This model is part of the OPUS-MT project, an effort to make neural machine translation models widely available and accessible for many languages in the world. All models are originally trained using the amazing framework of Marian NMT, an efficient NMT implementation written in pure C++. The models have been converted to pyTorch using the transformers library by huggingface. Training data is taken from OPUS and training pipelines use the procedures of OPUS-MT-train. Model Description:

Developed by: Language Technology Research Group at the University of Helsinki
Model Type: Translation (transformer-big)
Release: 2024-05-30
License: Apache-2.0
Language(s):
- Source Language(s): chm est fin fkv hun izh koi kom kpv krl liv mdf mrj myv sma sme smn udm vep vot vro
- Target Language(s): deu eng fra por spa
- Valid Target Language Labels: >>deu<< >>eng<< >>fra<< >>por<< >>spa<< >>xxx<<
Original Model: opusTCv20230926max50+bt+jhubc_transformer-big_2024-05-30.zip
Resources for more information:

This is a multilingual translation model with multiple target languages. A sentence initial language token is required in the form of >>id<< (id = valid target language ID), e.g. >>deu<<

Uses

This model can be used for translation and text-to-text generation.

Risks, Limitations and Biases

CONTENT WARNING: Readers should be aware that the model is trained on various public data sets that may contain content that is disturbing, offensive, and can propagate historical and current stereotypes.

Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)).

How to Get Started With the Model

A short example code:

python

from transformers import MarianMTModel, MarianTokenizer

src_text = [
    ">>deu<< Replace this with text in an accepted source language.",
    ">>spa<< This is the second sentence."
]

model_name = "pytorch-models/opus-mt-tc-bible-big-fiu-deu_eng_fra_por_spa"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))

for t in translated:
    print( tokenizer.decode(t, skip_special_tokens=True) )

You can also use OPUS-MT models with the transformers pipelines, for example:

python

from transformers import pipeline
pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-bible-big-fiu-deu_eng_fra_por_spa")
print(pipe(">>deu<< Replace this with text in an accepted source language."))

Training

Data: opusTCv20230926max50+bt+jhubc (source)
Pre-processing: SentencePiece (spm32k,spm32k)
Model Type: transformer-big
Original MarianNMT Model: opusTCv20230926max50+bt+jhubc_transformer-big_2024-05-30.zip
Training Scripts: GitHub Repo

Evaluation

Model scores at the OPUS-MT dashboard
test set translations: opusTCv20230926max50+bt+jhubc_transformer-big_2024-05-29.test.txt
test set scores: opusTCv20230926max50+bt+jhubc_transformer-big_2024-05-29.eval.txt
benchmark results: benchmark_results.txt
benchmark output: benchmark_translations.zip

langpair	testset	chr-F	BLEU	#sent	#words
est-deu	tatoeba-test-v2021-08-07	0.69451	53.9	244	1611
est-eng	tatoeba-test-v2021-08-07	0.72437	58.2	1359	8811
fin-deu	tatoeba-test-v2021-08-07	0.66025	47.3	2647	19163
fin-eng	tatoeba-test-v2021-08-07	0.69685	53.7	10690	80552
fin-fra	tatoeba-test-v2021-08-07	0.65900	48.3	1920	12193
fin-por	tatoeba-test-v2021-08-07	0.72250	54.0	477	3021
fin-spa	tatoeba-test-v2021-08-07	0.69600	52.1	2513	16912
hun-deu	tatoeba-test-v2021-08-07	0.62418	41.1	15342	127344
hun-eng	tatoeba-test-v2021-08-07	0.65626	48.7	13037	94699
hun-fra	tatoeba-test-v2021-08-07	0.66840	50.3	2494	16914
hun-por	tatoeba-test-v2021-08-07	0.65281	43.1	2500	16563
hun-spa	tatoeba-test-v2021-08-07	0.67467	48.7	2500	16670
est-deu	flores101-devtest	0.55353	25.7	1012	25094
est-eng	flores101-devtest	0.61930	34.7	1012	24721
est-fra	flores101-devtest	0.58199	31.3	1012	28343
est-por	flores101-devtest	0.54388	26.5	1012	26519
fin-eng	flores101-devtest	0.59914	32.2	1012	24721
fin-por	flores101-devtest	0.55156	27.1	1012	26519
hun-eng	flores101-devtest	0.61198	33.5	1012	24721
hun-fra	flores101-devtest	0.57776	30.8	1012	28343
hun-por	flores101-devtest	0.56263	28.4	1012	26519
hun-spa	flores101-devtest	0.49140	20.7	1012	29199
est-deu	flores200-devtest	0.55825	26.3	1012	25094
est-eng	flores200-devtest	0.62404	35.4	1012	24721
est-fra	flores200-devtest	0.58580	31.7	1012	28343
est-por	flores200-devtest	0.55070	27.3	1012	26519
est-spa	flores200-devtest	0.50188	21.5	1012	29199
fin-deu	flores200-devtest	0.54281	24.0	1012	25094
fin-eng	flores200-devtest	0.60642	33.1	1012	24721
fin-fra	flores200-devtest	0.57540	30.5	1012	28343
fin-por	flores200-devtest	0.55497	27.4	1012	26519
fin-spa	flores200-devtest	0.49847	21.4	1012	29199
hun-deu	flores200-devtest	0.55180	25.1	1012	25094
hun-eng	flores200-devtest	0.61466	34.0	1012	24721
hun-fra	flores200-devtest	0.57670	30.6	1012	28343
hun-por	flores200-devtest	0.56510	28.9	1012	26519
hun-spa	flores200-devtest	0.49681	21.3	1012	29199
hun-deu	newssyscomb2009	0.49819	17.9	502	11271
hun-eng	newssyscomb2009	0.52063	24.4	502	11818
hun-fra	newssyscomb2009	0.51589	22.0	502	12331
hun-spa	newssyscomb2009	0.51508	22.7	502	12503
hun-deu	newstest2008	0.50164	19.0	2051	47447
hun-eng	newstest2008	0.49802	20.4	2051	49380
hun-fra	newstest2008	0.51012	21.6	2051	52685
hun-spa	newstest2008	0.50719	22.3	2051	52586
hun-deu	newstest2009	0.49902	18.6	2525	62816
hun-eng	newstest2009	0.50950	22.3	2525	65399
hun-fra	newstest2009	0.50742	21.6	2525	69263
hun-spa	newstest2009	0.50788	22.2	2525	68111
fin-eng	newstest2015	0.55249	27.0	1370	27270
fin-eng	newstest2016	0.57961	30.7	3000	62945
fin-eng	newstest2017	0.59973	33.2	3002	61846
est-eng	newstest2018	0.59190	31.5	2000	45405
fin-eng	newstest2018	0.52373	24.4	3000	62325
fin-eng	newstest2019	0.57079	30.3	1996	36215
fin-eng	newstestB2017	0.56420	28.9	3002	61846
est-deu	ntrex128	0.51377	21.4	1997	48761
est-eng	ntrex128	0.58358	29.9	1997	47673
est-fra	ntrex128	0.52713	24.9	1997	53481
est-por	ntrex128	0.50745	22.2	1997	51631
est-spa	ntrex128	0.54304	27.5	1997	54107
fin-deu	ntrex128	0.50282	19.8	1997	48761
fin-eng	ntrex128	0.55545	26.3	1997	47673
fin-fra	ntrex128	0.50946	22.9	1997	53481
fin-por	ntrex128	0.50404	21.3	1997	51631
fin-spa	ntrex128	0.52641	25.5	1997	54107
hun-deu	ntrex128	0.49322	18.5	1997	48761
hun-eng	ntrex128	0.52964	23.3	1997	47673
hun-fra	ntrex128	0.49800	21.8	1997	53481
hun-por	ntrex128	0.48941	20.5	1997	51631
hun-spa	ntrex128	0.51123	24.2	1997	54107

Citation Information

Publications: Democratizing neural machine translation with OPUS-MT and OPUS-MT – Building open translation services for the World and The Tatoeba Translation Challenge – Realistic Data Sets for Low Resource and Multilingual MT (Please, cite if you use this model.)

bibtex

@article{tiedemann2023democratizing,
  title={Democratizing neural machine translation with {OPUS-MT}},
  author={Tiedemann, J{\"o}rg and Aulamo, Mikko and Bakshandaeva, Daria and Boggia, Michele and Gr{\"o}nroos, Stig-Arne and Nieminen, Tommi and Raganato, Alessandro and Scherrer, Yves and Vazquez, Raul and Virpioja, Sami},
  journal={Language Resources and Evaluation},
  number={58},
  pages={713--755},
  year={2023},
  publisher={Springer Nature},
  issn={1574-0218},
  doi={10.1007/s10579-023-09704-w}
}

@inproceedings{tiedemann-thottingal-2020-opus,
    title = "{OPUS}-{MT} {--} Building open translation services for the World",
    author = {Tiedemann, J{\"o}rg  and Thottingal, Santhosh},
    booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
    month = nov,
    year = "2020",
    address = "Lisboa, Portugal",
    publisher = "European Association for Machine Translation",
    url = "https://aclanthology.org/2020.eamt-1.61",
    pages = "479--480",
}

@inproceedings{tiedemann-2020-tatoeba,
    title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
    author = {Tiedemann, J{\"o}rg},
    booktitle = "Proceedings of the Fifth Conference on Machine Translation",
    month = nov,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.wmt-1.139",
    pages = "1174--1182",
}

Acknowledgements

The work is supported by the HPLT project, funded by the European Union’s Horizon Europe research and innovation programme under grant agreement No 101070350. We are also grateful for the generous computational resources and IT infrastructure provided by CSC -- IT Center for Science, Finland, and the EuroHPC supercomputer LUMI.

Model conversion info

transformers version: 4.45.1
OPUS-MT git hash: 0882077
port time: Tue Oct 8 10:53:49 EEST 2024
port machine: LM0-400-22516.local

WindstormLabs/origin-Helsinki-NLP-opus-mt-tc-bible-big-fiu-deu_eng_fra_por_spa

opus-mt-tc-bible-big-fiu-deu_eng_fra_por_spa

Table of Contents

Model Details