Hugging Face's logo

Helsinki-NLP
/
opus-mt-tc-big-gmw-gmw

Translation Transformers PyTorch TensorFlow Safetensors 10 languages marian text2text-generation opus-mt-tc Eval Results Inference Endpoints
Model card Files Community
4
Edit model card

opus-mt-tc-big-gmw-gmw

Table of Contents

Model Details

Neural machine translation model for translating from West Germanic languages (gmw) to West Germanic languages (gmw).

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: 2022-08-11
  • License: CC-BY-4.0
  • Language(s):
    • Source Language(s): afr deu eng enm fry gos gsw hrx ksh ltz multi nds nld pdc sco stq swg tpi yid
    • Target Language(s): afr ang deu eng enm fry gos ltz multi nds nld sco tpi yid
    • Language Pair(s): afr-deu afr-eng afr-nld deu-afr deu-deu deu-eng deu-nds deu-nld eng-afr eng-deu eng-eng eng-nld fry-eng fry-nld gos-deu gos-eng gos-nld hrx-deu hrx-eng ltz-deu ltz-eng ltz-nld multi-multi nds-deu nds-eng nds-nld nld-afr nld-deu nld-eng nld-fry nld-nds nld-nld
    • Valid Target Language Labels: >>act<< >>afr<< >>afs<< >>aig<< >>ang<< >>ang_Latn<< >>bah<< >>bar<< >>bis<< >>bjs<< >>brc<< >>bzj<< >>bzj_Latn<< >>bzk<< >>cim<< >>dcr<< >>deu<< >>djk<< >>djk_Latn<< >>drt<< >>drt_Latn<< >>dum<< >>eng<< >>enm<< >>enm_Latn<< >>fpe<< >>frk<< >>frr<< >>fry<< >>gcl<< >>gct<< >>geh<< >>gmh<< >>gml<< >>goh<< >>gos<< >>gpe<< >>gsw<< >>gul<< >>gyn<< >>hrx<< >>hrx_Latn<< >>hwc<< >>icr<< >>jam<< >>jvd<< >>kri<< >>ksh<< >>kww<< >>lim<< >>lng<< >>ltz<< >>mhn<< >>nds<< >>nld<< >>odt<< >>ofs<< >>ofs_Latn<< >>oor<< >>osx<< >>pcm<< >>pdc<< >>pdt<< >>pey<< >>pfl<< >>pih<< >>pih_Latn<< >>pis<< >>pis_Latn<< >>qlm<< >>rop<< >>sco<< >>sdz<< >>skw<< >>sli<< >>srm<< >>srm_Latn<< >>srn<< >>stl<< >>stq<< >>svc<< >>swg<< >>sxu<< >>tch<< >>tcs<< >>tgh<< >>tpi<< >>trf<< >>twd<< >>uln<< >>vel<< >>vic<< >>vls<< >>vmf<< >>wae<< >>wep<< >>wes<< >>wes_Latn<< >>wym<< >>ydd<< >>yec<< >>yid<< >>yih<< >>zea<<
  • Original Model: opusTCv20210807_transformer-big_2022-08-11.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. >>afr<<

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:

from transformers import MarianMTModel, MarianTokenizer

src_text = [
    ">>nds<< Red keinen Quatsch.",
    ">>eng<< Findet ihr das nicht etwas übereilt?"
]

model_name = "pytorch-models/opus-mt-tc-big-gmw-gmw"
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) )

# expected output:
#     Kiek ok bi: Rott.
#     Aren't you in a hurry?

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

from transformers import pipeline
pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-gmw-gmw")
print(pipe(">>nds<< Red keinen Quatsch."))

# expected output: Kiek ok bi: Rott.

Training

Evaluation

langpair testset chr-F BLEU #sent #words
afr-deu tatoeba-test-v2020-07-28-v2021-08-07 0.68633 50.3 1583 9105
afr-eng tatoeba-test-v2020-07-28-v2021-08-07 0.70502 56.4 1374 9622
afr-nld tatoeba-test-v2020-07-28-v2021-08-07 0.71500 55.5 1056 6710
deu-afr tatoeba-test-v2020-07-28-v2021-08-07 0.70191 54.2 1583 9507
deu-deu tatoeba-test-v2020-07-28-v2021-08-07 0.57304 34.6 2500 20797
deu-eng tatoeba-test-v2020-07-28-v2021-08-07 0.65919 48.4 17565 149415
deu-nds tatoeba-test-v2020-07-28-v2021-08-07 0.48028 23.2 9999 76119
deu-nld tatoeba-test-v2020-07-28-v2021-08-07 0.71366 54.4 10218 75208
deu-yid tatoeba-test-v2020-07-28-v2021-08-07 9.234 0.4 853 5353
eng-afr tatoeba-test-v2020-07-28-v2021-08-07 0.71940 56.4 1374 10314
eng-deu tatoeba-test-v2020-07-28-v2021-08-07 0.62912 41.8 17565 151539
eng-eng tatoeba-test-v2020-07-28-v2021-08-07 0.80136 66.3 12062 115099
eng-nld tatoeba-test-v2020-07-28-v2021-08-07 0.70929 54.3 12696 91769
eng-yid tatoeba-test-v2020-07-28-v2021-08-07 9.648 0.4 2483 16388
fry-eng tatoeba-test-v2020-07-28-v2021-08-07 0.40304 24.5 220 1573
fry-nld tatoeba-test-v2020-07-28-v2021-08-07 0.54939 40.5 260 1854
gos-deu tatoeba-test-v2020-07-28-v2021-08-07 0.45302 25.4 207 1168
gos-eng tatoeba-test-v2020-07-28-v2021-08-07 0.37587 23.9 1154 5634
gos-nld tatoeba-test-v2020-07-28-v2021-08-07 0.45701 26.1 1852 9902
hrx-deu tatoeba-test-v2020-07-28-v2021-08-07 0.51840 30.0 471 2805
hrx-eng tatoeba-test-v2020-07-28-v2021-08-07 0.42778 29.2 221 1235
ltz-deu tatoeba-test-v2020-07-28-v2021-08-07 0.37005 21.0 347 2208
ltz-eng tatoeba-test-v2020-07-28-v2021-08-07 0.37764 30.1 293 1840
ltz-nld tatoeba-test-v2020-07-28-v2021-08-07 0.32392 26.4 292 1685
multi-multi tatoeba-test-v2020-07-28-v2021-08-07 0.59400 40.4 10000 74505
nds-deu tatoeba-test-v2020-07-28-v2021-08-07 0.63898 45.5 9999 74544
nds-eng tatoeba-test-v2020-07-28-v2021-08-07 0.55112 38.4 2500 17584
nds-nld tatoeba-test-v2020-07-28-v2021-08-07 0.66676 49.8 1657 11489
nld-afr tatoeba-test-v2020-07-28-v2021-08-07 0.76610 62.3 1056 6823
nld-deu tatoeba-test-v2020-07-28-v2021-08-07 0.73047 56.7 10218 74121
nld-eng tatoeba-test-v2020-07-28-v2021-08-07 0.73940 60.2 12696 89970
nld-fry tatoeba-test-v2020-07-28-v2021-08-07 0.47959 31.0 260 1857
nld-nds tatoeba-test-v2020-07-28-v2021-08-07 0.43743 20.0 1657 11711
nld-nld tatoeba-test-v2020-07-28-v2021-08-07 0.63646 44.9 1000 7196
swg-deu tatoeba-test-v2020-07-28-v2021-08-07 0.40319 16.3 1523 15630
yid-deu tatoeba-test-v2020-07-28-v2021-08-07 6.304 0.1 853 5172
yid-eng tatoeba-test-v2020-07-28-v2021-08-07 3.715 0.1 2483 15449
yid-yid tatoeba-test-v2020-07-28-v2021-08-07 6.596 0.1 292 1802
deu-eng newssyscomb2009 0.54992 28.2 502 11821
eng-deu newssyscomb2009 0.53867 23.2 502 11271
deu-eng news-test2008 0.54584 27.2 2051 49380
eng-deu news-test2008 0.53204 23.7 2051 47427
deu-eng newstest2009 0.53749 25.9 2525 65402
eng-deu newstest2009 0.53283 22.9 2525 62816
deu-eng newstest2010 0.58356 30.6 2489 61724
eng-deu newstest2010 0.54886 25.8 2489 61511
deu-eng newstest2011 0.54883 26.3 3003 74681
eng-deu newstest2011 0.52712 23.1 3003 72981
deu-eng newstest2012 0.56160 28.5 3003 72812
eng-deu newstest2012 0.52662 23.3 3003 72886
deu-eng newstest2013 0.57770 31.4 3000 64505
eng-deu newstest2013 0.55774 27.8 3000 63737
deu-eng newstest2014-deen 0.59826 33.2 3003 67337
eng-deu newstest2014-deen 0.59441 29.6 3003 62964
deu-eng newstest2015-ende 0.59660 33.4 2169 46443
eng-deu newstest2015-ende 0.59889 32.3 2169 44260
deu-eng newstest2016-ende 0.64736 39.8 2999 64126
eng-deu newstest2016-ende 0.64429 38.3 2999 62670
deu-eng newstest2017-ende 0.60933 35.2 3004 64399
eng-deu newstest2017-ende 0.59258 30.7 3004 61291
deu-eng newstest2018-ende 0.66796 42.6 2998 67013
eng-deu newstest2018-ende 0.69605 46.5 2998 64276
deu-eng newstest2019-deen 0.63766 39.8 2000 39282
eng-deu newstest2019-ende 0.66880 43.3 1997 48969

Citation Information 

@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 European Language Grid as pilot project 2866, by the FoTran project, funded by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 771113), and the MeMAD project, funded by the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No 780069. We are also grateful for the generous computational resources and IT infrastructure provided by CSC -- IT Center for Science, Finland.

Model conversion info

  • transformers version: 4.16.2
  • OPUS-MT git hash: c1980b5
  • port time: Sun Oct 8 14:39:59 EEST 2023
  • port machine: LM0-400-22516.local
Downloads last month
9
Safetensors
Model size
213M params
Tensor type
FP16
·

Spaces using Helsinki-NLP/opus-mt-tc-big-gmw-gmw 2

Evaluation results

View on Papers With Code