Mini-project
IT, Software & SaaS
Small model that writes commit messages
Fine-tune a small open model to write commit messages from code changes.
The brief
Developers write vague commit messages like "fix" and "update". A small team wants a model that runs on its own machines and writes a good message from a diff. Fine-tune a small open model and compare it with a large hosted model on quality, speed and cost. Good looks like: a clean training set, a fair evaluation, and a recommendation backed by numbers.
What you will build
- Prepare diff and message pairs and a held-out test set
- Fine-tune a small open model with LoRA
- Generate messages with the small and a large model
- Score quality with human ratings and an LLM judge
- Compare quality, speed and cost and recommend one
Free, open data to build it with
- CommitPackFT ↗ Hugging Face (BigCode)
Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.