RasaHQ/rasa

rasa data validate error, schema issue

Closed this issue · 3 comments

rasa data validate fails on the Sara/rasa-demo config.yml with the error: You have provided an invalid key token_pattern for component 'CountVectorsFeaturizer' in your pipeline. Valid options for 'CountVectorsFeaturizer' are:

Sara uses token_pattern here.

The docs do not state that CountVectorsFeaturizer supports token_pattern but it looks like this is supported in the code.

Rasa version: 2.2.2

Hi @rgstephens, we decided not to expose this parameter as we're trying to keep tokenization-related parameters in the tokenizers. The Sara parameters are actually the default for "analyzer":"word". I'll make a new issue on the Sara repo to change this.

similar error throws on command rasa train on docker image rasa/rasa:2.4.0-full but instead of token_pattern its model_confidence.

@adamfeher64 can you share your config file?