The NiuTrans System for the WMT21 Efficiency Task
This work addresses efficiency bottlenecks for machine translation systems, particularly in high-throughput GPU applications, though it is incremental as it builds on prior techniques.
The paper tackles the challenge of improving machine translation efficiency while maintaining quality, achieving a speed of 247,000 words per second on an NVIDIA A100, which is 3 times faster than the previous year's system.
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).