We are happy to announce that Charmed Kubeflow 1.6 is now available in Beta. Kubeflow has evolved into an end-to-end MLOps platform for optimised complex model training. We’re looking for data scientists, ML engineers and developers to take the Beta release for a drive and share their feedback! Read on to learn more.
What is new in Kubeflow 1.6?
Kubeflow 1.6 is the latest version of the upstream project that will be released in the following weeks. There were many improvements on the roadmap; the enhancements on the training operator are one of the highlights. In general, it takes up to 15 iterations to take 50% of models into production. These enhancements can accelerate the entire process considerably for users who use Kubeflow as their MLOps platform.
Optimised AI training and modelling
Large volumes of data present a big challenge for various industries; finding ways to train more models efficiently is difficult. The MPI Operator was available in Alpha with limited support. The new version of Kubeflow expands its scope, making allreduce-style distributed training on Kubernetes easy to run. Unlike others, MPI Operator is decoupled, being compatible with other frameworks such as Tensorflow, PyTorch or Horovod.
Training operators were an important part of the Kubeflow 1.5 release, but there is much more to be done to improve model accuracy and lower infrastructure costs. PyTorch Elastic enhancements are still on the roadmap, aiming to improve the dynamicity of Kubeflow and allow users to get started faster on training models.
Are you from the energy sector and struggle with large volumes of data? Join us and learn more how to deal with them better.
Join us live: tech talk on Charmed Kubeflow 1.6
Tomorrow, 18 August 2022, at 5 PM GMT, Canonical will offer a livestream about Charmed Kubeflow 1.6 beta. Together with Dominik Fleischmann and Daniela Plascencia, our resident Charmed Kubeflow developers, we will answer your questions and talk about:
- The new release
- The importance of being part of an open source community
- The challenges involved in developing an open-source product
- The differences between the upstream release and Canonical’s Charmed Kubeflow
Deploy Charmed Kubeflow 1.6 Beta
Charmed Kubeflow 1.6 beta is driven by Juju, an enterprise Operator Lifecycle Manager (OLM) that provides model-driven application management and next-generation infrastructure-as-code.
Are you already familiar with Charmed Kubeflow?
If you are an old friend of Charmed Kubeflow, then your deployment process will be much quicker. However, you need to remove the existing version following the instructions from the uninstall tutorial.
juju deploy kubeflow --channel 1.6/beta --trust
Once you are all set, only one command is left, and the latest version of Kubeflow is deployed on your machine.
The bundle is available on CharmHub, where you can see all the released versions of Canonical’s Charmed Kubeflow.
Are you new to Charmed Kubeflow?
Before getting started on the latest version, you will have to follow a few steps from the quick start guide to Kubeflow. Check out the section called “Deploy Kubeflow”.
Shortly after you deploy and install MicroK8s and Juju, you will need to add the Kubeflow model and then make sure you have the latest version. Follow the instruction below to get this up and running:
juju deploy kubeflow --channel 1.6/beta --trust
The stable version will be released soon, so please report any bugs or submit your improvement ideas on Discourse. The known issues are also listed there. Your feedback matters to us and we would like to thank you in advance for it!
There is more coming up….
Once the stable version of Charmed Kubeflow 1.6 is released, we will update you and guide you through the deployment process.
Don’t forget to add the livestream from tomorrow to your calendar if you want to ask your questions live.
Learn more about Charmed Kubeflow
With Charmed Kubeflow, deployment and operations of Kubeflow are easy for any scenario.
Charmed Kubeflow is a collection of Python operators that define integration of the apps inside Kubeflow, like katib or pipelines-ui.
Use Kubeflow on-prem, desktop, edge, public cloud and multi-cloud.
Kubeflow makes deployments of Machine Learning workflows on Kubernetes simple, portable and scalable.
Kubeflow is the machine learning toolkit for Kubernetes. It extends Kubernetes ability to run independent and configurable steps, with machine learning specific frameworks and libraries.
The Kubeflow project is dedicated to making deployments of machine learning workflows on Kubernetes simple,
portable and scalable.
You can install Kubeflow on your workstation, local server or public cloud VM. It is easy to install with MicroK8s on any of these environments and can be scaled to high-availability.