Package for the design, training, pruning and verification of neural networks.
Project description
pyNeVer
Neural networks Verifier (NeVer 2) is a tool for the design, training and verification of neural networks. It supports sequential fully connected and convolutional neural networks with ReLU and Sigmoid activation functions. pyNeVer is the corresponding python package providing all the main capabilities of NeVer 2 and can be easily installed using pip.
Installation and setup
pyNeVer depends on several packages, which are all available via pip and should be installed automatically. The packages required for the correct execution are the following:
- numpy
- ortools
- onnx
- torch
- torchvision
- pysmt
- multipledispatch
To install pyNeVer, just run the command:
pip install pynever
To run some examples, further packages may be required. If an example requires a specific package, it will be detailed in the example directory.
Supported inputs
At present the pyNeVer package supports only the abstraction and verification of fully connected and convolutional
neural networks with ReLU and Sigmoid activation functions. The training and conversion supports also batch normalization
layers. A network with batchnorm layers following fully connected layers can be converted to a "pure" fully connected
neural networks using the capabilities provided in the utilities.py module.
The conversion package provides the capabilities for the conversion of PyTorch and ONNX
networks: therefore this kind of networks can be loaded using the respective frameworks and then converted to the
internal representation used by pyNeVer.
The properties for the verification and abstraction of the networks must be defined either in python code following
the specification which can be found in the documentation, or via an SMT-LIB file compliant to the
VNN-LIB standard.
Examples
NB: All the scripts should be executed INSIDE the related directory!
All the examples described below are guaranteed to work until Release v0.1.1a4. After this release, changes in the interface structure may add inconsistencies between test scripts and API, so the old examples will be removed and new examples will be created in future releases.
-
The directory examples/ contains some examples of application of the pyNeVer package. In particular the jupyter notebook shows a graphical example of the application of the abstraction module for the reachability of a small network with bi-dimensional input and outputs.
-
The pruning_example.py script show how to train and prune some small fully connected neural networks with relu activation function. It also show how it is possible to combine batch norm layer and fully connected layers to make the networks compliant with the requirements of the verification and abstraction modules.
-
The directory examples/submissions/ATVA2021 contains the experimental setup used for the experimental evaluation in our ATVA2021 paper. The experiments can be easily replicated by executing the python scripts acas_experiment.py from within the ATVA2021/ directory. The log files will be generated and will be saved in the logs/ directory.
Contributors
The main contributors of pyNeVer are Dario Guidotti and Stefano Demarchi, under the supervision of Professors
Armando Tacchella and Luca Pulina.
A significant contribution for the participation in VNN-COMP 2024 was
the help of Elena Botoeva.
Other contributors:
- Andrea Gimelli - Bound propagation integration
- Pedro Henrique Simão Achete - Command-line interface and convolutional linearization
- Karim Pedemonte - Design and refactoring
Publications
If you use NeVer2 or pyNeVer in your work, please kindly cite our papers. Here you can find the list of BibTeX entries.
@article{demarchi2024never2,
title={NeVer2: Learning and Verification of Neural Networks},
author={Demarchi, Stefano and Guidotti, Dario and Pulina, Luca and Tacchella, Armando},
journal={Soft Computing},
year={2024}
}
@inproceedings{demarchi2022formal,
title={Formal Verification Of Neural Networks: A Case Study About Adaptive Cruise Control.},
author={Demarchi, Stefano and Guidotti, Dario and Pitto, Andrea and Tacchella, Armando},
booktitle={ECMS},
pages={310--316},
year={2022}
}
@inproceedings{guidotti2021pynever,
title={pynever: A framework for learning and verification of neural networks},
author={Guidotti, Dario and Pulina, Luca and Tacchella, Armando},
booktitle={Automated Technology for Verification and Analysis: 19th International Symposium, ATVA 2021, Gold Coast, QLD, Australia, October 18--22, 2021, Proceedings 19},
pages={357--363},
year={2021},
organization={Springer}
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file pynever-1.0.0.tar.gz
.
File metadata
- Download URL: pynever-1.0.0.tar.gz
- Upload date:
- Size: 89.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.9.19
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 3a532064f06601312a1f9c8fc25f17dfb5702ebc252543fa144cee47272e7d33 |
|
MD5 | f1353436b0cdfd4b990f85542f26683c |
|
BLAKE2b-256 | 91c1277101b170bd120d6919072fa1aa08152975dff2d525281bc1fc2319963f |
File details
Details for the file pyNeVer-1.0.0-py3-none-any.whl
.
File metadata
- Download URL: pyNeVer-1.0.0-py3-none-any.whl
- Upload date:
- Size: 103.9 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.9.19
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 480d935960126b95201f79067ed58fddff61647ff5d417a0006e2be2f8130757 |
|
MD5 | 788b90d2332164084afe139e760ce3b2 |
|
BLAKE2b-256 | 58d536780aab40aa032a6a7fe925f5b021add27d9a7b6c272c0073b9c9327cf1 |