The NiuTrans System for the WMT21 Efficiency Task

TL;DR

This paper presents the NiuTrans system for the WMT21 efficiency task, achieving the fastest translation speed and lowest GPU memory usage through lightweight architectures, knowledge distillation, and optimization techniques.

Contribution

The paper introduces a highly efficient translation system combining lightweight Transformers, knowledge distillation, and various optimizations, setting new speed and memory benchmarks.

Findings

Translates 247,000 words/sec on NVIDIA A100

3x faster than previous system

Lowest GPU memory consumption

Abstract

This paper describes the NiuTrans system for the WMT21 translation efficiency task (http://statmt.org/wmt21/efficiency-task.html). Following last year's work, we explore various techniques to improve efficiency while maintaining translation quality. We investigate the combinations of lightweight Transformer architectures and knowledge distillation strategies. Also, we improve the translation efficiency with graph optimization, low precision, dynamic batching, and parallel pre/post-processing. Our system can translate 247,000 words per second on an NVIDIA A100, being 3$\times$ faster than last year's system. Our system is the fastest and has the lowest memory consumption on the GPU-throughput track. The code, model, and pipeline will be available at NiuTrans.NMT (https://github.com/NiuTrans/NiuTrans.NMT).