Finetuna Keep the context on the GPU.

Hello from finetuna.net

finetuna.net is the public home for Finetuna: install notes, short posts, and a place that is not the GitHub README.

What the CLI already does:

  • GPU-fit via /api/ps (size_vram / size) — not a VRAM guess
  • Named models you can ollama run forever (Modelfile-finetuna + ollama create)
  • --check / --verify with no create side effects
  • Auto-tune that only switches when the win beats measured noise (median + spread)
  • Remote Ollama via OLLAMA_HOST; companion discovery/chat on the network is ollanet

Install with npm i -g finetuna, or see /install. npm 1.1.1 ships with provenance from GitHub Actions.

This site is built with FilePress — git-native Markdown, static HTML, Cloudflare Pages.