Invention Grant
- Patent Title: Reconfigurable neural network processing based on subgraph recognition
-
Application No.: US18231024Application Date: 2023-08-07
-
Publication No.: US12045611B1Publication Date: 2024-07-23
- Inventor: Ron Diamant , Hongbin Zheng , Drazen Borkovic , Haichen Li
- Applicant: Amazon Technologies, Inc.
- Applicant Address: US WA Seattle
- Assignee: Amazon Technologies, Inc.
- Current Assignee: Amazon Technologies, Inc.
- Current Assignee Address: US WA Seattle
- Agency: Kilpatrick Townsend & Stockton LLP
- The original application number of the division: US17361992 2021.06.29
- Main IPC: G06F9/30
- IPC: G06F9/30 ; G06F7/548 ; G06F8/41 ; G06N3/04 ; G06N3/063

Abstract:
In one example, a method comprises: receiving input codes, wherein the input codes represent a computational dataflow graph; traversing the computational dataflow graph to identify single-entry-single-exit (SESE) subgraphs of the computational dataflow graph, wherein each SESE subgraph has a sequence of nodes comprising a root node and a child node and representing a sequence of element-wise operators, wherein the root node receives a single input tensor, and wherein the child node outputs a single output tensor; determining a merged operator for each SESE subgraph; and generating executable instructions for the computational dataflow graph to be executed by a hardware accelerator having a first execution unit and a second execution unit, wherein the executable instructions comprise first executable instructions for the merged operators targeted at the first execution unit, and second executable instructions for other operators of the computational dataflow graph targeted at the second execution unit.
Information query