UCNN：通过权重重复在深度神经网络中利用计算重用
UCNN: Exploiting Computational Reuse in Deep Neural Networks via Weight Repetition
摘要：
卷积神经网络（CNN）由于其每次操作的高精度和机器效率，已经开始渗透到电子社会的各个角落（从话音识别到场景生成）。在其核心，CNN计算由权重和输入向量之间的多维点积组成。本文研究了如何在CNN推理过程中利用权重重复--当相同的权重在权重向量中或权重向量之间多次出现时--来节省能量并提高性能。这概括了从CNN权重稀疏性提高效率的流行工作路线，因为由于重复的零权重而减少计算是由于重复的权重而减少计算的特殊情况。为了利用权重重复，本文提出了一种新的CNN加速器，称为唯一权重CNN加速器（UCNN）。UCNN使用权重重复来重用CNN子计算（例如，点积），并在存储在片外DRAM中时减小CNN模型大小-这两者都节省了能量。UCNN通过利用权重的稀疏性进一步提高了性能。我们使用加速器级循环和能量模型以及UCNN处理元素的RTL实现来评估UCNN。在三个当代CNN上，相对于使用Eyeriss风格稀疏优化的类似配置的基线加速器，UCNN将吞吐量归一化能耗提高了1.2 - 4倍。与此同时，UCNN处理元素相对于相同的基线仅增加了17-24%的面积开销。
英文摘要：
Convolutional Neural Networks (CNNs) have begun to permeate all corners of electronic society (from voice recognition to scene generation) due to their high accuracy and machine efficiency per operation. At their core, CNN computations are made up of multi-dimensional dot products between weight and input vectors. This paper studies how weight repetition ---when the same weight occurs multiple times in or across weight vectors--- can be exploited to save energy and improve performance during CNN inference. This generalizes a popular line of work to improve efficiency from CNN weight sparsity, as reducing computation due to repeated zero weights is a special case of reducing computation due to repeated weights. To exploit weight repetition, this paper proposes a new CNN accelerator called the Unique Weight CNN Accelerator (UCNN). UCNN uses weight repetition to reuse CNN sub-computations (e.g., dot products) and to reduce CNN model size when stored in off-chip DRAM ---both of which save energy. UCNN further improves performance by exploiting sparsity in weights. We evaluate UCNN with an accelerator-level cycle and energy model and with an RTL implementation of the UCNN processing element. On three contemporary CNNs, UCNN improves throughput-normalized energy consumption by 1.2x - 4x, relative to a similarly provisioned baseline accelerator that uses Eyeriss-style sparsity optimizations. At the same time, the UCNN processing element adds only 17-24% area overhead relative to the same baseline.
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。
