LSQ+: Improving low-bit quantization through learnable offsets and better initialization

The proposed method, called LSQ+, extends LSQ [7] by adding a simple yet effective learnable offset parameter for activation quantization to recover the lost accuracy on architectures employing Swish-like activations.

Data and Resources

Cite this as

Yash Bhalgat, Markus Nagel, Jinwon Lee, Tijmen Blankevoort, Nojun Kwak (2024). Dataset: LSQ+: Improving low-bit quantization through learnable offsets and better initialization. https://doi.org/10.57702/1baxurh3

DOI retrieved: December 16, 2024

Additional Info

Field Value
Created December 16, 2024
Last update December 16, 2024
Author Yash Bhalgat
More Authors
Markus Nagel
Jinwon Lee
Tijmen Blankevoort
Nojun Kwak