MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki

Timothee Mickus; Stig-Arne Grönroos; Joseph Attieh; Michele Boggia; Ona De Gibert; Shaoxiong Ji; Niki Andreas Loppi; Alessandro Raganato; Raúl Vázquez; Jörg Tiedemann

MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki

Timothee Mickus, Stig-Arne Grönroos, Joseph Attieh, Michele Boggia, Ona De Gibert, Shaoxiong Ji, Niki Andreas Loppi, Alessandro Raganato, Raúl Vázquez, Jörg Tiedemann

Forskningsoutput: Kapitel i bok/rapport/konferenshandling › Konferensbidrag › Vetenskaplig › Peer review

Sammanfattning

NLP in the age of monolithic large language models is approaching its limits in terms of size and information that can be handled. The trend goes to modularization, a necessary step into the direction of designing smaller sub-networks and components with specialized functionality. In this paper, we present the MAMMOTH toolkit: a framework designed for training massively multilingual modular machine translation systems at scale, initially derived from OpenNMT-py and then adapted to ensure efficient training across computation clusters.We showcase its efficiency across clusters of A100 and V100 NVIDIA GPUs, and discuss our design philosophy and plans for future information.The toolkit is publicly available online at https://github.com/Helsinki-NLP/mammoth.

Originalspråk	engelska
Titel på värdpublikation	Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics : System Demonstrations
Redaktörer	Nikolaos Aletras, Orphee De Clercq
Antal sidor	10
Utgivningsort	Kerrville
Förlag	The Association for Computational Linguistics
Utgivningsdatum	1 mars 2024
Sidor	127-136
ISBN (elektroniskt)	979-8-89176-091-2
Status	Publicerad - 1 mars 2024
MoE-publikationstyp	A4 Artikel i en konferenspublikation
Evenemang	Conference of the European Chapter of the Association for Computational Linguistics - St. Julians, Malta Varaktighet: 17 mars 2024 → 22 mars 2024 Konferensnummer: 18

Vetenskapsgrenar

6121 Språkvetenskaper
113 Data- och informationsvetenskap

Åtkomst av dokument

2024.eacl-demo.14Slutlig publicerad version, 713 KBLicens: CC BY

https://aclanthology.org/2024.eacl-demo.14Licens: CC BY

FoTran: Found in Translation - Natural Language Understanding with Cross-Lingual Grounding
Tiedemann, J. (Projektledare), Attieh, J. (deltagare), Aulamo, M. (deltagare), Celikkanat, H. (deltagare), De Gibert Bonet, O. (deltagare), Grönroos, S.-A. (deltagare), Mickus, T. (deltagare), Raganato, A. (deltagare), Scherrer, Y. (deltagare), Silfverberg, M. (deltagare), Sjöblom, E. I. (deltagare), Talman, A. (deltagare), Vazquez , R. (deltagare), Virpioja, S. P. (deltagare), Yli-Jyrä, A. (deltagare), Zosa, E. (deltagare), Celikkanat, H. (Deltagare), Raganato, A. (Deltagare), Silfverberg, M. (Deltagare), Sulubacak, U. (Deltagare) & Vazquez , R. (Deltagare)
01/09/2018 → 31/03/2024
Projekt: EU Horizon 2020: European Research Council: Consolidator Grant (H2020-ERC-COG)

Citera det här

Mickus, T., Grönroos, S.-A., Attieh, J., Boggia, M., De Gibert, O., Ji, S., Loppi, N. A., Raganato, A., Vázquez, R., & Tiedemann, J. (2024). MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki. I N. Aletras, & O. De Clercq (Red.), Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations (s. 127-136). The Association for Computational Linguistics. https://aclanthology.org/2024.eacl-demo.14

Mickus, Timothee ; Grönroos, Stig-Arne ; Attieh, Joseph et al. / MAMMOTH : Massively Multilingual Modular Open Translation @ Helsinki. Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations. redaktör / Nikolaos Aletras ; Orphee De Clercq. Kerrville : The Association for Computational Linguistics, 2024. s. 127-136

@inproceedings{3988190b9f05430b81415b392b8dae86,

title = "MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki",

abstract = "NLP in the age of monolithic large language models is approaching its limits in terms of size and information that can be handled. The trend goes to modularization, a necessary step into the direction of designing smaller sub-networks and components with specialized functionality. In this paper, we present the MAMMOTH toolkit: a framework designed for training massively multilingual modular machine translation systems at scale, initially derived from OpenNMT-py and then adapted to ensure efficient training across computation clusters.We showcase its efficiency across clusters of A100 and V100 NVIDIA GPUs, and discuss our design philosophy and plans for future information.The toolkit is publicly available online at https://github.com/Helsinki-NLP/mammoth.",

keywords = "6121 Languages, 113 Computer and information sciences",

author = "Timothee Mickus and Stig-Arne Gr{\"o}nroos and Joseph Attieh and Michele Boggia and {De Gibert}, Ona and Shaoxiong Ji and Loppi, {Niki Andreas} and Alessandro Raganato and Ra{\'u}l V{\'a}zquez and J{\"o}rg Tiedemann",

year = "2024",

month = mar,

day = "1",

language = "English",

pages = "127--136",

editor = "Nikolaos Aletras and {De Clercq}, Orphee",

booktitle = "Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics",

publisher = "The Association for Computational Linguistics",

address = "United States",

note = "Conference of the European Chapter of the<br/>Association for Computational Linguistics, EACL 2024 ; Conference date: 17-03-2024 Through 22-03-2024",

}

Mickus, T , Grönroos, S-A , Attieh, J, Boggia, M, De Gibert, O , Ji, S, Loppi, NA, Raganato, A , Vázquez, R & Tiedemann, J 2024, MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki. i N Aletras & O De Clercq (red), Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations. The Association for Computational Linguistics, Kerrville, s. 127-136, Conference of the European Chapter of the
Association for Computational Linguistics, St. Julians, Malta, 17/03/2024. <https://aclanthology.org/2024.eacl-demo.14>

MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki. / Mickus, Timothee ; Grönroos, Stig-Arne ; Attieh, Joseph et al.
Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations. red. / Nikolaos Aletras; Orphee De Clercq. Kerrville: The Association for Computational Linguistics, 2024. s. 127-136.

Forskningsoutput: Kapitel i bok/rapport/konferenshandling › Konferensbidrag › Vetenskaplig › Peer review

TY - GEN

T1 - MAMMOTH

T2 - Conference of the European Chapter of the<br/>Association for Computational Linguistics

AU - Mickus, Timothee

AU - Grönroos, Stig-Arne

AU - Attieh, Joseph

AU - Boggia, Michele

AU - De Gibert, Ona

AU - Ji, Shaoxiong

AU - Loppi, Niki Andreas

AU - Raganato, Alessandro

AU - Vázquez, Raúl

AU - Tiedemann, Jörg

N1 - Conference code: 18

PY - 2024/3/1

Y1 - 2024/3/1

N2 - NLP in the age of monolithic large language models is approaching its limits in terms of size and information that can be handled. The trend goes to modularization, a necessary step into the direction of designing smaller sub-networks and components with specialized functionality. In this paper, we present the MAMMOTH toolkit: a framework designed for training massively multilingual modular machine translation systems at scale, initially derived from OpenNMT-py and then adapted to ensure efficient training across computation clusters.We showcase its efficiency across clusters of A100 and V100 NVIDIA GPUs, and discuss our design philosophy and plans for future information.The toolkit is publicly available online at https://github.com/Helsinki-NLP/mammoth.

AB - NLP in the age of monolithic large language models is approaching its limits in terms of size and information that can be handled. The trend goes to modularization, a necessary step into the direction of designing smaller sub-networks and components with specialized functionality. In this paper, we present the MAMMOTH toolkit: a framework designed for training massively multilingual modular machine translation systems at scale, initially derived from OpenNMT-py and then adapted to ensure efficient training across computation clusters.We showcase its efficiency across clusters of A100 and V100 NVIDIA GPUs, and discuss our design philosophy and plans for future information.The toolkit is publicly available online at https://github.com/Helsinki-NLP/mammoth.

KW - 6121 Languages

KW - 113 Computer and information sciences

M3 - Conference contribution

SP - 127

EP - 136

BT - Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics

A2 - Aletras, Nikolaos

A2 - De Clercq, Orphee

PB - The Association for Computational Linguistics

CY - Kerrville

Y2 - 17 March 2024 through 22 March 2024

ER -

Mickus T , Grönroos SA , Attieh J, Boggia M, De Gibert O , Ji S et al. MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki. I Aletras N, De Clercq O, redaktörer, Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations. Kerrville: The Association for Computational Linguistics. 2024. s. 127-136

MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki

Sammanfattning

Vetenskapsgrenar

Åtkomst av dokument

Projekt

FoTran: Found in Translation - Natural Language Understanding with Cross-Lingual Grounding

Citera det här