Third-Party Integrations

Using Integrations

Ludwig provides an extendable interface to integrate with third-party systems. To activate a particular integration, simply insert its flag into the command line. Each integration may have specific requirements and use cases.

Available integrations

MLFlow

--mlflow - logs training metrics, hyperopt parameters, output artifacts, and trained models to MLflow. Set the environment variable MLFLOW_TRACKING_URI to log results to a remote tracking server.

Weights & Biases

--wandb - logs training metrics, configuration parameters, environment details, and trained model to Weights & Biases. For more details, refer to W&B Quickstart.

Comet ML

--comet - logs training metrics, environment details, test results, visualizations, and more to Comet.ML. Requires a free account. For more details, see Comet's Running Ludwig with Comet.

Aim

--aim - complete experimentation trackings with configuration, metadata, hyperparameters, losses and terminal logs. In order to see and end to end Aim-Ludwig training and tracking example please refer to our demo. For more details about Aim refer to the documentation.

Add more integrations

For more information about integration contributions, please see the Developer Guide.