When i add new action in my actions.py and trains the bot, it gives this error "ValueError: X.shape[1] = 256 should be equal to 128, the number of features at training time"

Rasa Open Source
1

I am running a weather bot, now i have added new entities and a new action in my bot. When i try to train it using rasa train it gives the below mentioned error:

ValueError: X.shape[1] = 256 should be equal to 128, the number of features at training time

Any idea how can i correct this?

2

hi @Sid - welcome to the forum! Can you please share your config file? This is happening when you run rasa train or rasa run ?

3

Hi @amn41 thanks for your reply! The error comes when i run “rasa shell” after “rasa train”. While interacting with the bot in “rasa shell” mode, as soon as i enter any text, for eg. Hi… the bot throws this error that i have mentioned in my first post.

This is my config.yml:

language: “en” pipeline: “pretrained_embeddings_spacy”

policies:

  • name: MemoizationPolicy
  • name: KerasPolicy epochs: 200

Basically, what i am trying to do it, I am using a pre-trained Weather bot that i got from Github. Now i am trying to add new intents, entities and custom actions in it to modify it for my requirements. After adding new training data in the data.json, once i train it and then use rasa shell, it throws this error.

Will appreciate your help in this!

Thanks, Sid

4

hi @Sid - could you try switching the pipeline to “supervised_embeddings” and checking if you still get this error after training a new model?