Towards Higher Ranks via Adversarial Weight Pruning

Authors: Yuchuan Tian, Hanting Chen, Tianyu Guo, Chao Xu, Yunhe Wang

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on various datasets and different tasks demonstrate the effectiveness of our algorithm in high sparsity.
Researcher Affiliation Collaboration Yuchuan Tian1, Hanting Chen2, Tianyu Guo2, Chao Xu1, Yunhe Wang2 1 National Key Lab of General AI, School of Intelligence Science and Technology, Peking University. 2 Huawei Noah s Ark Lab.
Pseudocode Yes The whole pruning framework is detailed in Algorithm 1.
Open Source Code Yes The codes are available at https://github.com/huawei-noah/Efficient-Computing/tree/ master/Pruning/RPG and https://gitee.com/mindspore/models/tree/ master/research/cv/RPG.
Open Datasets Yes CIFAR-10 is one of the most widely used benchmark for image classification. It consists of 60000 32 32 images: 50000 for training, and 10000 for validation. Image Net ISLVRC2012 [9] is a large scale image classification dataset. Mask R-CNN pruning on COCO val2017.
Dataset Splits Yes CIFAR-10... 50000 for training, and 10000 for validation. Image Net ISLVRC2012 [9]... It contains 1281K images in the training set and 50K images in the validation set.
Hardware Specification Yes Image Net experiments are run on 8 NVIDIA Tesla V100s.
Software Dependencies No We gratefully acknowledge the support of Mind Spore [38], CANN (Compute Architecture for Neural Networks) and Ascend AI Processor used for this research. (No version numbers provided for MindSpore or CANN).
Experiment Setup Yes We use the SGD optimizer with momentum 0.9, batchsize 128, learning rate 0.1, and weight decay 0.005.