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

Large language models (LLMs) commonly employ autoregressive generation duringinference, leading to high memory bandwidth demand and consequently extendedlatency. To mitigate this inefficiency, we present Bi-directional Tuning forlossless Acceleration (BiTA), an innovative method expediting LLMs viastreamlined semi-autoregressive generation and draft verification. Inspired bythe concept of prompt tuning, we enhance LLMs with a parameter-efficient designcalled bi-directional tuning for the capability in semi-autoregressivegeneration. Employing efficient tree-based decoding, the models perform draftcandidate generation and verification in parallel, ensuring outputs identicalto their autoregressive counterparts under greedy sampling. BiTA serves as alightweight plug-in module, seamlessly boosting the inference efficiency ofexisting LLMs without requiring additional assistance models or incurringsignificant extra memory costs. Applying the proposed BiTA, LLaMA-2-70B-Chatachieves a 2.7$\times$ speedup on the MT-Bench benchmark. Extensive experimentsconfirm our method surpasses state-of-the-art acceleration techniques.

BiTA: Bi-Directional Tuning for Lossless Acceleration in Large Language Models