CytonMT: an Efficient Neural Machine Translation Open-source Toolkit Implemented in C++

TL;DR

CytonMT is an open-source C++ toolkit for neural machine translation that offers high training efficiency, simplicity, and competitive translation quality, significantly accelerating training speed on GPUs.

Contribution

It introduces a C++-based NMT toolkit optimized for GPU acceleration, improving training speed while maintaining translation quality.

Findings

Training speed increased by up to 110.8%

Achieves competitive translation quality

Built entirely with C++ and NVIDIA libraries

Abstract

This paper presents an open-source neural machine translation toolkit named CytonMT (https://github.com/arthurxlw/cytonMt). The toolkit is built from scratch only using C++ and NVIDIA's GPU-accelerated libraries. The toolkit features training efficiency, code simplicity and translation quality. Benchmarks show that CytonMT accelerates the training speed by 64.5% to 110.8% on neural networks of various sizes, and achieves competitive translation quality.