Back to results
US 12,754,766 B2Granted
Neural network acceleration via graph partition
AssigneeBlack Sesame Technologies Inc.
InventorsTing Zhou, Fen Zhou, Yi Wang, Zexi Ye, Wei Zuo, Zheng Qi, Qiang Zhang
Life of this patent
Estimating the term…
Abstract
A method of constructing sub-graphs, includes receiving a directed acyclic graph (DAG), partitioning the directed acyclic graph into an at least one section, determining at least one hardware attribute, determining at least one DAG hardware limitation of the at least one section and determining a largest continuous node list of the at least one section in which the at least one hardware attribute meets the at least one DAG hardware limitation.
Claims10
A method of constructing sub-graphs, comprising:
- receiving a directed acyclic graph (DAG);
- partitioning the directed acyclic graph into an at least two sections;
- determining at least one hardware attribute;
- determining at least one DAG hardware limitation of the at least two sections;
- determining a largest continuous node list of the at least two sections in which the at least one hardware attribute meets the at least one DAG hardware limitation;
- executing the sections on a processor, wherein the section is a subnetwork of a neural network;
- determining if an at least one byte tensor move mask and weight (BTMW) of the at least two sections exceed a predetermined BTMW usage overflow;
- and
- updating the at least one hardware attribute with the at least one BTMW.
6 more elements
Dependencies are inferred from the claim text.