Machine Learning Toolkit for Kubernetes “Kubeflow 1.0” Released

March 3, the development team of Kubeflow, a machine learning toolkit for Kubernetes, released Kubeflow 1.0. It offers enhanced stability in the core technology for building and deploying a model.

Kubeflow is a toolkit to deploy machine learning workflow on Kubernetes. It was first released in December 2017 as an open source project. Focusing on simplicity, scalability, and portability, the goal is to run on any Kubernetes environment. One of the notable features is the dynamic link with “Jupyter notebook”, which is used widely in data science field. It also offers TensorFlow model training, allows export in “TensorFlow Serving” format, and introduces “Kuberflow Pipelines”, a comprehensive solution for deploying and managing machine-learning workflow .

In regards to Kybeflow 1.0, included in the stable components for developing, building, training and deploying models on Kybernetes are central dashboard UI, Jupiter notebook controller, TensorFlow Operator (TFJob), PyTorch Operator, command-line interface kfctl, and Profile controller.

Kubeflow 1.0 is available on the project website.

Kubeflow Project
https://www.kubeflow.org/