Model Approvals
Approval workflows ensure that model promotions (e.g., Staging to Production) are reviewed by team members before taking effect.
How Approvals Work
- A team member requests a model stage transition
- An approval request is created automatically
- Designated reviewers are notified
- Reviewers evaluate the model and approve or reject
- On approval, the stage transition completes
Viewing Pending Approvals
Navigate to MLOps > Approvals to see:
- All pending approval requests
- Model name and version
- Requested stage transition (e.g., Staging → Production)
- Requester name and timestamp
- Current review status
Reviewing an Approval
- Click on a pending approval
- Review the model version:
- Training metrics and performance
- Comparison with current production version
- Dataset and training parameters
- Model lineage
- Add a comment explaining your decision
- Click Approve or Reject
Approval Status
| Status | Description |
|---|---|
| Pending | Awaiting reviewer action |
| Approved | Reviewer approved the transition |
| Rejected | Reviewer rejected the transition |
Best Practices
- Always add comments - Explain why you approved or rejected
- Check metrics - Compare against the current production model
- Review lineage - Verify the training data and parameters
- Test first - Ensure the model was evaluated in Staging before Production promotion
Next Steps
- Model Registry - Manage model versions and stages
- Experiment Tracking - View training details