It is a part of the paper 'Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding' by Song Han, Huizi Mao, William J. Dally
Following packages are required for this project
- Python3.6+
- tqdm
- numpy
- pytorch, torchvision
- scipy
- scikit-learn
You can control other values such as
- random seed
- epochs
- sensitivity
- batch size
- learning rate
- and others