shell commands for using LoFT

Pretraining

python3 filterwise_IST_imagenet_2xresnet18.py --rank=0 --cuda-id=0 --resume=0 &

python3 filterwise_IST_imagenet_2xresnet18.py --rank=1 --cuda-id=1 --resume=0 &

python3 filterwise_IST_imagenet_2xresnet18.py --rank=2 --cuda-id=2 --resume=0 &

python3 filterwise_IST_imagenet_2xresnet18.py --rank=3 --cuda-id=3 --resume=0 &

Pruning

python prune.py --model './checkpoint_folder/checkpoint_name.pth' --save './prune_result_folder'

Fine-tuning

python finetune.py --refine './prune_result_folder/pruned.pth.tar' --save './target_folder' --epochs 90

to resume, use

python finetune.py --refine './prune_result_folder/pruned.pth.tar' --save './target_folder' --epochs 90 --resume './target_folder/apoint.pth.tar'