Depth-adaptive Transformer - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2019

Depth-adaptive Transformer

Abstract

State of the art sequence-to-sequence models for large scale tasks perform a fixed number of computations for each input sequence regardless of whether it is easy or hard to process. In this paper, we train Transformer models which can make output predictions at different stages of the network and we investigate different ways to predict how much computation is required for a particular sequence. Unlike dynamic computation in Universal Transformers, which applies the same set of layers iteratively, we apply different layers at every step to adjust both the amount of computation as well as the model capacity. On IWSLT German-English translation our approach matches the accuracy of a well tuned baseline Transformer while using less than a quarter of the decoder layers.

Keywords

Fichier principal
Vignette du fichier
paper.pdf (332.46 Ko) Télécharger le fichier
Origin Files produced by the author(s)
Loading...

Dates and versions

hal-02422914 , version 1 (23-12-2019)

Identifiers

  • HAL Id : hal-02422914 , version 1

Cite

Maha Elbayad, Jiatao Gu, Edouard Grave, Michael Auli. Depth-adaptive Transformer. ICLR 2020 - Eighth International Conference on Learning Representations, Apr 2020, Addis Ababa, Ethiopia. pp.1-14. ⟨hal-02422914⟩
349 View
451 Download

Share

Gmail Mastodon Facebook X LinkedIn More