arXivDaily arXiv每日学术速递 周一至周五更新
看起来今天可以划水了
arXiv 2006.12467cs.LGcs.CLstat.ML
收藏

自注意力的深度-宽度相互作用

The Depth-to-Width Interplay in Self-Attention

The Hebrew University of Jerusalem(耶路撒冷希伯来大学)

Yoav Levine, Noam Wies, Or Sharir, Hofit Bata, Amnon Shashua

摘要:

自注意力架构正在迅速推动自然语言处理的前沿,它表现出令人惊讶的深度效率低下的行为:以前的工作表明,增加内部表示(网络宽度)和增加自注意力层的数量(网络深度)一样有用。我们从理论上预测了自注意力的深度效率和深度效率低下之间的宽度依赖的过渡。我们对深度为6至48的网络进行了系统的经验消融,清楚地揭示了理论预测的行为,并提供了关于给定自注意力网络大小的最佳深度-宽度分配的明确定量建议。超越1-Triangle参数语言模型的竞赛使得增加自注意力力深度和宽度的明智指南成为一个必不可少的因素。我们的指南阐明了自注意力网络的深度与宽度的权衡,其规模高达GPT 3(我们预计其规模太深),甚至更高,标志着前所未有的30 K宽度对于1-Triangle参数网络来说是最佳的。

英文摘要:

Self-attention architectures, which are rapidly pushing the frontier in natural language processing, demonstrate a surprising depth-inefficient behavior: previous works indicate that increasing the internal representation (network width) is just as useful as increasing the number of self-attention layers (network depth). We theoretically predict a width-dependent transition between depth-efficiency and depth-inefficiency in self-attention. We conduct systematic empirical ablations on networks of depths 6 to 48 that clearly reveal the theoretically predicted behaviors, and provide explicit quantitative suggestions regarding the optimal depth-to-width allocation for a given self-attention network size. The race towards beyond 1-Trillion parameter language models renders informed guidelines for increasing self-attention depth and width in tandem an essential ingredient. Our guidelines elucidate the depth-to-width trade-off in self-attention networks of sizes up to the scale of GPT3 (which we project to be too deep for its size), and beyond, marking an unprecedented width of 30K as optimal for a 1-Trillion parameter network.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑