SwitchHead: Accelerating Transformers with Mixture-of-Experts Attention
CoRR(2023)
摘要
The costly self-attention layers in modern Transformers require memory and
compute quadratic in sequence length. Existing approximation methods usually
underperform and fail to obtain significant speedups in practice. Here we
present SwitchHead - a novel method that reduces both compute and memory
requirements and achieves wall-clock speedup, while matching the language
modeling performance of baseline Transformers with the same parameter budget.
SwitchHead uses Mixture-of-Experts (MoE) layers for the value and output
projections and requires 4 to 8 times fewer attention matrices than standard
Transformers. Our novel attention can also be combined with MoE MLP layers,
resulting in an efficient fully-MoE "SwitchHead" Transformer model. Our code is
public.
更多查看译文
AI 理解论文
溯源树
样例
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要