selkerdawy / FTWT

Fire Together Wire Together: A Dynamic Pruning Approach with Self-Supervised Mask Prediction

Fire Together Wire Together (FTWT)

Sara Elkerdawy¹, Mostafa Elhoushi², Hong Zhang¹, Nilanjan Ray¹

¹ Computing Science Departement, University of Alberta, Canada
² Toronto Heterogeneous Compilers Lab, Huawei, Canada

Sample training code for CIFAR fo dynamic pruning with self-supervised mask.

[Project Page], [Paper CVPR22], [Poster], [Video]

FLOPs reduction vs accuracy drop from baselines for various dynamic and static models on ResNet34 ImageNet.

Environment

virtualenv .envpy36 -p python3.6 #Initialize environment
source .envpy36/bin/activate
pip install -r req.txt # Install dependencies

Train baseline

sh job_baseline.sh #You can change model at line 5

Train dynamic

sh job_dynamic.sh #You can change model at line 5 and threshold at line 40

Results

dataset	model	mthresh	mode	Accuracy	FLOPS Reduction (%)
cifar10	vgg16-bn			93.82%	Baseline
		0.92	joint	93.55%	65%
		0.92	decoupled	93.73%	56%
		0.85	decoupled	93.19%	73%
		0.88	joint	92.65%	74%
	resnet56			93.66%	Baseline
		0.80	decoupled	92.63%	66%
		0.88	joint	92.28%	54%
	mobilenetv1			90.89%	Baseline
		1.00	decoupled	91.06%	78%
		1.00	joint	91.21%	78%
imagenet	resnet34			73.30%	Baseline
		0.97	decoupled	73.25%	25.86%
		0.95	decoupled	72.79%	37.77%
		0.93	decoupled	72.17%	47.42%
		0.92	decoupled	71.71%	52.24%
	resnet18			69.76%	Baseline
		0.91	decoupled	67.49%	51.56%
	mobilenetv1			69.57%	Baseline
		1.00	decoupled	69.66%	41.07%

Citation

@InProceedings{elkerdawy2022fire,
    author    = {Elkerdawy, Sara and Elhoushi, Mostafa and Zhang, Hong and Ray, Nilanjan},
    title     = {Fire Together Wire Together: A Dynamic Pruning Approach with Self-Supervised Mask Prediction},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
}