Hi,
Thanks for releasing the code.
I found out that the performance of resnet20 on cifar10 is not that good.
here is the script I use:
python main.py --config configs/smallscale/resnet18/resnet20-ukn-unsigned.yaml --multigpu 3 --name cifar10_resnet20 --data ../datasets/ --prune-rate 0.5
which achieves only 85% acc on test set.
the strange thing is that when I set the prune_rate to 0.9 which means I keeps more weights compared to 0.5, but I got only 50% acc on test set.
python main.py --config configs/smallscale/resnet18/resnet20-ukn-unsigned.yaml --multigpu 3 --name cifar10_resnet20 --data ../datasets/ --prune-rate 0.9
can you tell me if anything wrong in my script?
Hi,
Thanks for releasing the code.
I found out that the performance of resnet20 on cifar10 is not that good.
here is the script I use:
python main.py --config configs/smallscale/resnet18/resnet20-ukn-unsigned.yaml --multigpu 3 --name cifar10_resnet20 --data ../datasets/ --prune-rate 0.5
which achieves only 85% acc on test set.
the strange thing is that when I set the prune_rate to 0.9 which means I keeps more weights compared to 0.5, but I got only 50% acc on test set.
python main.py --config configs/smallscale/resnet18/resnet20-ukn-unsigned.yaml --multigpu 3 --name cifar10_resnet20 --data ../datasets/ --prune-rate 0.9
can you tell me if anything wrong in my script?