pytorch-admm-prunning

It is a pytorch implementation of DNN weight prunning with ADMM described in A Systematic DNN Weight Pruning Framework using Alternating Direction Method of Multipliers.

Train and test

• You can simply run code by

$ python main.py

• In the paper, authors use l2-norm regularization so you can easily add by

$ python main.py --l2

• Beyond this paper, if you don't want to use predefined prunning ratio, admm with l1 norm regularization can give a great solution and can be simply tested by

$ python main.py --l1

• There are two dataset you can test in this code: [mnist, cifar10]. Default setting is mnist, you can change dataset by

$ python main.py --dataset cifar10

Models

• In this code, there are two models: [LeNet, AlexNet]. I use LeNet for mnist, AlexNet for cifar10 by default.

Optimizer

• To prevent prunned weights from updated by optimizer, I modified Adam (named PruneAdam).

References

For this repository, I refer to KaiqiZhang's tensorflow implementation.