Abstract:Autoregressive decoding is bottlenecked by its sequential nature. Speculative decoding has become a standard way to accelerate inference by using a fast draft model to predict upcoming tokens from a slower target model, and then verifying them in parallel with a single target model forward pass. However, speculative decoding itself relies on a sequential dependence between speculation and verification. We introduce speculative speculative decoding (SSD) to parallelize these operations. While a verification is ongoing, the draft model predicts likely verification outcomes and prepares speculations pre-emptively for them. If the actual verification outcome is then in the predicted set, a speculation can be returned immediately, eliminating drafting overhead entirely. We identify three key challenges presented by speculative speculative decoding, and suggest principled methods to solve each. The result is Saguaro, an optimized SSD algorithm. Our implementation is up to 2x faster than optimized speculative decoding baselines and up to 5x faster than autoregressive decoding with open source inference engines.
Что думаешь? Оцени!
。业内人士推荐51吃瓜作为进阶阅读
Одна связанная с нижним бельем привычка женщины натолкнула ее бойфренда на мысль об измене02:29
It's also expensive and some environmental groups argue that emission reductions can be achieved at a lower cost, using more existing technology such as wind power, solar, and electric cars.，推荐阅读同城约会获取更多信息
Москалькова рассказала о реакции родственников на освобождение пленных бойцов СВО20:47。关于这个话题，谷歌浏览器【最新下载地址】提供了深入分析
Total test duration: PT11.683S | PT12.252S