this is a int8 ctranslate2 version of Helsinki-NLP/opus-mt-de-en
opus-mt-de-en
source languages: de
target languages: en
OPUS readme: de-en
dataset: opus
model: transformer-align
pre-processing: normalization + SentencePiece
download original weights: opus-2020-02-26.zip
test set translations: opus-2020-02-26.test.txt
test set scores: opus-2020-02-26.eval.txt
Benchmarks
| testset | BLEU | chr-F |
|---|---|---|
| newssyscomb2009.de.en | 29.4 | 0.557 |
| news-test2008.de.en | 27.8 | 0.548 |
| newstest2009.de.en | 26.8 | 0.543 |
| newstest2010.de.en | 30.2 | 0.584 |
| newstest2011.de.en | 27.4 | 0.556 |
| newstest2012.de.en | 29.1 | 0.569 |
| newstest2013.de.en | 32.1 | 0.583 |
| newstest2014-deen.de.en | 34.0 | 0.600 |
| newstest2015-ende.de.en | 34.2 | 0.599 |
| newstest2016-ende.de.en | 40.4 | 0.649 |
| newstest2017-ende.de.en | 35.7 | 0.610 |
| newstest2018-ende.de.en | 43.7 | 0.667 |
| newstest2019-deen.de.en | 40.1 | 0.642 |
| Tatoeba.de.en | 55.4 | 0.707 |
Provenance and EU AI Act Art. 53 note
- Upstream model: Helsinki-NLP/opus-mt-de-en β published by
Helsinki-NLP. - Upstream licence:
apache-2.0. This repository redistributes under the same terms; it grants no rights the upstream licence does not. - What was done here: format conversion and/or quantisation only (CTranslate2, INT8 precision). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
- Training data: documented β where it is documented at all β by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository. No training-content summary was found on the upstream model card at the time of writing; that documentation gap is upstream's and is not filled here.
- Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.
- Downloads last month
- 106