arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1804.03243cs.CL
收藏

基于GPU的精确格生成WFST解码器

A GPU-based WFST Decoder with Exact Lattice Generation

Shanghai Jiao Tong University(上海交通大学);NVIDIA(英伟达);Johns Hopkins University(约翰斯·霍普金斯大学)

Zhehuai Chen, Justin Luitjens, Hainan Xu, Yiming Wang, Daniel Povey, Sanjeev Khudanpur

摘要:

我们描述的Kaldi工具包的扩展,支持加权有限状态传感器(WFST)解码的图形处理单元(GPU)的初步工作。我们实现令牌重组作为一个原子GPU操作,以完全并行化的维特比束搜索,并提出了一个动态的负载均衡策略,更有效的令牌传递调度GPU线程之间。我们还重新设计了精确的格生成和格修剪算法,以更好地利用GPU。在Switchboard语料库上的实验表明,该方法在识别和置信度测量任务中获得了相同的1-best结果和格质量,而运行速度比单进程Kaldi解码器快3到15倍。上述结果是在不同的GPU架构上报告的。此外,我们获得了46倍的加速与序列并行和多进程服务(MPS)在GPU。

英文摘要:

We describe initial work on an extension of the Kaldi toolkit that supports weighted finite-state transducer (WFST) decoding on Graphics Processing Units (GPUs). We implement token recombination as an atomic GPU operation in order to fully parallelize the Viterbi beam search, and propose a dynamic load balancing strategy for more efficient token passing scheduling among GPU threads. We also redesign the exact lattice generation and lattice pruning algorithms for better utilization of the GPUs. Experiments on the Switchboard corpus show that the proposed method achieves identical 1-best results and lattice quality in recognition and confidence measure tasks, while running 3 to 15 times faster than the single process Kaldi decoder. The above results are reported on different GPU architectures. Additionally we obtain a 46-fold speedup with sequence parallelism and multi-process service (MPS) in GPU.

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。

↑