oswaldoludwig/kappaTune

A PyTorch-based optimizer wrapper for continual learning via selective fine-tuning, guided by the condition number κ (kappa) of model tensors. KappaTune identifies and updates only the least anisotropic parameters to preserve pre-trained knowledge and mitigate catastrophic forgetting.

36
/ 100
Emerging

This tool helps machine learning engineers and researchers to update large language models (LLMs) without losing previously learned knowledge. It takes a pre-trained PyTorch model and a new dataset, then intelligently selects and fine-tunes only the most stable parts of the model. The result is an updated model that retains its core capabilities while efficiently learning new information.

Use this if you need to continually update your AI models with new data or tasks without 'forgetting' what they already know, especially in scenarios like adapting LLMs to specific domains or evolving information.

Not ideal if you are starting model training from scratch, or if your primary goal is extreme parameter efficiency without specific concern for preserving pre-trained knowledge during sequential learning tasks.

continual-learning large-language-models model-adaptation machine-learning-engineering ai-research
No Package No Dependents
Maintenance 10 / 25
Adoption 5 / 25
Maturity 15 / 25
Community 6 / 25

How are scores calculated?

Stars

12

Forks

1

Language

Python

License

Apache-2.0

Last pushed

Mar 13, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/oswaldoludwig/kappaTune"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.