计算机科学
推论
变压器
解码方法
元建模
语言模型
人工智能
算法
程序设计语言
量子力学
物理
电压
作者
Youxian Sun,Li Dong,Shaohan Huang,Shuming Ma,Yuqing Xia,Jingling Xue,Jianyong Wang,Furu Wei
出处
期刊:Cornell University - arXiv
日期:2023-01-01
被引量:1
标识
DOI:10.48550/arxiv.2307.08621
摘要
In this work, we propose Retentive Network (RetNet) as a foundation architecture for large language models, simultaneously achieving training parallelism, low-cost inference, and good performance. We theoretically derive the connection between recurrence and attention. Then we propose the retention mechanism for sequence modeling, which supports three computation paradigms, i.e., parallel, recurrent, and chunkwise recurrent. Specifically, the parallel representation allows for training parallelism. The recurrent representation enables low-cost $O(1)$ inference, which improves decoding throughput, latency, and GPU memory without sacrificing performance. The chunkwise recurrent representation facilitates efficient long-sequence modeling with linear complexity, where each chunk is encoded parallelly while recurrently summarizing the chunks. Experimental results on language modeling show that RetNet achieves favorable scaling results, parallel training, low-cost deployment, and efficient inference. The intriguing properties make RetNet a strong successor to Transformer for large language models. Code will be available at https://aka.ms/retnet.
科研通智能强力驱动
Strongly Powered by AbleSci AI