arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2002.03090cs.LGstat.ML
收藏

BitPruning:学习比特长度以实现积极和准确的Quantization

BitPruning: Learning Bitlengths for Aggressive and Accurate Quantization

University of Toronto(多伦多大学);Mila(米拉研究所);IMT-Atlantique(IMT-大西洋高等电信学校)

Miloš Nikolić, Ghouthi Boukli Hacene, Ciaran Bannon, Alberto Delmas Lascorz, Matthieu Courbariaux, Yoshua Bengio, Vincent Gripon, Andreas Moshovos

摘要:

神经网络已经证明使用低位长整数量化实现了最先进的精度,在支持短位长的现有硬件设计上产生了执行时间和能量效益。然而,找到所需精度的最小位长的问题仍然是开放的。我们引入了一种训练方法,用于在保持准确性的同时在任何粒度下最小化推理位长度。也就是说,我们提出了一个正则化,惩罚整个架构的大位长表示,并展示了如何可以修改,以尽量减少其他可量化的标准,如操作或内存占用的数量。我们证明,我们的方法学习节俭的表示,同时保持准确性。使用ImageNet,该方法在AlexNet、ResNet 18和MobileNet V2上分别产生4.13、3.76和4.36位的平均每层位长,保持在基本TOP-1准确度的2.0%、0.5%和0.5%以内。

英文摘要:

Neural networks have demonstrably achieved state-of-the art accuracy using low-bitlength integer quantization, yielding both execution time and energy benefits on existing hardware designs that support short bitlengths. However, the question of finding the minimum bitlength for a desired accuracy remains open. We introduce a training method for minimizing inference bitlength at any granularity while maintaining accuracy. Namely, we propose a regularizer that penalizes large bitlength representations throughout the architecture and show how it can be modified to minimize other quantifiable criteria, such as number of operations or memory footprint. We demonstrate that our method learns thrifty representations while maintaining accuracy. With ImageNet, the method produces an average per layer bitlength of 4.13, 3.76 and 4.36 bits on AlexNet, ResNet18 and MobileNet V2 respectively, remaining within 2.0%, 0.5% and 0.5% of the base TOP-1 accuracy.

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。

↑