The new version of Red Hat OpenShift AI will be available in the coming days. It is designed to better support enterprises with their AI workloads.
This is done via a technology preview of a model registry, a central place for managing models. Better organizing models will make it easier to share, deploy, and track them. Multiple model registries can also be deployed.
Red Hat OpenShift AI 2.15 additionally introduces tools for detecting bias. This platform targets data scientists and AI engineers, who can better address increasing confidence by monitoring a model’s bias. The platform provides insight into how a model performs based on training data and how fairly it acts in production. Red Hat is partnering with TrustyAI, a community with a toolkit for developing responsible AI for these tools.
Support for Nvidia and AMD
Red Hat also supports Nvidia NIM, a set of microservices for accelerated delivery of generative AI applications. NIM supports many models for scalable inference on-premises or in the cloud via APIs. On the hardware front, there is also a new partnership with AMD, allowing users to leverage AMD GPUs for model development. As part of the partnership, there is access to images for training.