AI Blocks › Build and ship software
LitGPT model trainer
Pick a small open model and your data. Get it downloaded, trained on your examples and ready to chat with.
LitGPT model trainer is a free AI skill for the job "Build and ship software". Give the instructions to an AI assistant that supports the required files and tools, then review its output.
Free · No account · MIT licence · Made by Orchestra Research · Level: advanced
Made by Orchestra Research. See the original on GitHub; we copied it as it is, with its licence.
What it does
A small open model, downloaded, trained on your examples and ready to chat with, without writing training code from scratch. You pick the model and data and get working commands. It uses LitGPT, a clean library with more than 20 model designs, including Llama, Gemma, Phi, Qwen and Mistral. For one graphics card it suggests LoRA, a light method that fits in about 16 GB, and small models such as Phi-2 that use 8 to 12 GB. Full tuning of a 7B model needs 40 GB or more. It also covers pretraining from scratch and viewing logs. You need Python and a GPU.
You give it
- The model you want to start from
- Your training examples
- Your GPU memory
- Your goal for the model
You get
- Download commands for the model
- A tuning command sized to your GPU
- Settings for LoRA or full tuning
- Steps to chat with the result
- How to watch training logs
Use it when
- You want a small model that answers in your tone or format.
- You have a single GPU and need a method that fits.
- You want to learn how model designs differ by training one.
- You have a set of question and answer examples ready.
How to use it
- Add it to your AI once: download it and add the folder to Claude, ChatGPT or Codex as a skill.
- Ask for the job in your own words. Your AI opens the skill when it fits.
- It works best when your AI can run Python code on a computer with a GPU.