Hi folks!
AIKit is a side project of mine that's aimed to be a one-stop shop to easily and quickly get started to build, fine-tune and deploy large language models (LLMs) with only Docker!
I recently added extensible finetuning interface to AIKit. AIKit uses Unsloth under-the-hood for fast and efficent fine tuning of a model.
After finetuning, you can take the finetuned model and create a minimal container image for inference, which is OpenAI compatible powered by LocalAI, to create a seamless end-to-end experience.
I would love any thoughts and feedback!
Get started: https://sozercan.github.io/aikit
Demo of finetuning Mistral with OpenHermes and then serving the model as a minimal container image: https://youtu.be/FZuVb-9i-94