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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Neural network quantization is a promising compression technique to reducememory footprint and save energy consumption, potentially leading to real-timeinference. However, there is a performance gap between quantized andfull-precision models. To reduce it, existing quantization approaches requirehigh-precision INT32 or full-precision multiplication during inference forscaling or dequantization. This introduces a noticeable cost in terms ofmemory, speed, and required energy. To tackle these issues, we present F8Net, anovel quantization framework consisting of only fixed-point 8-bitmultiplication. To derive our method, we first discuss the advantages offixed-point multiplication with different formats of fixed-point numbers andstudy the statistical behavior of the associated fixed-point numbers. Second,based on the statistical and algorithmic analysis, we apply differentfixed-point formats for weights and activations of different layers. Weintroduce a novel algorithm to automatically determine the right format foreach layer during training. Third, we analyze a previous quantization algorithm-- parameterized clipping activation (PACT) -- and reformulate it usingfixed-point arithmetic. Finally, we unify the recently proposed method forquantization fine-tuning and our fixed-point approach to show the potential ofour method. We verify F8Net on ImageNet for MobileNet V1/V2 and ResNet18/50.Our approach achieves comparable and better performance, when compared not onlyto existing quantization techniques with INT32 multiplication or floating-pointarithmetic, but also to the full-precision counterparts, achievingstate-of-the-art performance.

F8Net: Fixed-Point 8-bit Only Multiplication for Network Quantization