Model documentation and templates

What kind of templates are available through ValidMind?

ValidMind supports the following template types:¹

¹ View documentation templates

• Documentation
• Validation Report
• Ongoing Monitoring

Can templates be configured per use case or to match our own?

Yes, the ValidMind Platform² allows you to configure versioned templates based on documentation requirements for each model or lifecycle use case.

² Accessing ValidMind

• ValidMind’s templates are fully customizable,³ and are complemented by the ability to manage validation guidelines.
• You can swap between different versions of templates or apply another version of the current template.⁴
• Template outlines can be modified right in the ValidMind Platform. We also offer an advanced raw YAML editor, enabling you to make detailed changes.

³ Customize documentation templates

⁴ Swap documentation templates

⁵ Manage permissions

By default, the Customer Admin role⁵ has sufficient permissions to manage documentation templates, and you must be a Developer or Validator to be able to swap documentation templates.

Can documentation be created right in the ValidMind Platform?

Yes, you can write documentation directly in the ValidMind Platform, without having to first generate anything using the ValidMind Library.⁶

⁶ Get started with the ValidMind Library

1. Add and edit text on any documentation within the platform using our content editing toolbar.⁷
2. Using the library, execute test suites and generate the corresponding supporting results. These results can then be added to your documentation within the platform.⁸

⁷ Work with content blocks

⁸ Work with test results

Can I run tests and log documentation without a model?

Yes! If you do not have a model ready, or your model can’t be loaded directly, or you only have access to model predictions, you can still run tests and log documentation using the ValidMind Library as long as you’re able to load the model predictions.

• Use assign_predictions()⁹ to load predictions from a separate file or a dataset with predictions.
• Call init_model()¹⁰ but instead of a trained model instance, pass an input_id and model metadata. ModelMetadata()¹¹ will use the provided metadata instead of trying to calculate it from the model’s library.

⁹ assign_predictions()

¹⁰ init_model()

¹¹ ModelMetadata()

For example:

  external_model = vm.init_model(
      input_id="spark_ml_model",
      metadata={...}
  )

If neither a trained model instance nor metadata is provided, init_model() will return an error.

However, tests that need a trained model will not work with “empty” models.

Can we attach files to models, findings, or documentation?

Yes, attachment type inventory fields are available for custom use.¹² Once created, attachment type fields allow you to upload supporting files to your model.

¹² Manage model inventory fields

• Out-of-the-box functionality is included for attaching files to model findings.
• You can also attach images to documentation content blocks and comments.

By default, the Customer Admin role has sufficient permissions to manage model inventory fields.

Do you support including images in model documentation?

Yes, as long as you can produce the image with Python or open the image from a file, you can include it in your documentation with ValidMind:¹³

¹³ Implement custom tests

• If you want to log an image as a test result, you can do so by passing the path to the image as a parameter to the custom test and then opening the file in the test function.
• If you are using a plotting library that isn’t directly supported by ValidMind, you can still return the image directly as a bytes-like object.

Can documentation be exported?

Yes, ValidMind currently supports exporting model documentation and validation reports in Word (.docx) format.¹⁴

¹⁴ Depending on the integration requirements of the systems used by your validation teams, such as connectivity via API, SharePoint, and more, ValidMind can work with you to automate the export and storage of documentation into these systems.

Learn more