from Hacker News

Show HN: Local fine tuning for Mistral and SDXL, GPU mem/latency optimization

by lewq on 12/21/23, 8:13 PM with 3 comments

100% bootstrapped new startup. It lets you fine tune Mistral-7B and SDXL. In particular, for the LLM fine tuning we implemented a dataprep pipeline that turns websites/pdfs/doc files into question-answer pairs for training the small LLM using an big LLM.

It includes a GPU scheduler that can do finegrained GPU memory scheduling (Kubernetes can only do whole-GPU, we do it per-GB of GPU memory to pack both inference and fine tuning jobs into the same fleet) to fit model instances into GPU memory to optimally trade off user facing latency with GPU memory utilization

It's a pretty simple stack of control plane and a fat container that runs anywhere you can get hold of a GPU (e.g. runpod).

Architecture: https://docs.helix.ml/docs/architecture

Demo walkthrough showing runner dashboard: https://docs.helix.ml/docs/overview

Run it yourself: https://docs.helix.ml/docs/controlplane

Discord: https://discord.gg/VJftd844GE

Please roast me!