公开(公告)号：US20240144030A1

公开(公告)日：2024-05-02

申请号：US18279820

申请日：2022-06-08

Applicant: Intel Corporation

Inventor： Juan Pablo Muñoz ,  Nilesh Jain ,  Chaunté Lacewell ,  Alexander Kozlov ,  Nikolay Lyalyushkin ,  Vasily Shamporov ,  Anastasia Senina

IPC: G06N3/0985

CPC classification number: G06N3/0985

Abstract: Methods, apparatus, systems, and articles of manufacture to modify pre-trained models to apply neural architecture search are disclosed. Example instructions, when executed, cause processor circuitry to at least access a pre-trained machine learning model, create a super-network based on the pre-trained machine learning model, create a plurality of subnetworks based on the super-network, and search the plurality of subnetworks to select a subnetwork.

公开(公告)号：US20250028965A1

公开(公告)日：2025-01-23

申请号：US18904364

申请日：2024-10-02

Applicant: Intel Corporation

Inventor： Alexander Kozlov ,  Andrey Anufriev ,  Nikolay Lyalyushkin ,  Dmitry Gorokhov ,  Yury Gorbachev

IPC: G06N3/082

Abstract: Systems, apparatuses and methods may provide for technology that selects a subset of linear layers from a plurality of linear layers in a pre-trained artificial intelligence (AI) model, wherein a quantization error of the subset of linear layers exceeds an error threshold. For each linear layer in the subset of linear layers, the technology solves a singular value decomposition (SVD) approximation, generates a first adapter layer and a second adapter layer based on the SVD decomposition, wherein the first adapter layer and the second adapter layer include weight matrices having a first dimension that is less than a first rank threshold and a second dimension that is greater than a second rank threshold, and determines an inference output based on the linear layer, the first adapter layer and the second adapter layer.