Neural network activation compression with outlier block floating-point
US12045724B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Dec 31, 2018 |
| Grant date | Jul 23, 2024 |
| Priority date | — |
| Expiry date | Mar 1, 2039 |
Classification
- Technology area (CPC H)Electricity
- CPC primaryH03M7/3059
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
Apparatus and methods for training a neural network accelerator using quantized precision data formats having outlier values are disclosed, and in particular for storing activation values from a neural network in a compressed format for use during forward and backward propagation training of the neural network. In certain examples of the disclosed technology, a computing system is configured to perform forward propagation for a layer of a neural network to produced first activation values in a first block floating-point format. In some examples, activation values generated by forward propagation are converted by the compressor to a second block floating-point format having a narrower numerical precision than the first block floating-point format. Outlier values, comprising additional bits of mantissa and/or exponent are stored in ancillary storage for subset of the activation values. The compressed activation values are stored in the memory, where they can be retrieved for use during back propagation.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.