Promoting models between stages in MLOps - Commands & Configuration

Introduction

When you train a machine learning model, you want to test it carefully before using it for real. Promoting models between stages means moving a model from testing to production safely. This helps avoid mistakes and keeps your app working well.

When you want to move a model from development to testing without breaking anything

When you need to approve a model after checking its accuracy before using it live

When you want to keep track of which model version is currently in production

When you want to roll back to a previous model if the new one causes problems

When you want to automate the process of moving models through different quality checks

Commands

This command moves version 3 of 'my-model' to the Production stage, making it the live model used by your app.

Terminal

mlflow models transition-stage --model-name my-model --version 3 --stage Production

Expected Output

Model version '3' of model 'my-model' transitioned to stage 'Production'.

→

--model-name - Specifies the name of the model to promote

→

--version - Specifies the version number of the model to promote

→

--stage - Specifies the target stage to promote the model to

This command lists all versions of 'my-model' and shows their current stages, so you can verify the promotion.

Terminal

mlflow models list-versions --model-name my-model

Expected Output

Version  Stage

1        Staging

2        Archived

3        Production

→

--model-name - Specifies the model to list versions for

This command moves version 2 of 'my-model' to the Archived stage, marking it as no longer active.

Terminal

mlflow models transition-stage --model-name my-model --version 2 --stage Archived

Expected Output

Model version '2' of model 'my-model' transitioned to stage 'Archived'.

→

--model-name - Specifies the model name

→

--version - Specifies the version to archive

→

--stage - Specifies the Archived stage

Key Concept

If you remember nothing else from this pattern, remember: promoting a model means changing its stage label to control which version your app uses.

Code Example

MLOps

import mlflow
from mlflow.tracking import MlflowClient

client = MlflowClient()

# Promote model version 3 to Production
client.transition_model_version_stage(name='my-model', version=3, stage='Production')
print("Model version 3 promoted to Production")

Common Mistakes

Trying to promote a model version that does not exist

The command will fail because the version number is invalid or missing

Check available versions with 'mlflow models list-versions' before promoting

Promoting a model without testing it first

This can cause your app to use a bad model, leading to wrong predictions

Always test models in Staging before moving them to Production

Not archiving old model versions after promotion

Old versions may confuse the system or cause accidental use

Use the Archived stage to mark old versions as inactive

Summary

Use 'mlflow models transition-stage' to move a model version between stages like Staging, Production, or Archived.

Check model versions and their stages with 'mlflow models list-versions' to confirm promotions.

Always test models before promoting and archive old versions to keep your system clean.