danielmamay/grokking

Implementation of OpenAI's 'Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets' paper.

42
/ 100
Emerging

This project helps machine learning researchers explore the 'grokking' phenomenon, where models generalize to unseen data long after overfitting on the training data. You can feed in small algorithmic datasets and observe how different model architectures and training parameters affect this delayed generalization. It's designed for researchers focused on understanding the fundamental learning dynamics of neural networks, particularly in areas like interpretability and generalization theory.

No commits in the last 6 months.

Use this if you are a machine learning researcher studying how neural networks achieve generalization, especially on small, synthetic datasets, and want to experiment with factors influencing the 'grokking' effect.

Not ideal if you are looking to apply machine learning to real-world, large-scale datasets or build production-ready applications, as this tool is for fundamental research into model behavior.

machine-learning-research neural-network-generalization algorithmic-datasets model-interpretability deep-learning-theory
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 8 / 25
Maturity 16 / 25
Community 18 / 25

How are scores calculated?

Stars

42

Forks

12

Language

Python

License

MIT

Last pushed

Sep 23, 2023

Commits (30d)

0

Get this data via API

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

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