ALEXANDRIA, Va., June 12 -- United States Patent no. 12,299,577, issued on May 13, was assigned to NVIDIA Corp. (Santa Clara, Calif.).
"Tensor processing using low precision format" was invented by Boris Ginsburg (Santa Clara, Calif.), Sergei Nikolaev (Santa Clara, Calif.), Ahmad Kiswani (Santa Clara, Calif.), Hao Wu (Hanzhou, China), Amir Gholaminejad (Santa Clara, Calif.), Slawomir Kierat (Mountain View, Calif.), Michael Houston (Saratoga, Calif.) and Alex Fit-Florea (Belmont, Calif.).
According to the abstract* released by the U.S. Patent & Trademark Office: "Aspects of the present invention are directed to computer-implemented techniques for improving the training of artificial neural networks using a reduced precision (e.g., float16)...