Rasa: Allow config path to be passed to Rasa X

[ortsaCniveK]

Description of Problem:

Because users are able to pass custom config paths to rasa train, there should be CLI parity for the same.

Overview of the Solution:

Accept custom config path to override default config path, unless none is passed, then use default.

That, or allow for a config/ folder to be passed for all configuration-related files.

Examples (if relevant):

Mainly would help clean up some of the folder structure; without using Rasa X, you can have something like this:

├── actions
├── config
    ├── config.yml
    ├── credentials.yml
    └── endpoints.yml
├── data
    ├── nlu
    └── stories
├── logs
└── domain.yml

and pass the config file by passing --config config/config.yml

instead of having the config files clutter up the root directory.

Blockers (if relevant): N/A

Definition of Done:

• [ ] Tests are added
• [ ] config path handling is changed
• [ ] Feature mentioned in the changelog

[sara-tagger]

Thanks for submitting this feature request 🚀@wochinge will get back to you about it soon!✨

[wochinge]

@ortsaCniveK Thanks for bringing this. I agree, that this could be useful 👍

[SvenBecker]

Furthermore I would like to set a path to domain.yml and the models folder? Are there any plans to make this configurable too?

[ortsaCniveK]

I think that's a good idea; there should be absolute parity between the rasa train command options, of which candidates are:

  -c CONFIG, --config CONFIG
                        The policy and NLU pipeline configuration of your bot.
                        (default: config.yml)
  -d DOMAIN, --domain DOMAIN
                        Domain specification (yml file). (default: domain.yml)
  --out OUT             Directory where your models should be stored.
                        (default: models)
  --augmentation AUGMENTATION
                        How much data augmentation to use during training.
                        (default: 50)
  --debug-plots         If enabled, will create plots showing checkpoints and
                        their connections between story blocks in a file
                        called `story_blocks_connections.html`. (default:
                        False)
  --dump-stories        If enabled, save flattened stories to a file.
                        (default: False)
  --fixed-model-name FIXED_MODEL_NAME
                        If set, the name of the model file/directory will be
                        set to the given name. (default: None)
  --persist-nlu-data    Persist the nlu training data in the saved model.
                        (default: False)
  --force               Force a model training even if the data has not
                        changed. (default: False)

Not sure if this is something we'd want to add to the CLI or just abstract into another config file, though.

[ArjaanBuijk]

The original request is now implemented:

-c CONFIG, --config CONFIG
                        The policy and NLU pipeline configuration of your bot.
                        (default: config.yml)

@wochinge , what do you think about implementation of the other command options:

  -d DOMAIN, --domain DOMAIN
                        Domain specification (yml file). (default: domain.yml)
  --out OUT             Directory where your models should be stored.
                        (default: models)
  --augmentation AUGMENTATION
                        How much data augmentation to use during training.
                        (default: 50)
  --debug-plots         If enabled, will create plots showing checkpoints and
                        their connections between story blocks in a file
                        called `story_blocks_connections.html`. (default:
                        False)
  --dump-stories        If enabled, save flattened stories to a file.
                        (default: False)
  --fixed-model-name FIXED_MODEL_NAME
                        If set, the name of the model file/directory will be
                        set to the given name. (default: None)
  --persist-nlu-data    Persist the nlu training data in the saved model.
                        (default: False)
  --force               Force a model training even if the data has not
                        changed. (default: False)

[wochinge]

I think

• it makes sense for --domain and --out
• it doesn't make sense for --fixed-model-name, --force,--dump-stories and --debug-plots
• it would make sense to allow --augmentation, --persist-nlu-data to be specified in the config,yml file

What do you think?

[ortsaCniveK]

@wochinge I agree with above, it would be really nice to just run rasa train and specify training-related parameters in the config file and let the CLI just handle where output is sent or where input is found. It would make a bot more portable if the training-related parameters are specified in the config.