§ feed · storyline

Nous Research speeds up LLM pre-training by 2.5x with token

Nous Research releases Token Superposition Training, accelerating LLM pre-training by up to 2.5x across 270M–10B parameter models without architectural changes.

May 14 · 07:46:32 · primary fetch1 sourceupdated May 14 · 07:46:32

Nous Research released Token Superposition Training (TST), accelerating LLM pre-training by up to 2.5x across models from 270M to 10B parameters without changing architecture or data.

read full article on marktechpost.com ↗

§ sources1 publication · timeline below

marktechpost.comNous Research Releases Token Superposition Training to Speed Up LLM Pre-Training by Up to 2.5x Across 270M to 10B Parameter Modelsprimary07:46:32