Training and testing Rasa NLU models

[vmugra]

AS A Service Manager I WANT TO train and test Rasa NLU models created via Training Module SO THAT I could find errors if they exist and be confident that only valid models will move on to production

Acceptance Criteria

The scope of this task is to remotely initiate Rasa terminal commands - spike, @turnerrainer

Rasa tests run as terminal commands are applied to

• [ ] CI/CD pipeline for automated testing
• [ ] NLU model tests, Data and stories validation tests, i.e. Command line: rasa test --out results/model
• [ ] Cross-Validation tests on NLU models, Dialogue model evaluations, i.e. Command line: rasa test --cross-validation --folds 8 --out results/model/cross - where you can choose specific folds amount
• [ ] Data and stories validation tests

If training has bee executed, rasa train and rasa test command lines will be started!

• [ ] Epoch testing on Tensorboard - not being done!

GUI

• [x] #60

DSL

• [ ] Commands can be initiated on demand
• [ ] Commands can be initiated on pre-defined time (cronjob)

[turnerrainer]

1. Provide Rasa testing results (PDF, images) via GUI;
2. Launch Rasa train via GUI (manually + pre-defined timeslots);
3. Rasa must provide information about whether Rasa is currently in progress of training or not (DSL provides and endpoint for this).

[janyprus]

sample input for /rasa/training/command

{
    "command": "rasa test --cross-validation --folds 8 --out results/model/cross"
}

Pull request #199