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

Distilling state-of-the-art transformer models into lightweight studentmodels is an effective way to reduce computation cost at inference time. Thestudent models are typically compact transformers with fewer parameters, whileexpensive operations such as self-attention persist. Therefore, the improvedinference speed may still be unsatisfactory for real-time or high-volume usecases. In this paper, we aim to further push the limit of inference speed bydistilling teacher models into bigger, sparser student models -- bigger in thatthey scale up to billions of parameters; sparser in that most of the modelparameters are n-gram embeddings. Our experiments on six single-sentence textclassification tasks show that these student models retain 97% of theRoBERTa-Large teacher performance on average, and meanwhile achieve up to 600xspeed-up on both GPUs and CPUs at inference time. Further investigation revealsthat our pipeline is also helpful for sentence-pair classification tasks, andin domain generalization settings.

Sparse Distillation: Speeding Up Text Classification by Using Bigger Student Models