AI Blocks › Build and ship software
Model settings tuner
Give it a model and a compute budget. Get a tuning plan and the best settings found.
Model settings tuner 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 Burhan Sebin · Level: expert
Made by Burhan Sebin. See the original on GitHub; we copied it as it is, with its licence.
What it does
The best settings for your model, found within the compute you can spend. You get a clear plan for the search, then the winning combination. First it lists every setting to try, with ranges, using log scales for values like learning rate. Then it picks a search method to fit your budget: grid for small spaces, random as a strong default, Bayesian search when each trial is expensive, or Hyperband when you train deep networks and can stop weak trials early. It scores each trial with cross-validation or a holdout set, runs trials in parallel, prunes the poor ones, and reports which settings mattered most.
You give it
- Your model and training code
- The score to improve
- Your compute or time budget
- Settings you already know are fixed (optional)
- Spreadsheet (optional)
You get
- A search space with ranges
- A chosen search method and why
- Search code with early stopping
- The best settings found
- A ranking of which settings mattered
Use it when
- Your model scores fine, and you suspect better settings would add a few points.
- You have a fixed GPU budget and need to spend it well.
- You do not know whether to use grid or random search.
- You want to know which setting matters most.
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 code and read your project files.