Timezone: »

 
Poster
Scaling Vision with Sparse Mixture of Experts
Carlos Riquelme · Joan Puigcerver · Basil Mustafa · Maxim Neumann · Rodolphe Jenatton · André Susano Pinto · Daniel Keysers · Neil Houlsby

Tue Dec 07 08:30 AM -- 10:00 AM (PST) @
in Poster Session 1 »

Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are "dense", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet.

Keywords: Transformers · Vision · Language

Author Information

Carlos Riquelme (Google Brain)
Joan Puigcerver (Universitat Politecnica de Valencia)
Basil Mustafa (Google)
Maxim Neumann (Google)
Rodolphe Jenatton (Amazon)
André Susano Pinto (Google)
Daniel Keysers (Google Research, Brain Team)
Neil Houlsby (Google)

More from the Same Authors