Marian: Cost-effective High-Quality Neural Machine Translation in C++
This work addresses the need for cost-effective and efficient neural machine translation systems, though it is incremental as it builds on existing methods like Transformer variants.
The paper tackled optimizing the Transformer model for neural machine translation by combining teacher-student training, low-precision matrix products, and auto-tuning, resulting in high-quality, high-performance models that dominated the Pareto frontier in the WNMT 2018 shared task.
This paper describes the submissions of the "Marian" team to the WNMT 2018 shared task. We investigate combinations of teacher-student training, low-precision matrix products, auto-tuning and other methods to optimize the Transformer model on GPU and CPU. By further integrating these methods with the new averaging attention networks, a recently introduced faster Transformer variant, we create a number of high-quality, high-performance models on the GPU and CPU, dominating the Pareto frontier for this shared task.