SimA: Simple Softmax-free Attention for Vision Transformers

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Koohpayegani, Soroush Abbasi, Pirsiavash, Hamed
Natura: Preprint
Pubblicazione: 2022
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911809921875968
author Koohpayegani, Soroush Abbasi
Pirsiavash, Hamed
author_facet Koohpayegani, Soroush Abbasi
Pirsiavash, Hamed
contents Recently, vision transformers have become very popular. However, deploying them in many applications is computationally expensive partly due to the Softmax layer in the attention block. We introduce a simple but effective, Softmax-free attention block, SimA, which normalizes query and key matrices with simple $\ell_1$-norm instead of using Softmax layer. Then, the attention block in SimA is a simple multiplication of three matrices, so SimA can dynamically change the ordering of the computation at the test time to achieve linear computation on the number of tokens or the number of channels. We empirically show that SimA applied to three SOTA variations of transformers, DeiT, XCiT, and CvT, results in on-par accuracy compared to the SOTA models, without any need for Softmax layer. Interestingly, changing SimA from multi-head to single-head has only a small effect on the accuracy, which simplifies the attention block further. The code is available here: https://github.com/UCDvision/sima
format Preprint
id arxiv_https___arxiv_org_abs_2206_08898
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle SimA: Simple Softmax-free Attention for Vision Transformers
Koohpayegani, Soroush Abbasi
Pirsiavash, Hamed
Computer Vision and Pattern Recognition
Recently, vision transformers have become very popular. However, deploying them in many applications is computationally expensive partly due to the Softmax layer in the attention block. We introduce a simple but effective, Softmax-free attention block, SimA, which normalizes query and key matrices with simple $\ell_1$-norm instead of using Softmax layer. Then, the attention block in SimA is a simple multiplication of three matrices, so SimA can dynamically change the ordering of the computation at the test time to achieve linear computation on the number of tokens or the number of channels. We empirically show that SimA applied to three SOTA variations of transformers, DeiT, XCiT, and CvT, results in on-par accuracy compared to the SOTA models, without any need for Softmax layer. Interestingly, changing SimA from multi-head to single-head has only a small effect on the accuracy, which simplifies the attention block further. The code is available here: https://github.com/UCDvision/sima
title SimA: Simple Softmax-free Attention for Vision Transformers
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2206.08898