Rasa stopwords issue

Rasa Open Source
1

Earlier my config.yml was:

language: en
pipeline: supervised_embeddings

policies:
  - name: MemoizationPolicy
  - name: KerasPolicy
  - name: MappingPolicy

But then I had to configure stopwords, so now my file is:

language: en
pipeline:
- name: "WhitespaceTokenizer"
- name: "RegexFeaturizer"
- name: "EmbeddingIntentClassifier"
- name: "CRFEntityExtractor"
- name: "EntitySynonymMapper"
- name: "CountVectorsFeaturizer"
  stop_words: {'I', 'in', 'the'}

policies:
  - name: MemoizationPolicy
  - name: KerasPolicy
  - name: MappingPolicy

But I’m getting below error on executing rasa train: ValueError: Cannot feed value of shape (64,) for Tensor ‘a:0’, which has shape ‘(?, 417)’

How to resolve this, Also is the 2nd config.yml equilvalent to supervised_embeddings other than stopwords change?

2

I realize that this

stop_words: None # string {‘english’}, list, or None (default)

Could be a little confusing. I believe you want

- name: "CountVectorsFeaturizer"
  stop_words: ['I', 'in', 'the']

As for the pipeline, it’s not exactly equivalent, as the components are out of order – not sure if that is causing your error, as output from previous components are used as input for later ones. Here is the supervised embeddings pipeline:

pipeline:
- name: "WhitespaceTokenizer"
- name: "RegexFeaturizer"
- name: "CRFEntityExtractor"
- name: "EntitySynonymMapper"
- name: "CountVectorsFeaturizer"
- name: "EmbeddingIntentClassifier"
3

Thanks @erohmensing, That was the issue :slight_smile:

4

Awesome, happy to help :rocket: