Dynamic Sparse Attention for Scalable Transformer Acceleration

Transformers are the mainstream of NLP applications and are becoming increasingly popular in other domains such as Computer Vision. Despite the improvements in model quality, the enormous computation costs make Transformers difficult at deployment, especially when the sequence length is large in eme...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE transactions on computers Ročník 71; číslo 12; s. 1 - 14
Hlavní autoři: Liu, Liu, Qu, Zheng, Chen, Zhaodong, Tu, Fengbin, Ding, Yufei, Xie, Yuan
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York IEEE 01.01.2022
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Témata:
Complexity
Complexity theory
Computational modeling
Computer vision
Costs
Hardware
Model accuracy
Redundancy
Sparse matrices
Transformers
ISSN:0018-9340, 1557-9956
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Transformers are the mainstream of NLP applications and are becoming increasingly popular in other domains such as Computer Vision. Despite the improvements in model quality, the enormous computation costs make Transformers difficult at deployment, especially when the sequence length is large in emerging applications. Processing attention mechanism as the essential component of Transformer is the bottleneck of execution due to the quadratic complexity. Prior art explores sparse patterns in attention to support long sequence modeling, but those pieces of work are on static or fixed patterns. We demonstrate that the sparse patterns are dynamic, depending on input sequences. Thus, we propose the Dynamic Sparse Attention (DSA) that can efficiently exploit dynamic sparse patterns in attention. Compared with other methods, our approach can achieve better trade-offs between accuracy and model complexity. Moving forward, we identify challenges and provide solutions to implement DSA on existing hardware (GPUs) and specialized hardware in order to achieve practical speedup and efficiency improvements for Transformer execution.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0018-9340
1557-9956
DOI:10.1109/TC.2022.3208206