Have you heard the news? Charmed Kubeflow 1.8 is available in Beta. Kubeflow is the foundation of Canonical MLOps. The latest release brings improved capabilities to personalise different components of the platform, including the images that can be used in Notebooks.
We are looking for data scientists, machine learning engineers, creators and AI enthusiasts to take Charmed Kubeflow 1.8 Beta for a test drive and share their feedback with us.
What’s new in Kubeflow 1.8?
- New hyperparameter tuning algorithms
- Improved security planning for future releases
- A resource scheduler plugin
With these new features, Kubeflow is in an even stronger position to help organisations optimise their models for different use cases including GenAI, large language models (LLMs) or predictive analytics.
Besides all the features introduced in the upstream project’s version, Canonical’s Charmed Kubeflow offers dynamic enablement of the sidebar from the dashboard, as well as the opportunity to add new images in Notebooks. Additionally, Charmed Kubeflow can now be deployed in fully air-gapped environments.
Dynamic Notebook image selector
Depending on the use case, each user might need different Notebook images. Whereas Tensorflow and Pytorch are by default in the bundle, niche industries or applications would get better results if they could use their own images. This increases model performance, as well as the chance to move beyond experimentation.
In the machine learning operations (MLOps) space, there is a fast-growing number of Notebook images specialised in different activities. For example, NVIDIA Nemo is specialised in GenAI and it enables a certain category of users. Giving users the freedom to add their own images allows them to focus on the use case rather than the tooling used. However, not everyone needs all of them and adding a huge list of images by default in the platform would require too many resources.
Charmed Kubeflow 1.8 has the capability to add any Notebook image and benefit from its capabilities. This enables data scientists and machine learning engineers to focus on building machine learning models that could reach their best performance, rather than worry about tools.
Run Kubeflow offline
Depending on your company policy, computing power available and various security and compliance restrictions, you may prefer running machine learning workflows in different environments. Especially in highly regulated industries, organisations often have a need to run in air-gapped environments.
To address this, companies look for MLOps platforms that can be deployed and then work offline, on different clouds. This should allow them to complete most of the machine learning workflow within one tool, to avoid even more time spent on connecting more of those dots.
Charmed Kubeflow is an end-to-end MLOps platform that runs on Kubernetes and allows professionals to develop and deploy machine learning models. Once data is ingested, all activities such as training, automation, model monitoring and model serving can be performed inside the tool. From its initial design, Charmed Kubeflow could run on any cloud platform and has the ability to support various scenarios, including hybrid-cloud and multi-cloud scenarios. Charmed Kubeflow is validated and constantly tested to ensure that all capabilities work in offline environments.
Join us live: tech talk on Charmed Kubeflow 1.8
Tomorrow, Oct 5, 2023, 5pm GMT, Canonical will host a live stream about Charmed Kubeflow 1.8 Beta. Together with Noha Ihab, we will continue the tradition that started with the previous releases. We will answer your questions and talk about:
- The latest release: Kubeflow 1.8 and how our distribution handles it
- Key features covered in Charmed Kubeflow 1.8
- The differences between the upstream release and Canonical’s Charmed Kubeflow
Charmed Kubeflow 1.8 Beta: try it now
Are you already a Charmed Kubeflow user?
If you are already familiar with Charmed Kubeflow, you will only have to upgrade to the latest version. We already prepared a guide, with all the steps you need to take.
Please be mindful that this is not a stable version, so there is always a risk that something might go wrong. Save your work to proceed with caution. If you encounter any difficulties, Canonical’s MLOps team is here to hear your feedback and help you out. Since this is a Beta version, Canonical does not recommend running or upgrading it on any production environment.
Are you new to Charmed Kubeflow?
You are a real adventurer, you can go ahead and start directly with the beta version. This might result in a few more challenges for you. For all the prerequisites, follow the tutorial and please check out the section “Get started”.
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 instructions below to get this up and running:
juju deploy kubeflow –channel 1.7/beta –trust
Don’t be shy. Share your feedback.
Charmed Kubeflow is an open source project that grows because of the care, time and feedback that our community gives. The latest release in beta is no exception, so if you have any feedback or questions about Charmed Kubeflow 1.8, please don’t hesitate to let us know.
Meet us at Canonical AI Roadshow
Feel like talking to our team in person and sharing your feedback? Meet us at the Canonical AI Roadshow. Access key dates on the event webpage.
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.
The Kubeflow community is meeting again this year to celebrate their success, learn from their users and discuss some of the challenges of the open source...
The AI Roadshow lands in the USA Date: 25-26 October 2023 Location: Renaissance Austin Hotel, Austin, Texas Book a meeting The Canonical AI Roadshow will soon...
Series of events will highlight generative AI use cases powered by open source software London, UK. 6 September 2023. Canonical, the publisher of Ubuntu, is...