Included
- Dataset card and model card templates
- License allowlist and source tracker
- Secret scanning and deduplication steps
- QLoRA training and adapter export scripts
Model Upgrade
Fine-Tune Starter Kit
Turn owned code into a custom coding adapter.
A compliance-first LoRA/QLoRA pipeline for preparing code instruction data, scanning licenses/secrets, training adapters, and shipping a local model package.
A compliance-first path from owned code samples to an adapter package.
PythonTransformersPEFTTRLHugging Face
준비중
Customization path
- Bring owned repositories as training data
- Choose a local, enterprise, or laptop-class model path
- Export adapter-only packages for cleaner distribution
Best buyer
Teams that want a model tuned to their own codebase without training from scratch.
Try the local flow
The first buyer experience is a working repo index, an index report, and a cited coding answer.
$ Planned packageExpected demo output
The demo proves that retrieval is wired before the buyer swaps in a real repository.
Preview coming soonPurchase delivery
- Planned package
What changes for the buyer
The demo project gets replaced with the buyer's owned repo, docs, and style guide. The template keeps the same local indexing, report, citation, and review workflow.
Buyer clarity
Before purchase
Clear answers for delivery, customization, and license boundaries.
Does fine-tuning remove the base model license?
No. The starter kit is designed to track model, dataset, and adapter boundaries so the buyer can package upgrades without hiding upstream obligations.
Product proof
Planned package
