Dynamic Transformer for Efficient Machine Translation on Embedded Devices

TL;DR

This paper introduces Dynamic-HAT, a resource-aware Transformer model for machine translation on embedded devices, enabling real-time adaptation to hardware constraints with minimal performance loss.

Contribution

It presents a novel dynamic Transformer approach that efficiently switches between sub-models at runtime based on resource availability, optimizing translation quality and latency.

Findings

Switching time between sub-transformers is less than 1 second.

BLEU score loss is under 1.5% when switching without retraining.

Performance can be scaled across different hardware with a 1% BLEU score improvement.

Abstract

The Transformer architecture is widely used for machine translation tasks. However, its resource-intensive nature makes it challenging to implement on constrained embedded devices, particularly where available hardware resources can vary at run-time. We propose a dynamic machine translation model that scales the Transformer architecture based on the available resources at any particular time. The proposed approach, 'Dynamic-HAT', uses a HAT SuperTransformer as the backbone to search for SubTransformers with different accuracy-latency trade-offs at design time. The optimal SubTransformers are sampled from the SuperTransformer at run-time, depending on latency constraints. The Dynamic-HAT is tested on the Jetson Nano and the approach uses inherited SubTransformers sampled directly from the SuperTransformer with a switching time of <1s. Using inherited SubTransformers results in a BLEU score loss of <1.5% because the SubTransformer configuration is not retrained from scratch after sampling. However, to recover this loss in performance, the dimensions of the design space can be reduced to tailor it to a family of target hardware. The new reduced design space results in a BLEU score increase of approximately 1% for sub-optimal models from the original design space, with a wide range for performance scaling between 0.356s - 1.526s for the GPU and 2.9s - 7.31s for the CPU.