Red Hat aims to democratize AI development by launching Red Hat Enterprise Linux AI (RHEL AI). This platform is specifically engineered to be accessible, customizable, and open source for AI development.
So, RHEL AI is not an AI tool, it is a platform for AI development!
Think of RHEL AI as a complete, ready-to-go platform for building and deploying AI. It's specifically designed for businesses that want to use open-source Granite generative AI models in their applications, making the whole process of developing, testing, and actually getting those models running much smoother. What's really clever about it is that it's built on top of the InstructLab open-source project. This means RHEL AI brings together IBM Research's powerful Granite large language models (LLMs) with InstructLab's specialized tools for fine-tuning those models. The best part? It's all wrapped up in a super-optimized, bootable RHEL image, which makes setting it up on your servers a breeze.
What's under the umbrella ??
First off, you've got the Granite Models. Think of these as the foundational large language models (LLMs) – they're completely open source, under an Apache 2.0 license, and were built through a collaboration between IBM Research and the folks over at the InstructLab community. What's really neat is that you get full transparency: access to all the data sources and the model weights. This means you can actually take these base models and train them with your own specialized information, essentially creating custom LLMs tailored to your specific needs. And you get to choose whether you want to share your tweaked models with the wider community or keep them just for your own use.
Then there's InstructLab. This is another open-source project that really simplifies the whole process of customizing LLMs. It uses a unique method called "Large-scale Alignment for ChatBots" (LAB), which is driven by a taxonomy. What this means for you is that even if you're not a seasoned data scientist, you can still fine-tune these models with your own domain-specific knowledge. How? By helping generate high-quality "synthetic data" that the models then train on. InstructLab even comes with its own command-line tools that interact with a local collection of skills and knowledge, making it easier to train and then serve up your custom models.
Underneath it all, RHEL AI comes as an Optimized RHEL Image. This isn't just any old Red Hat image; it's specifically tweaked to handle demanding AI workloads right out of the box. It's built to work seamlessly with popular hardware accelerators from major players like AMD, Intel, and NVIDIA, ensuring it runs efficiently whether you're deploying in public clouds or on specialized AI servers. Plus, it's designed to play nicely with Red Hat Certified Ecosystem hardware and can be deployed on platforms like IBM Cloud without a hitch.
And finally, for that extra peace of mind, RHEL AI subscriptions will include Enterprise Support and Indemnification directly from Red Hat. This means you'll get comprehensive support, a full product lifecycle, and intellectual property indemnification, giving organizations the confidence they need to deploy AI solutions backed by Red Hat's robust support system.
Features & Benefits ?
1. LLMs for the enterprise - Tune smaller, purpose-built models with your own data using RAG.
2. Increased efficiency
3. Cloud-native scalability - Image mode - ring any bells ? Containers ?
4. Community collaboration - well what do you know - of course it is there.
WARNING !! : May contain further awesomeness. Proceed to the deep dive?
- You can get started training open source Granite models by downloading InstructLab cli on your laptop/desktop or jump right to the RHEL AI developer preview.
- https://developers.redhat.com/products/rhel-ai/overview
- https://www.redhat.com/en/topics/ai/what-is-instructlab
- https://www.redhat.com/en/products/ai/inference-server
- https://www.redhat.com/en/about/open-source-assurance
