English

IOT: Instance-wise Layer Reordering for Transformer Structures

Computation and Language 2021-03-08 v1 Artificial Intelligence

Abstract

With sequentially stacked self-attention, (optional) encoder-decoder attention, and feed-forward layers, Transformer achieves big success in natural language processing (NLP), and many variants have been proposed. Currently, almost all these models assume that the layer order is fixed and kept the same across data samples. We observe that different data samples actually favor different orders of the layers. Based on this observation, in this work, we break the assumption of the fixed layer order in the Transformer and introduce instance-wise layer reordering into the model structure. Our Instance-wise Ordered Transformer (IOT) can model variant functions by reordered layers, which enables each sample to select the better one to improve the model performance under the constraint of almost the same number of parameters. To achieve this, we introduce a light predictor with negligible parameter and inference cost to decide the most capable and favorable layer order for any input sequence. Experiments on 3 tasks (neural machine translation, abstractive summarization, and code generation) and 9 datasets demonstrate consistent improvements of our method. We further show that our method can also be applied to other architectures beyond Transformer. Our code is released at Github.

Keywords

Cite

@article{arxiv.2103.03457,
  title  = {IOT: Instance-wise Layer Reordering for Transformer Structures},
  author = {Jinhua Zhu and Lijun Wu and Yingce Xia and Shufang Xie and Tao Qin and Wengang Zhou and Houqiang Li and Tie-Yan Liu},
  journal= {arXiv preprint arXiv:2103.03457},
  year   = {2021}
}

Comments

Accepted at ICLR-2021

R2 v1 2026-06-23T23:47:08.212Z