AutoMoE: Heterogeneous Mixture-of-Experts with Adaptive Computation for Efficient Neural Machine Translation

Ganesh Jawahar, Subhabrata Mukherjee, Xiaodong Liu, Young Jin Kim, Muhammad Abdul-Mageed, Laks Lakshmanan, V.S., Ahmed Hassan Awadallah, Sebastien Bubeck, Jianfeng Gao

Findings: Machine Learning for NLP Findings Paper

Session 1: Machine Learning for NLP (Virtual Poster)
Conference Room: Pier 7&8
Conference Time: July 10, 11:00-12:30 (EDT) (America/Toronto)
Global Time: July 10, Session 1 (15:00-16:30 UTC)
Spotlight Session: Spotlight - Metropolitan Centre (Spotlight)
Conference Room: Metropolitan Centre
Conference Time: July 10, 19:00-21:00 (EDT) (America/Toronto)
Global Time: July 10, Spotlight Session (23:00-01:00 UTC)
Keywords: model compression methods
TLDR: Mixture-of-Expert (MoE) models have obtained state-of-the-art performance in Neural Machine Translation (NMT) tasks. Existing works in MoE mostly consider a homogeneous design where the same number of experts of the same size are placed uniformly throughout the network. Furthermore, existing MoE wor...
You can open the #paper-P2243 channel in a separate window.
Abstract: Mixture-of-Expert (MoE) models have obtained state-of-the-art performance in Neural Machine Translation (NMT) tasks. Existing works in MoE mostly consider a homogeneous design where the same number of experts of the same size are placed uniformly throughout the network. Furthermore, existing MoE works do not consider computational constraints (e.g., FLOPs, latency) to guide their design. To this end, we develop AutoMoE -- a framework for designing heterogeneous MoE's under computational constraints. AutoMoE leverages Neural Architecture Search (NAS) to obtain efficient sparse MoE sub-transformers with 4x inference speedup (CPU) and FLOPs reduction over manually designed Transformers, with parity in BLEU score over dense Transformer and within 1 BLEU point of MoE SwitchTransformer, on aggregate over benchmark datasets for NMT. Heterogeneous search space with dense and sparsely activated Transformer modules (e.g., how many experts? where to place them? what should be their sizes?) allows for adaptive compute -- where different amounts of computations are used for different tokens in the input. Adaptivity comes naturally from routing decisions which send tokens to experts of different sizes. AutoMoE code, data, and trained models are available at https://aka.ms/AutoMoE.