Back to results
US 2022/0092383 A1Published application
System and method for post-training quantization of deep neural networks with per-channel quantization mode selection
AssigneeSamsung Electronics Co., Ltd.
InventorsFatih CAKIR, Mostafa EL-KHAMY, Jungwon LEE
Life of this patent
Estimating the term…
Abstract
A method and system are provided. The method includes topologically sorting layers of a neural network, selecting a quantization process that utilizes a quantization of a previous layer, and determining, with the selected quantization process, a quantization mode of one layer in the neural network based on the quantization of a previous layer.
Claims20
A method, comprising:
- topologically sorting layers of a neural network;
- selecting a quantization process that utilizes a quantization of a previous layer;
- and
- determining, with the selected quantization process, a quantization mode of one layer in the neural network based on the quantization of a previous layer.
1 more element
A system, comprising:
- a memory;
- and
- a processor configured to:
- topologically sort layers of a neural network;
- select a quantization process that utilizes a quantization of a previous layer;
- and
- determine, with the selected quantization process, a quantization mode of one layer in the neural network based on the quantization of a previous layer.
4 more elements
Dependencies are inferred from the claim text.