https://github.com/dk56/snare

SNARE: Scorebased Neural Architecture REduction
https://github.com/dk56/snare

deep-learning keras neural-networks python tensorflow

Last synced: 26 days ago
JSON representation

SNARE: Scorebased Neural Architecture REduction

• Host: GitHub
• URL: https://github.com/dk56/snare
• Owner: DK56
• License: mit
• Created: 2019-09-30T08:45:44.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2023-03-24T22:24:04.000Z (over 1 year ago)
• Last Synced: 2024-09-30T10:03:18.213Z (about 1 month ago)
• Topics: deep-learning, keras, neural-networks, python, tensorflow
• Language: Python
• Homepage:
• Size: 69.3 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# SNARE

[![PyPI version](https://badge.fury.io/py/snare-ml.svg)](https://badge.fury.io/py/snare-ml) [![MIT license](https://img.shields.io/badge/License-MIT-blue.svg)](https://lbesson.mit-license.org/)

SNARE is a new optimization approach to reduce trained neural networks in size. This implementation uses a sequential Keras model with TensorFlow backend as an input and iteratively performs score-based pruning and re-training. SNARE ouputs a smaller Keras-compatible DNN that is optimized to achieve a similar accuracy as the original network. 

## Installation

SNARE requires [TensorFlow] v1.15.2 to run. To install SNARE and all dependencies use:

[TensorFlow]: 

```sh

$ pip3 install snare-ml

```

## License

MIT