BriefGPT.xyz
Dec, 2022
FiDO：面向更强性能和更快推理优化的解码器融合
FiDO: Fusion-in-Decoder optimized for stronger performance and faster inference
HTML
PDF
Michiel de Jong, Yury Zemlyanskiy, Joshua Ainslie, Nicholas FitzGerald, Sumit Sanghai...
TL;DR
使用两项简单的变更加速FiD架构的推理速度，并允许更大的Decoder。我们称具有以上修改的FiD为FiDO，并表明它在各种推理预算范围内均表现出更好的性能。
Abstract
fusion-in-decoder
(FiD) is a powerful
retrieval-augmented language model
that sets the state-of-the-art on many knowledge-intensive NLP tasks. However, FiD suffers from very expensive
→