Publish a model to a model risk manager, or for peer review. When a model is published, it can be reviewed and accepted or rejected before it is deployed to production.
Note that model review is a mandatory process. Models have to be published to a reviewer and accepted before they can be deployed to production.
To publish a model for review
- Click the publish icon to publish the model you want to review.
- Upload any document related to the model or use case. E.g: a description of the use case with objectives and expected outcomes.
- Fill in the needed information
- Special Comment for deployment
Comment by model creator
Click Change reviewer to select a reviewer
Click Publish to Reviewer to publish the model to the user selected in iii.
- An auto-generated email with information about the published model is sent, requesting the reviewer to log into the platform to review the model (see model manager for details on what the reviewer does).
- Once the model is reviewed and either accepted or rejected, the model creator gets an email with instructions on what to do.
- If it is accepted, then the model can be deployed to production.
- If rejected, see the comments, prepare and publish a new version again.
Note that a published model is no longer seen in the ML Engine leaderboard.