https://github.com/pytorchfi/pytorchfi
A runtime fault injection tool for PyTorch :fire:
https://github.com/pytorchfi/pytorchfi
Last synced: 2 months ago
JSON representation
A runtime fault injection tool for PyTorch :fire:
- Host: GitHub
- URL: https://github.com/pytorchfi/pytorchfi
- Owner: pytorchfi
- License: other
- Created: 2019-04-30T05:52:04.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2024-07-25T10:17:34.000Z (9 months ago)
- Last Synced: 2025-02-09T15:22:27.211Z (3 months ago)
- Language: Python
- Homepage: https://pytorchfi.dev
- Size: 17.5 MB
- Stars: 108
- Watchers: 5
- Forks: 24
- Open Issues: 15
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-safety-critical-ai - `pytorchfi/pytorchfi`
- awesome-safety-critical-ai - `pytorchfi/pytorchfi`
README
PyTorchFI
Background •
Usage •
Code •
Contributors •
Citation •
License
## Background
PyTorchFI is a runtime perturbation tool for deep neural networks (DNNs), implemented for the popular PyTorch deep learning platform. PyTorchFI enables users to perform perturbation on weights or neurons of a DNN during runtime. It is extremely versatile for dependability and reliability research, with applications including resiliency analysis of classification networks, resiliency analysis of object detection networks, analysis of models robust to adversarial attacks, training resilient models, and for DNN interpertability.
An example of a use case for PyTorchFI is to simulate an error by performaing a fault-injection on an object recognition model.
| Golden Output | Output with Fault Injection |
| :------------------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------: |
|  |  |
## Usage
Download on PyPI [here](https://pypi.org/project/pytorchfi/), or take a look at our documentation at [pytorchfi.dev](https://pytorchfi.dev/).
You can also learn more with our [interactive demo](https://colab.research.google.com/drive/1BMB4LbsTU_K_YXUFzRyfIynpGu5Yhr1Y).
### Installing
**From Pip**
Install using `pip install pytorchfi`
**From Source**
Download this repository into your project folder.
### Importing
Import the entire package:
```python
import pytorchfi
```
Import a specific module:
```python
from pytorchfi import core
```
### Testing
```bash
pytest
```
## Code
### Structure
The main source code of PyTorchFI is held in `pytorchfi`, which carries both `core` and `error_models.py` implementations.
### Formatting
All python code is formatted with [black](https://black.readthedocs.io/en/stable/).
## Contributors
Before contributing, please refer to our [contributing guidelines](https://github.com/pytorchfi/pytorchfi/blob/master/CONTRIBUTING.md).
- [Sarita V. Adve](http://sadve.cs.illinois.edu/) (UIUC)
- [Neeraj Aggarwal](https://neerajaggarwal.com) (UIUC)
- [Christopher W. Fletcher](http://cwfletcher.net/) (UIUC)
- [Siva Kumar Sastry Hari](https://research.nvidia.com/person/siva-hari) (NVIDIA)
- [Abdulrahman Mahmoud](http://amahmou2.web.engr.illinois.edu/) (UIUC)
- [Alex Nobbe](https://github.com/Alexn99) (UIUC)
- [Jose Rodrigo Sanchez Vicarte](https://jose-sv.github.io/) (UIUC)
## Citation
View the [published paper](http://rsim.cs.illinois.edu/Pubs/20-DSML-PyTorchFI.pdf). If you use or reference PyTorchFI, please cite:
```
@INPROCEEDINGS{PytorchFIMahmoudAggarwalDSML20,
author={A. {Mahmoud} and N. {Aggarwal} and A. {Nobbe} and J. R. S. {Vicarte} and S. V. {Adve} and C. W. {Fletcher} and I. {Frosio} and S. K. S. {Hari}},
booktitle={2020 50th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W)},
title={PyTorchFI: A Runtime Perturbation Tool for DNNs},
year={2020},
pages={25-31},
}
```
## Funding Sources
This project was funded in part by the [Applications Driving Architectures (ADA) Research Center](https://adacenter.org/), a JUMP Center co-sponsored by SRC and DARPA, and in collaboration with NVIDIA Research.
## License
[NCSA](https://opensource.org/licenses/NCSA) License. Copyright © 2021 [RSim Research Group](http://rsim.cs.uiuc.edu/).