Skip to Main content Skip to Navigation
Conference papers

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 : machine translation
Document type :
Conference papers
Complete list of metadata

Cited literature [17 references]  Display  Hide  Download

https://hal.inria.fr/hal-02422914
Contributor : Maha Elbayad Connect in order to contact the contributor
Submitted on : Monday, December 23, 2019 - 1:52:16 PM
Last modification on : Tuesday, May 11, 2021 - 11:37:25 AM
Long-term archiving on: : Tuesday, March 24, 2020 - 1:27:28 PM

File

paper.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02422914, version 1

Citation

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⟩

Share

Metrics

Record views

399

Files downloads

1146