The patent badge is an abbreviated version of the USPTO patent document. The patent badge does contain a link to the full patent document.
The patent badge is an abbreviated version of the USPTO patent document. The patent badge covers the following: Patent number, Date patent was issued, Date patent was filed, Title of the patent, Applicant, Inventor, Assignee, Attorney firm, Primary examiner, Assistant examiner, CPCs, and Abstract. The patent badge does contain a link to the full patent document (in Adobe Acrobat format, aka pdf). To download or print any patent click here.
Patent No.:
Date of Patent:
Dec. 10, 2024
Filed:
Feb. 14, 2019
Microsoft Technology Licensing, Llc, Redmond, WA (US);
Daniel Lo, Bothell, WA (US);
Bita Darvish Rouhani, Bellevue, WA (US);
Eric S. Chung, Woodinville, WA (US);
Yiren Zhao, Cambridge, GB;
Amar Phanishayee, Seattle, WA (US);
Ritchie Zhao, Ithaca, NY (US);
Microsoft Technology Licensing, LLC, Redmond, WA (US);
Abstract
Apparatus and methods for training a neural network accelerator using quantized precision data formats are disclosed, and, in particular, for adjusting floating-point formats used to store activation values during training. In certain examples of the disclosed technology, a computing system includes processors, memory, and a floating-point compressor in communication with the memory. The computing system is configured to produce a neural network comprising activation values expressed in a first floating-point format, select a second floating-point format for the neural network based on a performance metric, convert at least one of the activation values to the second floating-point format, and store the compressed activation values in the memory. Aspects of the second floating-point format that can be adjusted include the number of bits used to express mantissas, exponent format, use of non-uniform mantissas, and/or use of outlier values to express some of the mantissas.