We are happy to announce the availability of MLflow 1.5.0!

In addition to bug and documentation fixes, MLflow 1.5.0 includes the following major features and improvements:

• New support for a LightGBM flavor.
• New support for a XGBoost flavor.
• New support for a Gluon flavor and autologging.
• Runs automatically created by mlflow.tensorflow.autolog() and mlflow.keras.autolog() are now automatically ended after training and/or exporting your model. See the docs for more details.

For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.

We are happy to announce the availability of MLflow 1.4.0!

In addition to bug and documentation fixes, MLflow 1.4.0 includes the following major features and improvements:

• Model Registry (Beta). MLflow 1.4.0 adds an experimental model registry feature, where you can manage, version, and keep lineage of your production models.
• TensorFlow updates
  • MLflow Keras model saving, loading, and logging has been updated to be compatible with TensorFlow 2.0.
  • Autologging for tf.estimator and tf.keras models has been updated to be compatible with TensorFlow 2.0. The same functionalities of autologging in TensorFlow 1.x are available in TensorFlow 2.0, namely when fitting tf.keras models and when exporting saved tf.estimator models.
  • Examples and READMEs for both TensorFlow 1.X and TensorFlow 2.0 have been added to mlflow/examples/tensorflow.

For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.

We are happy to announce the availability of MLflow 1.3.0!

In addition to several bug and documentation fixes, MLflow 1.3.0 includes the following major features and improvements:

• The Python client now supports logging & loading models using TensorFlow 2.0
• Significant performance improvements when fetching runs and experiments in MLflow servers that use SQL database-backed storage
• New GetExperimentByName REST API endpoint, used in the Python client to speed up set_experiment and get_experiment_by_name
• New mlflow.delete_run, mlflow.delete_experiment fluent APIs in the Python client
• New CLI command (mlflow experiments csv) to export runs of an experiment into a CSV
• Directories can now be logged as artifacts via mlflow.log_artifact in the Python fluent API
• HTML and geojson artifacts are now rendered in the run UI
• Keras autologging support for fit_generator Keras API
• MLflow models packaged as docker containers can be executed via Google Cloud Run
• Artifact storage configurations are propagated to containers when executing docker-based MLflow projects locally
• The Python, Java, R clients and UI now retry HTTP requests on 429 (Too Many Requests) errors

For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.

We are happy to announce the availability of MLflow 1.2.0!

In addition to several bug and documentation fixes, MLflow 1.2.0 includes the following major features and improvements:

• Experiments now have editable tags and descriptions
• Search latency has been significantly reduced in the SQLAlchemyStore

For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.

We are happy to announce the availability of MLflow 1.1.0!

MLflow 1.1.0 introduces several major features:

• Automatic logging from TensorFlow and Keras
• Parallel coordinate plots in the tracking UI
• Pandas DataFrame based search API
• Java Fluent API
• Kubernetes execution backend for MLflow projects
• Search Pagination

For a comprehensive list of features, see the release change log, and check out the latest documentation on mlflow.org.

We are happy to announce the availability of MLflow 0.9.1!

MLflow 0.9.1 is a patch release on top of 0.9.0 containing mostly bug fixes and internal improvements. We have also included a one breaking API change in preparation for additions in MLflow 1.0 and later. This release also includes significant improvements to the Search API. Please visit the release change log to read more about the fixes and updates in this release.

We are happy to announce the availability of MLflow 0.9.0!

MLflow 0.9.0 introduces several major features:

• Support for running MLflow Projects in Docker containers.
• Database stores for the MLflow Tracking Server.
• Simplified custom Python model packaging.
• Plugin systems allowing third party libraries to extend MLflow functionality.
• Support for HTTP authentication to the Tracking Server in the R client.

And a few breaking changes:

• [Scoring] The pyfunc scoring server now expects requests with the application/json content type to contain json-serialized pandas dataframes in the split format, rather than the records format. Also, when reading the pandas dataframes from JSON, the scoring server no longer automatically infers data types as it can result in unintentional conversion of data types.
• [API] Removed GetMetric & GetParam from the REST API as they are subsumed by GetRun.

For a comprehensive list of features, see the release change log, and check out the latest documentation on mlflow.org.

We are happy to announce the availability of MLflow 0.8.2!

MLflow 0.8.2 is a patch release on top of 0.8.1 containing bug fixes and documentation updates. Please visit the change log to read more about the fixes and updates introduced in this release.

We are happy to announce the availability of MLflow 0.8.1! MLflow 0.8.1 introduces several significant improvements:

• Improved UI responsiveness and load time, especially when displaying experiments containing hundreds to thousands of runs.

• Improved visualizations, including interactive scatter plots for MLflow run comparisons.

• Expanded support for scoring Python models as Spark UDFs. For more information, see the updated documentation for this feature.

• By default, saved models will now include a Conda environment specifying all of the dependencies necessary for loading them in a new environment.

• MLflow projects can now be run from ZIP files.

The release includes additional bugfixes and improvements across the Python client, tracking UI, and documentation. Visit the change log to read more about the new features.

We are happy to announce the availability of MLflow 0.8.0! MLflow 0.8.0 introduces several major features:

• Dramatically improved UI for comparing experiment run results:

  • Metrics and parameters are by default grouped into a single column, to avoid an explosion of mostly-empty columns. Individual metrics and parameters can be moved into their own column to help compare across rows.
  • Runs that are "nested" inside other runs (e.g., as part of a hyperparameter search or multistep workflow) now show up grouped by their parent run, and can be expanded or collapsed altogether. Runs can be nested by calling mlflow.start_run or mlflow.run while already within a run.
  • Run names (as opposed to automatically generated run UUIDs) now show up instead of the run ID, making comparing runs in graphs easier.
  • The state of the run results table, including filters, sorting, and expanded rows, is persisted in browser local storage, making it easier to go back and forth between an individual run view and the table.

• Support for deploying models as Docker containers directly to Azure Machine Learning Service Workspace (as opposed to the previously-recommended solution of Azure ML Workbench).

The release also includes bugfixes and improvements across the Python and Java clients, tracking UI, and documentation. Visit the change log to read more about the new features.