Experiments with MLflow

Machine Learning requires experimenting with a wide range of datasets, data preparation steps, and algorithms to build a model that maximizes a target metric. Once you have built a model, you also need to deploy it to a production system, monitor its performance, and continuously retrain it on new data and compare it with alternative models.

CML lets you train, reuse, and deploy models with any library, and package them into reproducible artifacts that other data scientists can use.

CML packages the ML models in a reusable, reproducible form so you can share it with other data scientists or transfer it to production.

CML is compatible with the MLflow™ tracking API and makes use of the MLflow client library as the default method to log experiments. Existing projects with existing experiments are still available and usable.

The functionality described in this document is for the new version of the Experiments feature, which replaces an older version of the Experiments feature that could not be used from within Sessions. In Projects that have existing Experiments created using the previous feature, you can continue to view these existing Experiments. New projects use the new Experiments feature.