Welcome KitOps v0.2! This update brings two major features for working with LLMs, as well numerous smaller enhancements. We are excited to introduce KitOps’ Dev mode - making it possible to test LLMs locally (even if you don’t have an internet connection or GPU).
Dive into the world of large language models with our step-by-step tutorial on fine-tuning using LoRA, powered by tools like llama.cpp and KitOps. LoRA (Low-Rank Adaptation) is an efficient technique for adapting pre-trained models, minimizing computational overhead. We'll guide you through setting up your environment, creating a Kitfile, building a LoRA adapter, and deploying your fine-tuned model. By the end, you'll have a packaged model ready for deployment.
In AI projects the biggest (and most solvable) source of friction are the handoffs between data scientists, application developers, testers, and infrastructure engineers as the project moves from development to production. This friction exists at every company size, in every industry, and every vertical. Gartner’s research shows that AI/ML projects are rarely deployed in under 9 months despite the use of ready-to-go large language models (LLMs) like Llama, Mistral, and Falcon.
ModelKits, much like other OCI artifacts, can be identified using tags that are comprehensible to humans. This blog explores various strategies for effectively tagging your ModelKits.
One of the goals of KitOps is streamlining AI integration with your existing DevOps practices. Our latest contribution on this evolving landscape is a GitHub Action that simplifies integrating Kit CLI into your existing CI/CD toolset. This development is particularly exciting for those looking to leverage ModelKits, our OCI-compliant AI project packaging, directly within their CI/CD pipelines.
AI/ML is a wildfire of a trend. It’s being integrated into just about every application you can think of. When compared to other technical innovations over the past years, like the blockchain, the infrastructure and tooling was being built (and still is) in parallel to discovering a meaningful application.
