Setting augmentation factor in rasaX

[Deprecated] Rasa X Community Edition
1

I just started with rasa and rasaX, so kind of stuck with setting augmentation factor, how do I set that while using rasaX? I have read that when using Embedding Policy, augmentation factor has to be 0. Also I would like know about changing Embedding Policy defaults, where do I change that? Thanks.

2

Hi @faheemv, welcome to the forum! You can specify the augmentation factor when training a model in the command line. E.g.

rasa train --augmentation 100

The default value is 50. Please check out the Rasa Tutorial. You can always run rasa train --help for an overview of command-line options.

As for the embedding policy, data augmentation is applied to all training stories and is independent of the chosen policy. Have a look at https://rasa.com/docs/rasa/core/policies on how to change policy defaults.

I hope that helps!

3

Thanks for the help,

4

Your answer is for command-line.

Origin question is about Rasa x.

I still want to know how to set the augmentation factor in rasa x / openshift. For when i use either augmentation_factor or augmentation, I’m getting an error:

Blockquote

ERROR:rasax.community.api.blueprints.models:500, Internal Server Error, body=‘b’{“version”:“1.4.2”,“status”:“failure”,“message”:“An unexpected error occurred during training. Error: init() got an unexpected keyword argument 'augmentation_factor'”,“reason”:“TrainingError”,“details”:{},“help”:null,“code”:500}‘’ ERROR:rasax.community.api.blueprints.models:500, Internal Server Error, body=‘b’{“version”:“1.4.2”,“status”:“failure”,“message”:“An unexpected error occurred during training. Error: init() got an unexpected keyword argument 'augmentation'”,“reason”:“TrainingError”,“details”:{},“help”:null,“code”:500}‘’

Blockquote

These are my policies:

policies:

  • name: KerasPolicy max_history: 1
  • name: MappingPolicy
  • name: MemoizationPolicy max_history: 1
  • name: AugmentedMemoizationPolicy augmentation: 0