DetoxAI is a Python package for debiasing neural networks. It provides a simple and efficient way to remove bias from your models while maintaining their performance. The package is designed to be easy to use and integrate into existing projects. For more information about the package, see the website https://detoxai.github.io and documentation https://detoxai.readthedocs.io
Installation
DetoxAI is available on PyPI, and can be installed by running the following command:
pip install detoxai
Quickstart
The snippet below shows the high-level API of DetoxAI and how to use it.
import detoxai
model = ...
dataloader = ... # has to output a tuple of three tensors: (x, y, protected attributes)
corrected = detoxai.debias(model, dataloader)
metrics = corrected["SAVANIAFT"].get_all_metrics() # Get metrics for the model debiased with SavaniAFT method
model = corrected["SAVANIAFT"].get_model()
A shortest snippet that would actually run and shows how to plug DetoxAI into your code is below.
import torch
import torchvision
import detoxai
model = torchvision.models.resnet18(pretrained=True)
model.fc = torch.nn.Linear(model.fc.in_features, 2) # Make it binary classification
X = torch.rand(128, 3, 224, 224)
Y = torch.randint(0, 2, size=(128,))
PA = torch.randint(0, 2, size=(128,))
dataloader = torch.utils.data.DataLoader(list(zip(X, Y, PA)), batch_size=32)
results: dict[str, detoxai.CorrectionResult] = detoxai.debias(model, dataloader)
Too see more examples of detoxai in use, navigate to the github repo https://github.com/DetoxAI/detoxai and see examples/ folder.
Acknowledgment
If you use this library in your work please cite as:
@misc{detoxai2025,
author={Ignacy St\k{e}pka and Lukasz Sztukiewicz and Micha\l{} Wili\'{n}ski and Jerzy Stefanowski},
title={{DetoxAI}: a {Python} Package for Debiasing Neural Networks},
year={2025},
url={https://github.com/DetoxAI/detoxai},
}
License
MIT License
3 maintainers
Metadata
3 maintainers
Release files for detoxai 0.3.8
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| detoxai-0.3.8.tar.gz | 8.3 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| detoxai-0.3.8-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 8.4 MB
Release files / detoxai-0.3.8.tar.gz
| Download URL | detoxai-0.3.8.tar.gz |
|---|---|
| Size | 8.3 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
f8f43877317600cfcca4bfd4392754dc2216726afc3656002f59e6bc719537ec
|
|
BLAKE2b-256 checksum How to use checksums |
34a84215368c1700a9cd2f370b4f9382de775b802293fa874e5d22e103bfeae9
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.7.2
|
Release files / detoxai-0.3.8-py3-none-any.whl
| Download URL | detoxai-0.3.8-py3-none-any.whl |
|---|---|
| Size | 92.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
e555696948460f75dd2dc1b39b6655badf53e68af352c67851579785c45a782d
|
|
BLAKE2b-256 checksum How to use checksums |
6363314a70a853236588f3ee8d9d014336c5a2a9c94f016ccb83352161837ef0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.7.2
|
