The patent badge is an abbreviated version of the USPTO patent document. The patent badge does contain a link to the full patent document.

The patent badge is an abbreviated version of the USPTO patent document. The patent badge covers the following: Patent number, Date patent was issued, Date patent was filed, Title of the patent, Applicant, Inventor, Assignee, Attorney firm, Primary examiner, Assistant examiner, CPCs, and Abstract. The patent badge does contain a link to the full patent document (in Adobe Acrobat format, aka pdf). To download or print any patent click here.

Date of Patent:
Jul. 28, 2026

Filed:

Mar. 31, 2023
Applicant:

Amazon Technologies, Inc., Seattle, WA (US);

Inventors:

Praphruetpong Athiwaratkun, Jersey City, NJ (US);

Jiacheng Guo, Kirkland, WA (US);

Yuchen Tian, Santa Clara, CA (US);

Haifeng Qian, Mount Kisco, NY (US);

Vamshidhar Krishnamurthy Dantu, Sunnyvale, CA (US);

Matthew Lee, Elmhurst, NY (US);

Ramesh M. Nallapati, San Jose, CA (US);

Parminder Bhatia, Kearny, NJ (US);

Srinivas Iragavarapu, Redmond, WA (US);

Rama Krishna Sandeep Pokkunuri, Redmond, WA (US);

Sudipta Sengupta, Bellevue, WA (US);

Bing Xiang, Mount Kisco, NY (US);

Assignee:

Amazon Technologies, Inc., Seattle, WA (US);

Attorney:
Primary Examiner:
Assistant Examiner:
Int. Cl.
CPC ...
G06F 8/33 (2018.01); G06F 40/40 (2020.01);
U.S. Cl.
CPC ...
G06F 8/33 (2013.01); G06F 40/40 (2020.01);
Abstract

Techniques for improved attention calculations are described. In some examples, the attention calculations are for a Transformer-based model. In some examples, a trained model is applied to input data, wherein the application of the trained model at least includes determining at least one of logits or an output of an attention-block using incremental decoding, wherein the incremental decoding calculation is decomposed into at least an incremental calculation and a context calculation.


Find Patent Forward Citations

Loading…