Fetching the paper…

Spectra: Surprising Effectiveness of Pretraining Ternary Language Models at Scale · Around