Standard
How far can we get with one GPU in 100 hours? CoAStaL at MultiIndicMT Shared Task. / Aralikatte, Rahul; Murrieta Bello, Héctor Ricardo; Hershcovich, Daniel; Bollmann, Marcel; Søgaard, Anders.
Proceedings of the 8th Workshop on Asian Translation (WAT2021). Association for Computational Linguistics, 2021. p. 205-211.
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Harvard
Aralikatte, R, Murrieta Bello, HR
, Hershcovich, D, Bollmann, M
& Søgaard, A 2021,
How far can we get with one GPU in 100 hours? CoAStaL at MultiIndicMT Shared Task. in
Proceedings of the 8th Workshop on Asian Translation (WAT2021). Association for Computational Linguistics, pp. 205-211, 8th Workshop on Asian Translation (WAT2021), Online,
05/08/2021.
https://doi.org/10.18653/v1/2021.wat-1.24
APA
Aralikatte, R., Murrieta Bello, H. R.
, Hershcovich, D., Bollmann, M.
, & Søgaard, A. (2021).
How far can we get with one GPU in 100 hours? CoAStaL at MultiIndicMT Shared Task. In
Proceedings of the 8th Workshop on Asian Translation (WAT2021) (pp. 205-211). Association for Computational Linguistics.
https://doi.org/10.18653/v1/2021.wat-1.24
Vancouver
Aralikatte R, Murrieta Bello HR
, Hershcovich D, Bollmann M
, Søgaard A.
How far can we get with one GPU in 100 hours? CoAStaL at MultiIndicMT Shared Task. In Proceedings of the 8th Workshop on Asian Translation (WAT2021). Association for Computational Linguistics. 2021. p. 205-211
https://doi.org/10.18653/v1/2021.wat-1.24
Author
Aralikatte, Rahul ; Murrieta Bello, Héctor Ricardo ; Hershcovich, Daniel ; Bollmann, Marcel ; Søgaard, Anders. / How far can we get with one GPU in 100 hours? CoAStaL at MultiIndicMT Shared Task. Proceedings of the 8th Workshop on Asian Translation (WAT2021). Association for Computational Linguistics, 2021. pp. 205-211
Bibtex
@inproceedings{c3d3230ac4a04faca8ee5677d0b6c2e0,
title = "How far can we get with one GPU in 100 hours?: CoAStaL at MultiIndicMT Shared Task",
abstract = "This work shows that competitive translation results can be obtained in a constrained setting by incorporating the latest advances in memory and compute optimization. We train and evaluate large multilingual translation models using a single GPU for a maximum of 100 hours and get within 4-5 BLEU points of the top submission on the leaderboard. We also benchmark standard baselines on the PMI corpus and re-discover well-known shortcomings of translation systems and metrics.",
author = "Rahul Aralikatte and {Murrieta Bello}, {H{\'e}ctor Ricardo} and Daniel Hershcovich and Marcel Bollmann and Anders S{\o}gaard",
year = "2021",
doi = "10.18653/v1/2021.wat-1.24",
language = "English",
pages = "205--211",
booktitle = "Proceedings of the 8th Workshop on Asian Translation (WAT2021)",
publisher = "Association for Computational Linguistics",
note = "8th Workshop on Asian Translation (WAT2021) ; Conference date: 05-08-2021 Through 06-08-2021",
}
RIS
TY - GEN
T1 - How far can we get with one GPU in 100 hours?
T2 - 8th Workshop on Asian Translation (WAT2021)
AU - Aralikatte, Rahul
AU - Murrieta Bello, Héctor Ricardo
AU - Hershcovich, Daniel
AU - Bollmann, Marcel
AU - Søgaard, Anders
PY - 2021
Y1 - 2021
N2 - This work shows that competitive translation results can be obtained in a constrained setting by incorporating the latest advances in memory and compute optimization. We train and evaluate large multilingual translation models using a single GPU for a maximum of 100 hours and get within 4-5 BLEU points of the top submission on the leaderboard. We also benchmark standard baselines on the PMI corpus and re-discover well-known shortcomings of translation systems and metrics.
AB - This work shows that competitive translation results can be obtained in a constrained setting by incorporating the latest advances in memory and compute optimization. We train and evaluate large multilingual translation models using a single GPU for a maximum of 100 hours and get within 4-5 BLEU points of the top submission on the leaderboard. We also benchmark standard baselines on the PMI corpus and re-discover well-known shortcomings of translation systems and metrics.
U2 - 10.18653/v1/2021.wat-1.24
DO - 10.18653/v1/2021.wat-1.24
M3 - Article in proceedings
SP - 205
EP - 211
BT - Proceedings of the 8th Workshop on Asian Translation (WAT2021)
PB - Association for Computational Linguistics
Y2 - 5 August 2021 through 6 August 2021
ER -