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

Neural network quantization is a very promising solution in the field ofmodel compression, but its resulting accuracy highly depends on atraining/fine-tuning process and requires the original data. This not onlybrings heavy computation and time costs but also is not conducive to privacyand sensitive information protection. Therefore, a few recent works arestarting to focus on data-free quantization. However, data-free quantizationdoes not perform well while dealing with ultra-low precision quantization.Although researchers utilize generative methods of synthetic data to addressthis problem partially, data synthesis needs to take a lot of computation andtime. In this paper, we propose a data-free mixed-precision compensation(DF-MPC) method to recover the performance of an ultra-low precision quantizedmodel without any data and fine-tuning process. By assuming the quantized errorcaused by a low-precision quantized layer can be restored via thereconstruction of a high-precision quantized layer, we mathematically formulatethe reconstruction loss between the pre-trained full-precision model and itslayer-wise mixed-precision quantized model. Based on our formulation, wetheoretically deduce the closed-form solution by minimizing the reconstructionloss of the feature maps. Since DF-MPC does not require any original/syntheticdata, it is a more efficient method to approximate the full-precision model.Experimentally, our DF-MPC is able to achieve higher accuracy for an ultra-lowprecision quantized model compared to the recent methods without any data andfine-tuning process.

Data-Free Quantization via Mixed-Precision Compensation without Fine-Tuning