Speculative sampling: LLMs writing a lot faster using smaller LLMsblog.dust.tt4 points·spolu··0 commentsOpen articleSaveView on HN