Canonical Charmed Kubeflow delivered on NVIDIA DGX systems Canonical and NVIDIA working together to help enterprises benefit from AI ML Canonical is the company behind Ubuntu the most popular operating system for container cloud and hyperscale computing NVIDIA leverages Ubuntu as the basis for the NVIDIA DGX base operating system NGC containers and more With a focus on AI Canonical s Charmed Kubeflow is certified as part of the DGX Ready Software program and has been tested on both single node and multi node clusters Charmed Kubeflow can be run on Canonical Kubernetes and MicroK8s which provide a hardened conformant multi cloud Kubernetes with full lifecycle automation Just like Charmed Kubeflow Canonical Kubernetes and MicroK8s are certified as part of the DGX Ready Software program contributing to a seamless experience for the AI practitioner Charmed Kubeflow A secure MLOps platform to take machine learning models to production Quick deployment Run the entire ML lifecycle Composable architectures Reproducibility portability scalability Canonical s Charmed Kubeflow provides an open source secure end to end MLOps platform that automates machine learning workflows It runs on any Kubernetes and on any cloud being the base of an open source MLOps ecosystem Charmed Kubeflow is running on NVIDIA DGX optimised Ubuntu host images and provides unparalleled integrations and operations on any compute environment Combined with NVIDIA AI Enterprise an end to end suite of AI software included with DGX systems Charmed Kubeflow and DGX provide the tooling AI experts need to train models at scale Charmed Kubeflow translates steps in the data science workflow into Kubernetes jobs It is one of the official distributions of the Kubeflow upstream project Charmed Kubeflow can be used for different purposes from experimentation using Notebooks to training using the Kubeflow Pipelines or tuning using Katib Its main goal is to automate ML workloads encouraging reproducibility and allowing bigger teams to work efficiently Machine learning work ow within a single tool Multi cloud and hybrid cloud compatibility Integrations with external tooling Single node and multi node clusters Optimised software stack Unmatched AI leadership Scalable AI clusters Access to AI expertise Advantages of Charmed Kubeflow on NVIDIA DGX With NVIDIA AI Enterprise and NVIDIA DGX Charmed Kubeflow improves the performance of AI workflows by utilising the hardware to its maximum extent and accelerating project delivery Charmed Kubeflow can significantly speed up the training phase of AI projects especially when coupled with DGX systems With a wide range of integrations Charmed Kubeflow and NVIDIA DGX bring the MLOps practice to a new level making the scaling process easier and seamless Charmed Kubeflow runs on top of any Kubernetes However having it deployed on Canonical s Charmed Kubernetes and MicroK8s offers compelling integrations and additional features such as resource scheduling Both products are certified as part of the DGX Ready Software program as well With model driven operations single node and high availability multi node you get a complete stack for AI projects that consists of NVIDIA DGX systems Ubuntu OS Charmed Kubernetes and Charmed Kubeflow From experimentation to production using one tool Security patching upgrades and updates User management Runs on supported versions of Kubernetes Workload ML Labs Healthcare AI HPC Charmed Kubernetes Containers Security Lifecycle Management Signed Drivers Fast Upgrades WSL Support Ubuntu Certi ed Hardware Platform NVIDIA DGX Enterprise security and support Day 0 support on the latest upstream Runs on any Kubernetes Security maintenance enterprise support Lifecycle Management Tested on AI specialised hardware up to date Canonical s AI ML advisory Enterprise training and workshops with Cluster lifecycle automation with Juju Open Kubeflow versions and managed Kubeflow with the latest drivers Read more about the benefits of using compatible hardware and software for AI Get started with Charmed Kubeflow on DGX systems Contact us to speed up your AI journey with NVIDIA and Canonical For more information about Charmed Kubeflow please visit charmed kubeflow io To learn about the NVIDIA DGX Ready Software program please visit nvidia com dgx mlops To explore more about Kubeflow on DGX systems please check the blog Canonical Limited 2023 Ubuntu Kubuntu Canonical and their associated logos are the registered trademarks of Canonical Limited All other trademarks are the properties of their respective owners Canonical Ubuntu
View Full Document
Unlocking...