EfficientLIF-Net

Pytorch code for "Sharing Leaky-Integrate-and-Fire Neurons for Memory-Efficient Spiking Neural Networks"

Dependencies

• Python 3.9
• PyTorch 1.10.0
• Spikingjelly

Training Details

For ResNet19/VGG16 architecture,

(1) Standard SNN

python train_lifshare.py --dataset 'cifar10' --arch 'resnet19' --lifshare 'noshare' --batch_size 128 --learning_rate 1e-1
python train_lifshare.py --dataset 'cifar10' --arch 'vgg16' --lifshare 'noshare' --batch_size 128 --learning_rate 1e-1

(2) Cross-Layer sharing SNN

python train_lifshare.py --dataset 'cifar10' --arch 'resnet19' --lifshare 'layer' --batch_size 128 --learning_rate 1e-1
python train_lifshare.py --dataset 'cifar10' --arch 'vgg16' --lifshare 'layer' --batch_size 128 --learning_rate 1e-1

(3) Cross-Channel sharing SNN

python train_lifshare.py --dataset 'cifar10' --arch 'resnet19' --lifshare 'channel' --ch_group_num 2 --batch_size 128 --learning_rate 1e-1
python train_lifshare.py --dataset 'cifar10' --arch 'vgg16' --lifshare 'channel' --ch_group_num 2 --batch_size 128 --learning_rate 1e-1

(3) Cross-Layer+Channel sharing SNN

python train_lifshare.py --dataset 'cifar10' --arch 'resnet19' --lifshare 'layerchannel' --batch_size 128 --learning_rate 1e-1
python train_lifshare.py --dataset 'cifar10' --arch 'vgg16' --lifshare 'layerchannel' --batch_size 128 --learning_rate 1e-1