arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.07253cs.LGstat.ML
收藏

带反馈的动态模型剪枝

Dynamic Model Pruning with Feedback

EPFL(洛桑联邦理工学院);ETH Zurich(苏黎世联邦理工学院)

Tao Lin, Sebastian U. Stich, Luis Barba, Daniil Dmitriev, Martin Jaggi

摘要:

深度神经网络通常有数百万个参数。这可能会阻碍它们部署到低端设备,不仅是因为高内存要求,还因为推理延迟增加。我们提出了一种新的模型压缩方法,生成一个稀疏的训练模型,而没有额外的开销:通过允许(i)动态分配的稀疏模式和(ii)纳入反馈信号,以重新激活过早修剪的权重,我们获得一个性能稀疏模型在一个单一的训练通过(不需要再训练,但可以进一步提高性能)。我们在CIFAR-10和ImageNet上对我们的方法进行了评估,并表明所获得的稀疏模型可以达到密集模型的最新性能。此外,它们的性能超过了所有以前提出的修剪计划所产生的模型。

英文摘要:

Deep neural networks often have millions of parameters. This can hinder their deployment to low-end devices, not only due to high memory requirements but also because of increased latency at inference. We propose a novel model compression method that generates a sparse trained model without additional overhead: by allowing (i) dynamic allocation of the sparsity pattern and (ii) incorporating feedback signal to reactivate prematurely pruned weights we obtain a performant sparse model in one single training pass (retraining is not needed, but can further improve the performance). We evaluate our method on CIFAR-10 and ImageNet, and show that the obtained sparse models can reach the state-of-the-art performance of dense models. Moreover, their performance surpasses that of models generated by all previously proposed pruning schemes.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑