Fast Neural Machine Translation Implementation
This work addresses efficiency for neural machine translation practitioners, but it is incremental as it builds on existing methods like the Amun engine.
The authors tackled the problem of improving the speed of neural machine translation inference on GPUs by implementing efficient mini-batching and fusing softmax with k-best extraction, resulting in their submissions achieving the top three fastest positions in a GPU efficiency track.
This paper describes the submissions to the efficiency track for GPUs at the Workshop for Neural Machine Translation and Generation by members of the University of Edinburgh, Adam Mickiewicz University, Tilde and University of Alicante. We focus on efficient implementation of the recurrent deep-learning model as implemented in Amun, the fast inference engine for neural machine translation. We improve the performance with an efficient mini-batching algorithm, and by fusing the softmax operation with the k-best extraction algorithm. Submissions using Amun were first, second and third fastest in the GPU efficiency track.