MadryLab/trak

A fast, effective data attribution method for neural networks in PyTorch

45
/ 100
Emerging

This tool helps machine learning engineers and researchers understand which specific training examples most influenced a neural network's predictions. You input a trained PyTorch neural network and its training dataset, and it outputs scores indicating the positive or negative impact of each training example on specific target predictions. This allows you to identify problematic training data, debug unexpected model behavior, or improve dataset quality.

232 stars. No commits in the last 6 months.

Use this if you need to quickly and efficiently understand the impact of individual training data points on your PyTorch model's predictions, especially when dealing with large datasets.

Not ideal if you are working with models outside of PyTorch or if your primary goal is not data attribution for debugging and understanding model behavior.

machine-learning-debugging dataset-curation model-interpretability deep-learning-analysis model-auditing
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 19 / 25

How are scores calculated?

Stars

232

Forks

37

Language

Python

License

MIT

Last pushed

Nov 18, 2024

Commits (30d)

0

Get this data via API

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

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