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

Increased focus on the computational efficiency of NLP systems has motivatedthe design of efficient model architectures and improvements to underlyinghardware accelerators. However, the resulting increases in computationalthroughput and reductions in floating point operations have not directlytranslated to improvements in wall-clock inference latency. We demonstrate thatthese discrepancies can be largely attributed to bottlenecks introduced by deeplearning frameworks. We denote this phenomenon as the \textit{framework tax},and observe that the disparity is growing as hardware speed increases overtime. In this work, we examine this phenomenon through a series of case studiesanalyzing the effects of model design decisions, framework paradigms, andhardware platforms on total model latency. Code is available athttps://github.com/JaredFern/Framework-Tax.

The Framework Tax: Disparities Between Inference Efficiency in NLP Research and Deployment