US 12,093,836 B2
Automatic multi-objective hardware optimization for processing of deep learning networks
Mattias Marder, Haifa (IL); Estelle Aflalo, Tel Aviv (IL); Avrech Ben-David, Haifa (IL); Shauharda Khadka, Lake Forest, WA (US); Somdeb Majumdar, Mission Viejo, CA (US); Santiago Miret, Oakland, CA (US); and Hanlin Tang, San Francisco, CA (US)
Assigned to INTEL CORPORATION, Santa Clara, CA (US)
Filed by Intel Corporation, Santa Clara, CA (US)
Filed on Dec. 21, 2020, as Appl. No. 17/129,521.
Prior Publication US 2021/0150371 A1, May 20, 2021
Int. Cl. G06N 3/10 (2006.01); G06N 3/045 (2023.01); G06N 3/086 (2023.01)
CPC G06N 3/10 (2013.01) [G06N 3/045 (2023.01); G06N 3/086 (2013.01)] 20 Claims
OG exemplary drawing
 
1. One or more non-transitory computer-readable storage mediums having stored thereon instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving an input including client preferences for a plurality of performance indicators for processing of a deep learning workload on a computing system, the computing system including one or more processors;
performing automatic hardware optimization for the processing of the deep learning workload, including:
generating, by the one or more processors, a workload representation for the deep learning workload, wherein generating the workload representation includes generating a graph neural network, and
providing the workload representation to machine learning processing by the one or more processors to generate, by the one or more processors, a workload executable, generating the workload executable including generating hardware mapping, the hardware mapping being generated based on the client preferences for the plurality of performance indicators; and
applying the workload executable in processing of the deep learning workload by the one or more processors, including implementing the hardware mapping for the processing of the deep learning workload.