CaltechAUTHORS
  A Caltech Library Service

Long-Short Transformer: Efficient Transformers for Language and Vision

Zhu, Chen and Ping, Wei and Xiao, Chaowei and Shoeybi, Mohammad and Goldstein, Tom and Anandkumar, Anima and Catanzaro, Bryan (2021) Long-Short Transformer: Efficient Transformers for Language and Vision. . (Unpublished) https://resolver.caltech.edu/CaltechAUTHORS:20210831-203935342

[img] PDF - Submitted Version
Creative Commons Attribution Non-commercial Share Alike.

1MB

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20210831-203935342

Abstract

Transformers have achieved success in both language and vision domains. However, it is prohibitively expensive to scale them to long sequences such as long documents or high-resolution images, because self-attention mechanism has quadratic time and memory complexities with respect to the input sequence length. In this paper, we propose Long-Short Transformer (Transformer-LS), an efficient self-attention mechanism for modeling long sequences with linear complexity for both language and vision tasks. It aggregates a novel long-range attention with dynamic projection to model distant correlations and a short-term attention to capture fine-grained local correlations. We propose a dual normalization strategy to account for the scale mismatch between the two attention mechanisms. Transformer-LS can be applied to both autoregressive and bidirectional models without additional complexity. Our method outperforms the state-of-the-art models on multiple tasks in language and vision domains, including the Long Range Arena benchmark, autoregressive language modeling, and ImageNet classification. For instance, Transformer-LS achieves 0.97 test BPC on enwik8 using half the number of parameters than previous method, while being faster and is able to handle 3x as long sequences compared to its full-attention version on the same hardware. On ImageNet, it can obtain the state-of-the-art results (e.g., a moderate size of 55.8M model solely trained on 224x224 ImageNet-1K can obtain Top-1 accuracy 84.1%), while being more scalable on high-resolution images. The source code and models are released at https://github.com/NVIDIA/transformer-ls.


Item Type:Report or Paper (Discussion Paper)
Related URLs:
URLURL TypeDescription
http://arxiv.org/abs/2107.02192arXivDiscussion Paper
https://github.com/NVIDIA/transformer-lsRelated ItemSource code and models
Additional Information:Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)
Record Number:CaltechAUTHORS:20210831-203935342
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20210831-203935342
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:110656
Collection:CaltechAUTHORS
Deposited By: George Porter
Deposited On:01 Sep 2021 14:12
Last Modified:01 Sep 2021 14:12

Repository Staff Only: item control page