute response. Leaving an untrained CountVectorizer for it

2019-09-14 22:24:27 INFO rasa.nlu.model - Starting to train component WhitespaceTokenizer 2019-09-14 22:24:27 INFO rasa.nlu.model - Finished training component. 2019-09-14 22:24:27 INFO rasa.nlu.model - Starting to train component RegexFeaturizer 2019-09-14 22:24:27 INFO rasa.nlu.model - Finished training component. 2019-09-14 22:24:27 INFO rasa.nlu.model - Starting to train component CRFEntityExtractor 2019-09-14 22:24:27 INFO rasa.nlu.model - Finished training component. 2019-09-14 22:24:27 INFO rasa.nlu.model - Starting to train component EntitySynonymMapper 2019-09-14 22:24:27 INFO rasa.nlu.model - Finished training component. 2019-09-14 22:24:27 INFO rasa.nlu.model - Starting to train component CountVectorsFeaturizer 2019-09-14 22:24:27 WARNING rasa.nlu.featurizers.count_vectors_featurizer - Unable to train CountVectorizer for message attribute response. Leaving an untrained CountVectorizer for it 2019-09-14 22:24:27 INFO rasa.nlu.model - Finished training component. 2019-09-14 22:24:27 INFO rasa.nlu.model - Starting to train component CountVectorsFeaturizer 2019-09-14 22:24:27 WARNING rasa.nlu.featurizers.count_vectors_featurizer - Unable to train CountVectorizer for message attribute response. Leaving an untrained CountVectorizer for it 2019-09-14 22:24:27 INFO rasa.nlu.model - Finished training component. 2019-09-14 22:24:27 INFO rasa.nlu.model - Starting to train component EmbeddingIntentClassifier Epochs: 100%|█████████████████████████████████████████████████| 300/300 [00:06<00:00, 46.89it/s, loss=0.479, acc=1.000] 2019-09-14 22:24:36 INFO rasa.utils.train_utils - Finished training embedding policy, train loss=0.479, train accuracy=1.000

Hi @subitha, if you aren’t trying to use the new ResponseSelector, no need to worry about this warning! We are currently working on a fix to make sure that the warning only shows up if you’re trying to use it and something goes wrong. :slight_smile:

Hi @subitha , are you specifying any retrieval actions or response phrases in your training data for response selector? If not, then the log is just a warning and you can ignore it. If you do have some data for response selector in your training data and you are still seeing this, please post a small snippet of your data and exact Rasa version.

In the first case you shouldn’t see the warning if you update the Rasa package to version 1.3.3 anyways.