English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Recently low-precision deep learning accelerators (DLAs) have become populardue to their advantages in chip area and energy consumption, yet thelow-precision quantized models on these DLAs bring in severe accuracydegradation. One way to achieve both high accuracy and efficient inference isto deploy high-precision neural networks on low-precision DLAs, which is rarelystudied. In this paper, we propose the PArallel Low-precision Quantization(PalQuant) method that approximates high-precision computations via learningparallel low-precision representations from scratch. In addition, we present anovel cyclic shuffle module to boost the cross-group information communicationbetween parallel low-precision groups. Extensive experiments demonstrate thatPalQuant has superior performance to state-of-the-art quantization methods inboth accuracy and inference speed, e.g., for ResNet-18 network quantization,PalQuant can obtain 0.52\% higher accuracy and 1.78$\times$ speedupsimultaneously over their 4-bit counter-part on a state-of-the-art 2-bitaccelerator. Code is available at \url{https://github.com/huqinghao/PalQuant}.

PalQuant: Accelerating High-precision Networks on Low-precision Accelerators