Sep 19, 2023

Retentive Network: A Successor to Transformer for Large Language Models (Paper Explained)

Posted by in category: computing

Retention is an alternative to Attention in Transformers that can both be written in a parallel and in a recurrent fashion. This means the architecture achieves training parallelism while maintaining low-cost inference. Experiments in the paper look very promising.

0:00 — Intro.
2:40 — The impossible triangle.
6:55 — Parallel vs sequential.
15:35 — Retention mechanism.
21:00 — Chunkwise and multi-scale retention.
24:10 — Comparison to other architectures.
26:30 — Experimental evaluation.


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 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 this https URL.

Authors: yutao sun, li dong, shaohan huang, shuming ma, yuqing xia, jilong xue, jianyong wang, furu wei.


Leave a reply