arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2009.10656cs.DCcs.ARcs.LG
收藏

E-BATCH:高能效和高产能的RNN配料

E-BATCH: Energy-Efficient and High-Throughput RNN Batching

Universitat Politècnica de Catalunya(加泰罗尼亚理工大学)

Franyell Silfa, Jose Maria Arnau, Antonio Gonzalez

摘要:

递归神经网络(RNN)推理由于跨时间步长的严格数据依赖性而表现出较低的硬件利用率。批处理多个请求可以提高吞吐量。然而,RNN排序需要大量的填充,因为批量输入序列的长度可能会有很大的不同。每隔几个时间步动态更新一次批处理的方案避免了填充。然而,它们需要在短时间内执行不同的RNN层,从而降低了能源效率。因此,我们提出了E-BATCH,这是一种为RNN加速器量身定制的低延迟和节能的并行计算方案。它包括一个运行时系统和有效的硬件支持。运行时连接多个序列以创建大的批处理,从而大大节省了能源。此外,当序列的评估完成时,加速器会通知它,以便可以立即将新序列添加到批处理中,从而大大减少填充量。E-BATCH动态控制每个批次评估的时间步长数,以实现给定硬件平台的延迟和能效之间的最佳权衡。我们在E-PUR和TPU的基础上评估E-BATCH。在E-PUR中,E-BATCH将吞吐量提高了1.8倍,能效提高了3.6倍,而在TPU中,它将吞吐量提高了2.1倍,能效提高了1.6倍。

英文摘要:

Recurrent Neural Network (RNN) inference exhibits low hardware utilization due to the strict data dependencies across time-steps. Batching multiple requests can increase throughput. However, RNN batching requires a large amount of padding since the batched input sequences may largely differ in length. Schemes that dynamically update the batch every few time-steps avoid padding. However, they require executing different RNN layers in a short timespan, decreasing energy efficiency. Hence, we propose E-BATCH, a low-latency and energy-efficient batching scheme tailored to RNN accelerators. It consists of a runtime system and effective hardware support. The runtime concatenates multiple sequences to create large batches, resulting in substantial energy savings. Furthermore, the accelerator notifies it when the evaluation of a sequence is done, so that a new sequence can be immediately added to a batch, thus largely reducing the amount of padding. E-BATCH dynamically controls the number of time-steps evaluated per batch to achieve the best trade-off between latency and energy efficiency for the given hardware platform. We evaluate E-BATCH on top of E-PUR and TPU. In E-PUR, E-BATCH improves throughput by 1.8x and energy-efficiency by 3.6x, whereas in TPU, it improves throughput by 2.1x and energy-efficiency by 1.6x, over the state-of-the-art.

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。

↑