Invention Grant
- Patent Title: Task activating for accelerated deep learning
-
Application No.: US16604108Application Date: 2018-04-17
-
Publication No.: US11157806B2Publication Date: 2021-10-26
- Inventor: Sean Lie , Michael Morrison , Srikanth Arekapudi , Michael Edwin James , Gary R. Lauterbach
- Applicant: Cerebras Systems Inc.
- Applicant Address: US CA Los Altos
- Assignee: Cerebras Systems Inc.
- Current Assignee: Cerebras Systems Inc.
- Current Assignee Address: US CA Los Altos
- Agency: PatentVentures
- Agent Bennett Smith; Korbin Van Dyke
- International Application: PCT/IB2018/052651 WO 20180417
- International Announcement: WO2018/193370 WO 20181025
- Main IPC: G06N3/063
- IPC: G06N3/063 ; G06N3/04 ; G06F13/00 ; G06N3/08 ; G06F9/30 ; H04L12/935 ; G06F9/38 ; H04L12/54 ; G06F5/06 ; G06F13/40 ; H04L12/931 ; G06F30/392

Abstract:
Techniques in advanced deep learning provide improvements in one or more of accuracy, performance, and energy efficiency. An array of processing elements performs flow-based computations on wavelets of data. Each processing element has a compute element and a routing element. Each router enables communication via wavelets with at least nearest neighbors in a 2D mesh. Routing is controlled by virtual channel specifiers in each wavelet and routing configuration information in each router. Execution of an activate instruction or completion of a fabric vector operation activates one of the virtual channels. A virtual channel is selected from a pool comprising previously activated virtual channels and virtual channels associated with previously received wavelets. A task corresponding to the selected virtual channel is activated by executing instructions corresponding to the selected virtual channel.
Public/Granted literature
- US20210004674A1 TASK ACTIVATING FOR ACCELERATED DEEP LEARNING Public/Granted day:2021-01-07
Information query