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/

Global: Zendy Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We introduce a Power-of-Two low-bit post-training quantization(PTQ) methodfor deep neural network that meets hardware requirements and does not call forlong-time retraining. Power-of-Two quantization can convert the multiplicationintroduced by quantization and dequantization to bit-shift that is adopted bymany efficient accelerators. However, the Power-of-Two scale factors have fewercandidate values, which leads to more rounding or clipping errors. We propose anovel Power-of-Two PTQ framework, dubbed RAPQ, which dynamically adjusts thePower-of-Two scales of the whole network instead of statically determining themlayer by layer. It can theoretically trade off the rounding error and clippingerror of the whole network. Meanwhile, the reconstruction method in RAPQ isbased on the BN information of every unit. Extensive experiments on ImageNetprove the excellent performance of our proposed method. Without bells andwhistles, RAPQ can reach accuracy of 65% and 48% on ResNet-18 and MobileNetV2respectively with weight INT2 activation INT4. We are the first to propose themore constrained but hardware-friendly Power-of-Two quantization scheme forlow-bit PTQ specially and prove that it can achieve nearly the same accuracy asSOTA PTQ method. The code was released.

RAPQ: Rescuing Accuracy for Power-of-Two Low-bit Post-training Quantization