TaWQ

The official implementation of TaWQ

Requirements

pytorch

spikingjelly==0.0.0.12

timm==0.6.12

ImageNet

train with 8 GPUs or NPUs

python -m torch.distributed.launch --nproc_per_node=8 main.py --accum_iter 1 --batch_size 64 --blr 6e-4 --model qkformer_tawq_10_768 --output_dir ./qkformer_tawq_10_768 --num_workers 79 --data_path ../Data_ImageNet2012/ --epochs 200

CIFAR100

python train.py