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

Efficient machine translation models are commercially important as they canincrease inference speeds, and reduce costs and carbon emissions. Recently,there has been much interest in non-autoregressive (NAR) models, which promisefaster translation. In parallel to the research on NAR models, there have beensuccessful attempts to create optimized autoregressive models as part of theWMT shared task on efficient translation. In this paper, we point out flaws inthe evaluation methodology present in the literature on NAR models and weprovide a fair comparison between a state-of-the-art NAR model and theautoregressive submissions to the shared task. We make the case for consistentevaluation of NAR models, and also for the importance of comparing NAR modelswith other widely used methods for improving efficiency. We run experimentswith a connectionist-temporal-classification-based (CTC) NAR model implementedin C++ and compare it with AR models using wall clock times. Our results showthat, although NAR models are faster on GPUs, with small batch sizes, they arealmost always slower under more realistic usage conditions. We call for morerealistic and extensive evaluation of NAR models in future work.

Non-Autoregressive Machine Translation: It's Not as Fast as it Seems