AI Blocks › Track your numbers
Model run tracker
Point it at your PyMC models. Get every run saved with its settings and checks, so you can compare them later.
Model run tracker is a free AI skill for the job "Track your numbers". 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 PyMC Labs · Level: expert
Made by PyMC Labs. See the original on GitHub; we copied it as it is, with its licence.
What it does
Every PyMC model fit saved with its settings, checks and results, so you can compare runs and bring any one back later. The skill ties PyMC to MLflow. It sets up a named experiment and logs your data source, package versions, random seed and sampler settings before the slow fitting starts. Runs sit inside a tracked block, so failures show up as failed runs. After the fit it saves the full results file and the quality checks, and labels runs as exploring, reportable or mock. Guides cover plain PyMC, PyMC-Marketing auto-logging and saving marketing mix models. It needs Python.
You give it
- Your PyMC model code
- Where to store runs
- Which runs are for exploring or for reporting
- PyMC-Marketing details if you use it (optional)
You get
- Code that logs each fit
- Saved results and checks per run
- Run labels for tracking status
- A way to compare runs
- Steps to restore a saved model
Use it when
- You have fit the same Bayesian model six times with small changes.
- A report needs the exact settings behind one result.
- You want to save and reload a marketing mix model.
- A failed fit left you unsure which run was the last good one.
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 and have MLflow installed.