This repository contains a Pytorch implementation of the paper Multi-prize Lottery Ticket Hypothesis.
pip3 install -r requirements.txt
python3 main.py
--lr
: Learning rate- Default :
1.2e-3
- Default :
--epochs
: Number of cycle of pruning that should be done.- Default :
50
- Default :
--test_freq
: Frequency for Validation- Default :
50
- Default :
--batch_size
: Batch size- Default :
60
- Default :
--dataset
: Choice of dataset- Options :
mnist
,cifar10
- Default :
mnist
- Options :
--arch_type
: Type of architecture- Options :
fc1
- Simple fully connected network,lenet5
- LeNet5,AlexNet
- AlexNet,resnet18
- Resnet18,vgg16
- VGG16 - Default :
fc1
- Options :
--prune_percent
: Percentage of weight to be pruned after each cycle.- Default :
5
- Default :
--mini_batch
: Experiment on mini-batch- Default :
False
- Default :
--score
: Using score matrix to determine the pruning mask- Default :
False
- Default :
--binarize
: Model binarization- Default :
False
- Default :
The-Lottery-Ticket-Hypothesis-Binary-Neural-Networks-Pruning
├── models
│ ├── cifar10
│ │ ├── AlexNet.py
│ │ ├── fc1.py
│ │ ├── LeNet5.py
│ │ ├── resnet.py
│ │ ├── SmallVGG.py
│ │ └── vgg.py
│ └── mnist
│ ├── AlexNet.py
│ ├── fc1.py
│ ├── LeNet5.py
│ ├── resnet.py
│ ├── SmallVGG.py
│ └── vgg.py
├── dumps
├── main.py
├── plots
├── README.md
├── requirements.txt
├── saves
└── utils.py