arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.15767cs.LGcs.CV
收藏

[转载报道]彩票作弊:让所有彩票中奖者

[Reproducibility Report] Rigging the Lottery: Making All Tickets Winners

University of Wisconsin Madison(威斯康星大学麦迪逊分校);Stanford University(斯坦福大学)

Varun Sundar, Rajat Vadiraj Dwaraknath

摘要:

$\textit{RigL}$是一种稀疏训练算法,它声称可以直接训练稀疏网络,这些网络在固定的参数计数和计算预算下匹配或超过现有的密集到稀疏训练技术(如修剪)的性能。我们在Pytorch中从头开始实现了$\textit{RigL}$,并在CIFAR-10上重现了其性能,误差在报告值的0.1%以内。在两个CIFAR-10/100上,中心主张都成立--给定固定的训练预算,$\textit{RigL}$在目标稀疏度范围内优于现有的动态稀疏训练方法。通过训练更长的时间,性能可以匹配或超过迭代修剪,同时在整个训练过程中消耗恒定的FLOP。我们还表明,在调整$\textit{RigL}$的超参数的每一个稀疏,初始化对-超参数的参考选择往往是接近最佳性能的好处不大。超越了原来的文件,我们发现,最佳的初始化方案取决于训练约束。虽然Erdos-Renyi-Kernel分布对于固定参数计数优于均匀分布,但对于固定FLOP计数,后者表现更好。最后,在训练时重新分配逐层稀疏性可以弥合两种初始化方案之间的性能差距,但会增加计算成本。

英文摘要:

$\textit{RigL}$, a sparse training algorithm, claims to directly train sparse networks that match or exceed the performance of existing dense-to-sparse training techniques (such as pruning) for a fixed parameter count and compute budget. We implement $\textit{RigL}$ from scratch in Pytorch and reproduce its performance on CIFAR-10 within 0.1% of the reported value. On both CIFAR-10/100, the central claim holds -- given a fixed training budget, $\textit{RigL}$ surpasses existing dynamic-sparse training methods over a range of target sparsities. By training longer, the performance can match or exceed iterative pruning, while consuming constant FLOPs throughout training. We also show that there is little benefit in tuning $\textit{RigL}$'s hyper-parameters for every sparsity, initialization pair -- the reference choice of hyperparameters is often close to optimal performance. Going beyond the original paper, we find that the optimal initialization scheme depends on the training constraint. While the Erdos-Renyi-Kernel distribution outperforms the Uniform distribution for a fixed parameter count, for a fixed FLOP count, the latter performs better. Finally, redistributing layer-wise sparsity while training can bridge the performance gap between the two initialization schemes, but increases computational cost.

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。

↑