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): deu eng fra por spa
- Target Language(s): bas bem bnt bss cce cjk cwe dig dug gog gwr hay heh her jmc kam kdc kdn kik kin kki kkj kmb kng kon ksb kua ldi lem lin lon lsm lua lug luy mcp myx nbl nde ndo nim nnb nso nuj nya nyf nyn nyo nyy old ozm pkb rim run seh sna sot ssw suk swa swc swh sxb thk tlj toh toi tsn tso tum umb ven vmw vun wmw xho xog zul
- Valid Target Language Labels: >>abb<< >>agh<< >>akw<< >>asa<< >>auh<< >>axk<< >>baf<< >>bag<< >>bas<< >>bbg<< >>bbi<< >>bbm<< >>bcp<< >>bdp<< >>bdu<< >>beb<< >>bem<< >>beq<< >>bez<< >>bhy<< >>bip<< >>biw<< >>biz<< >>bja<< >>bkf<< >>bkh<< >>bkj<< >>bkp<< >>bkt<< >>bkw<< >>bli<< >>blv<< >>bmb<< >>bmg<< >>bml<< >>bmw<< >>bng<< >>bni<< >>bnm<< >>bnt_Latn<< >>bnx<< >>boh<< >>bok<< >>bou<< >>boy<< >>bpj<< >>bqm<< >>bqu<< >>bqz<< >>brf<< >>bri<< >>brl<< >>bsi<< >>bss<< >>btb<< >>btc<< >>buf<< >>bui<< >>bum<< >>buu<< >>buw<< >>bvb<< >>bvg<< >>bvx<< >>bwc<< >>bwg<< >>bwl<< >>bws<< >>bwt<< >>bww<< >>bwz<< >>bxc<< >>bxg<< >>bxp<< >>byi<< >>bzm<< >>bzo<< >>cce<< >>ccl<< >>cgg<< >>chw<< >>cjk<< >>cjk_Latn<< >>coh<< >>cuh<< >>cwa<< >>cwb<< >>cwe<< >>dav<< >>dde<< >>dez<< >>dhm<< >>dhs<< >>dig<< >>dii<< >>diu<< >>diz<< >>dma<< >>dmx<< >>dne<< >>doe<< >>dov<< >>dua<< >>dug<< >>dzn<< >>ebo<< >>ebu<< >>ekm<< >>eko<< >>eto<< >>ewo<< >>fan<< >>fip<< >>flr<< >>fwe<< >>gev<< >>gey<< >>gmx<< >>gog<< >>guz<< >>gwe<< >>gwr<< >>gyi<< >>han<< >>haq<< >>hav<< >>hay<< >>hba<< >>heh<< >>hem<< >>her<< >>hij<< >>hka<< >>hke<< >>hol<< >>hom<< >>hoo<< >>hum<< >>ifm<< >>ikz<< >>ilb<< >>isn<< >>iyx<< >>jgb<< >>jit<< >>jmc<< >>job<< >>kam<< >>kbj<< >>kbs<< >>kck<< >>kcu<< >>kcv<< >>kcw<< >>kcz<< >>kdc<< >>kde<< >>kdg<< >>kdn<< >>keb<< >>ked<< >>khu<< >>khx<< >>khy<< >>kik<< >>kin<< >>kiv<< >>kiz<< >>kki<< >>kkj<< >>kkq<< >>kkw<< >>kmb<< >>kme<< >>kmw<< >>kng<< >>kny<< >>koh<< >>kon<< >>koo<< >>koq<< >>kqn<< >>ksb<< >>ksf<< >>ksv<< >>ktf<< >>ktu<< >>kty<< >>kua<< >>kuj<< >>kwc<< >>kwm<< >>kwn<< >>kws<< >>kwu<< >>kxx<< >>kya<< >>kzn<< >>kzo<< >>kzy<< >>lag<< >>lai<< >>lam<< >>lch<< >>ldi<< >>lea<< >>leb<< >>leh<< >>lej<< >>lel<< >>lem<< >>leo<< >>lfa<< >>lgm<< >>lgz<< >>lie<< >>lik<< >>lin<< >>liz<< >>lke<< >>llb<< >>lli<< >>lnb<< >>lol<< >>lon<< >>loo<< >>loq<< >>loz<< >>lse<< >>lsm<< >>lua<< >>lub<< >>lue<< >>lug<< >>luj<< >>lum<< >>lun<< >>lup<< >>luy<< >>lwa<< >>lyn<< >>mbm<< >>mbo<< >>mck<< >>mcp<< >>mcx<< >>mdn<< >>mdp<< >>mdq<< >>mdt<< >>mdu<< >>mdw<< >>mer<< >>mfu<< >>mgg<< >>mgh<< >>mgq<< >>mgr<< >>mgs<< >>mgv<< >>mgw<< >>mgy<< >>mgz<< >>mhb<< >>mhm<< >>mho<< >>mhw<< >>mjh<< >>mkk<< >>mkw<< >>mlb<< >>mlk<< >>mmu<< >>mmz<< >>mny<< >>mow<< >>mpa<< >>mvw<< >>mwe<< >>mwn<< >>mws<< >>mwz<< >>mxc<< >>mxg<< >>mxo<< >>myc<< >>mye<< >>myx<< >>mzd<< >>nba<< >>nbd<< >>nbl<< >>nda<< >>ndc<< >>nde<< >>ndg<< >>ndh<< >>ndj<< >>ndk<< >>ndl<< >>ndn<< >>ndo<< >>ndq<< >>ndw<< >>ngc<< >>ngd<< >>ngl<< >>ngo<< >>ngp<< >>ngq<< >>ngy<< >>ngz<< >>nih<< >>nim<< >>nix<< >>njx<< >>njy<< >>nka<< >>nkc<< >>nkn<< >>nkt<< >>nkv<< >>nkw<< >>nlj<< >>nlo<< >>nmd<< >>nmg<< >>nmq<< >>nnb<< >>nnb_Latn<< >>nne<< >>nnq<< >>noq<< >>now<< >>nql<< >>nra<< >>nse<< >>nso<< >>nsx<< >>nte<< >>ntk<< >>nto<< >>nui<< >>nuj<< >>nvo<< >>nxd<< >>nxi<< >>nxo<< >>nya<< >>nyc<< >>nye<< >>nyf<< >>nyg<< >>nyj<< >>nyk<< >>nym<< >>nyn<< >>nyo<< >>nyr<< >>nyu<< >>nyy<< >>nzb<< >>nzd<< >>old<< >>olu<< >>oml<< >>ozm<< >>pae<< >>pbr<< >>pem<< >>phm<< >>pic<< >>piw<< >>pkb<< >>pmm<< >>pof<< >>poy<< >>puu<< >>reg<< >>rim<< >>rnd<< >>rng<< >>rnw<< >>rof<< >>rub<< >>ruc<< >>ruf<< >>run<< >>rwk<< >>rwm<< >>sak<< >>sbk<< >>sbm<< >>sbp<< >>sbs<< >>sbw<< >>sby<< >>sdj<< >>seg<< >>seh<< >>sgm<< >>shc<< >>shq<< >>shr<< >>sie<< >>skt<< >>slx<< >>smd<< >>smx<< >>sna<< >>sng<< >>snq<< >>soc<< >>sod<< >>soe<< >>soo<< >>sop<< >>sot<< >>sox<< >>soz<< >>ssc<< >>ssw<< >>sub<< >>suj<< >>suk<< >>suw<< >>swa<< >>swb<< >>swc<< >>swh<< >>swj<< >>swk<< >>sxb<< >>sxe<< >>syi<< >>syx<< >>szg<< >>szv<< >>tap<< >>tbt<< >>tck<< >>teg<< >>tek<< >>tga<< >>thk<< >>tii<< >>tke<< >>tlj<< >>tll<< >>tmv<< >>tny<< >>tog<< >>toh<< >>toi<< >>toi_Latn<< >>tsa<< >>tsc<< >>tsn<< >>tso<< >>tsv<< >>ttf<< >>ttj<< >>ttl<< >>tum<< >>tvs<< >>tvu<< >>twl<< >>two<< >>twx<< >>tyi<< >>tyx<< >>ukh<< >>umb<< >>vau<< >>ven<< >>vid<< >>vif<< >>vin<< >>vmk<< >>vmr<< >>vmw<< >>vum<< >>vun<< >>wbh<< >>wbi<< >>wdd<< >>wlc<< >>wmw<< >>wni<< >>won<< >>wum<< >>wun<< >>xdo<< >>xho<< >>xku<< >>xkv<< >>xma<< >>xmc<< >>xog<< >>xsq<< >>yaf<< >>yao<< >>yas<< >>yat<< >>yav<< >>yel<< >>yey<< >>yko<< >>ymk<< >>yns<< >>yom<< >>zaj<< >>zak<< >>zdj<< >>zga<< >>zin<< >>zmb<< >>zmf<< >>zmn<< >>zmp<< >>zmq<< >>zms<< >>zmw<< >>zmx<< >>zul<<
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. >>bas<<

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

model_name = "pytorch-models/opus-mt-tc-bible-big-deu_eng_fra_por_spa-bnt"
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-deu_eng_fra_por_spa-bnt")
print(pipe(">>bas<< 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
eng-run	tatoeba-test-v2021-08-07	0.44207	11.8	1703	6710
eng-swa	tatoeba-test-v2021-08-07	0.60298	32.7	387	1888
fra-run	tatoeba-test-v2021-08-07	0.42664	11.2	1274	5081
spa-run	tatoeba-test-v2021-08-07	0.41921	10.5	963	3886
eng-lin	flores101-devtest	0.43748	13.2	1012	26769
eng-nso	flores101-devtest	0.47122	19.4	1012	31298
eng-sna	flores101-devtest	0.44294	9.4	1012	20105
eng-xho	flores101-devtest	0.50110	11.6	1012	18227
fra-sna	flores101-devtest	0.40676	6.2	1012	20105
por-lin	flores101-devtest	0.41675	10.7	1012	26769
spa-lin	flores101-devtest	0.40631	8.8	1012	26769
deu-lin	flores200-devtest	0.40763	9.9	1012	26769
deu-xho	flores200-devtest	0.40586	4.8	1012	18227
eng-kin	flores200-devtest	0.41492	11.1	1012	22774
eng-lin	flores200-devtest	0.45568	14.7	1012	26769
eng-nso	flores200-devtest	0.48626	20.8	1012	31298
eng-nya	flores200-devtest	0.45067	10.7	1012	22180
eng-sna	flores200-devtest	0.45629	10.1	1012	20105
eng-sot	flores200-devtest	0.45331	15.4	1012	31600
eng-ssw	flores200-devtest	0.43635	7.1	1012	18508
eng-tsn	flores200-devtest	0.45233	17.7	1012	33831
eng-tso	flores200-devtest	0.48529	18.3	1012	29548
eng-xho	flores200-devtest	0.51974	13.1	1012	18227
eng-zul	flores200-devtest	0.53320	14.0	1012	18556
fra-lin	flores200-devtest	0.44410	13.0	1012	26769
fra-sna	flores200-devtest	0.42053	6.9	1012	20105
fra-xho	flores200-devtest	0.44537	7.1	1012	18227
fra-zul	flores200-devtest	0.41291	5.7	1012	18556
por-lin	flores200-devtest	0.42944	11.7	1012	26769
por-xho	flores200-devtest	0.41363	5.8	1012	18227
spa-lin	flores200-devtest	0.41938	9.4	1012	26769
deu-swa	ntrex128	0.48979	18.0	1997	46859
deu-tsn	ntrex128	0.41894	15.4	1997	71271
eng-nya	ntrex128	0.46801	14.9	1997	43727
eng-ssw	ntrex128	0.42880	6.7	1997	36169
eng-swa	ntrex128	0.60117	33.4	1997	46859
eng-tsn	ntrex128	0.46599	22.2	1997	71271
eng-xho	ntrex128	0.48847	11.2	1997	35439
eng-zul	ntrex128	0.49764	10.7	1997	34438
fra-swa	ntrex128	0.45494	17.5	1997	46859
fra-tsn	ntrex128	0.41426	15.3	1997	71271
fra-xho	ntrex128	0.41206	5.2	1997	35439
por-swa	ntrex128	0.46465	18.0	1997	46859
por-tsn	ntrex128	0.40236	14.5	1997	71271
por-xho	ntrex128	0.40070	5.0	1997	35439
spa-swa	ntrex128	0.46670	18.1	1997	46859
spa-tsn	ntrex128	0.40263	14.2	1997	71271
spa-xho	ntrex128	0.40247	4.9	1997	35439
eng-kin	tico19-test	0.40952	11.3	2100	55034
eng-lin	tico19-test	0.44670	15.5	2100	61116
eng-swa	tico19-test	0.56798	28.0	2100	58846
eng-zul	tico19-test	0.53624	14.4	2100	44098
fra-swa	tico19-test	0.44926	16.8	2100	58846
fra-zul	tico19-test	0.40588	6.0	2100	44098
por-lin	tico19-test	0.41729	12.5	2100	61116
por-swa	tico19-test	0.49303	19.6	2100	58846
spa-lin	tico19-test	0.41645	12.1	2100	61116
spa-swa	tico19-test	0.48614	18.8	2100	58846
spa-zul	tico19-test	0.40058	5.3	2100	44098

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 09:00:33 EEST 2024
port machine: LM0-400-22516.local

Helsinki-NLP/opus-mt-tc-bible-big-deu_eng_fra_por_spa-bnt

opus-mt-tc-bible-big-deu_eng_fra_por_spa-bnt

Table of Contents

Model Details