MobileNMT: Enabling Translation in 15MB and 30ms
This work addresses the problem of enabling efficient, low-latency, and privacy-preserving translation on mobile devices for users in offline or resource-constrained scenarios, representing a strong specific gain rather than a foundational advancement.
The paper tackles the challenge of deploying neural machine translation (NMT) models on mobile devices by introducing MobileNMT, a system that achieves translation in 15MB and 30ms, with a 47.0x speedup and 99.5% memory savings compared to existing systems, at the cost of only 11.6% BLEU loss.
Deploying NMT models on mobile devices is essential for privacy, low latency, and offline scenarios. For high model capacity, NMT models are rather large. Running these models on devices is challenging with limited storage, memory, computation, and power consumption. Existing work either only focuses on a single metric such as FLOPs or general engine which is not good at auto-regressive decoding. In this paper, we present MobileNMT, a system that can translate in 15MB and 30ms on devices. We propose a series of principles for model compression when combined with quantization. Further, we implement an engine that is friendly to INT8 and decoding. With the co-design of model and engine, compared with the existing system, we speed up 47.0x and save 99.5% of memory with only 11.6% loss of BLEU. The code is publicly available at https://github.com/zjersey/Lightseq-ARM.