Helsinki-NLP/OPUS-MT-train
1
1
Fork 0
mirror of https://github.com/Helsinki-NLP/OPUS-MT-train.git synced 2024-08-17 16:20:50 +03:00
Code Issues Packages Projects Releases Wiki Activity

latest models

Browse Source
This commit is contained in:
Joerg Tiedemann 2020-02-26 21:20:17 +02:00
parent 44182291dc
commit 2805bf49e7
18 changed files with 364 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

62
Makefile.dist
View File

@ -55,7 +55,10 @@ best_dist_all:
## (BLEU needs to be above MIN_BLEU_SCORE)
## NEW: don't trust models tested with GNOME test sets!
best_dist:
## OLD version of finding the best model
## --> this didn't properly look at different variants in the same folder
best_dist_old:
@m=0;\
s=''; \
echo "------------------------------------------------"; \
@ -85,6 +88,54 @@ best_dist:
fi
## new version of finding the best model
## --> look at different model variants in each work-dir
## --> take only the best one to publish
best_dist:
@m=0;\
s=''; \
echo "------------------------------------------------"; \
echo "search best model for ${LANGPAIRSTR}"; \
for d in ${ALT_MODEL_DIR}; do \
e=`ls work-$$d/${LANGPAIRSTR}/val/*.trg | xargs basename | sed 's/\.trg//'`; \
echo "evaldata = $$e"; \
if [ "$$e" != "GNOME" ]; then \
I=`find work-$$d/${LANGPAIRSTR}/ -maxdepth 1 -name "$$e.*.eval" -printf "%f\n"`; \
for i in $$I; do \
x=`echo $$i | cut -f3 -d. | cut -f1 -d-`; \
y=`echo $$i | cut -f3 -d. | cut -f2 -d- | sed 's/[0-9]$$//'`; \
z=`echo $$i | cut -f2 -d.`; \
v=`echo $$i | cut -f4 -d.`; \
b=`grep 'BLEU+' work-$$d/${LANGPAIRSTR}/$$e.$$z.$$x-$$y[0-9].$$v.*.eval | cut -f3 -d' '`; \
if (( $$(echo "$$m-$$b < 0" |bc -l) )); then \
echo "$$d/$$i ($$b) is better than $$s ($$m)!"; \
m=$$b; \
E=$$i; \
s=$$d; \
else \
echo "$$d/$$i ($$b) is worse than $$s ($$m)!"; \
fi \
done; \
fi \
done; \
echo "------------------------------------------------"; \
if [ "$$s" != "" ]; then \
if (( $$(echo "$$m > ${MIN_BLEU_SCORE}" |bc -l) )); then \
x=`echo $$E | cut -f3 -d. | cut -f1 -d-`; \
y=`echo $$E | cut -f3 -d. | cut -f2 -d- | sed 's/[0-9]$$//'`; \
z=`echo $$E | cut -f2 -d.`; \
v=`echo $$E | cut -f4 -d.`; \
${MAKE} \
MODELSHOME=${PWD}/models \
PRE_SRC=$$x PRE_TRG=$$y \
DATASET=$$z \
MODELTYPE=$$v \
MODELS_URL=https://object.pouta.csc.fi/OPUS-MT-models dist-$$s; \
fi; \
fi
## make a package for distribution
@ -119,6 +170,9 @@ endif
POSTPROCESS_SCRIPT = postprocess-${PREPROCESS_TYPE}.sh
## make the distribution package including test evaluation files and README
${DIST_PACKAGE}: ${MODEL_FINAL}
ifneq (${SKIP_DIST_EVAL},1)
@${MAKE} $(TEST_EVALUATION)
@ -147,10 +201,10 @@ endif
echo '## Benchmarks' >> ${WORKDIR}/README.md; \
echo '' >> ${WORKDIR}/README.md; \
cd ${WORKDIR}; \
grep -H BLEU *k${NR}.*eval | \
grep -H BLEU *.${DATASET}.${PRE_SRC}-${PRE_TRG}${NR}.${MODELTYPE}.*.eval | \
tr '.' '/' | cut -f1,5,6 -d '/' | tr '/' "." > $@.1; \
grep BLEU *k${NR}.*eval | cut -f3 -d ' ' > $@.2; \
grep chrF *k${NR}.*eval | cut -f3 -d ' ' > $@.3; \
grep BLEU *.${DATASET}.${PRE_SRC}-${PRE_TRG}${NR}.${MODELTYPE}.*.eval | cut -f3 -d ' ' > $@.2; \
grep chrF *.${DATASET}.${PRE_SRC}-${PRE_TRG}${NR}.${MODELTYPE}.*.eval | cut -f3 -d ' ' > $@.3; \
echo '| testset | BLEU | chr-F |' >> README.md; \
echo '|-----------------------|-------|-------|' >> README.md; \
paste $@.1 $@.2 $@.3 | sed "s/\t/ | /g;s/^/| /;s/$$/ |/" >> README.md; \

1
Makefile.env
View File

@ -20,6 +20,7 @@ NR_GPUS = 1
HPC_NODES = 1
HPC_DISK = 500
HPC_QUEUE = serial
HPC_GPUQUEUE = gpu
# HPC_MODULES = nlpl-opus python-env/3.4.1 efmaral moses
# HPC_MODULES = nlpl-opus moses cuda-env marian python-3.5.3-ml
HPC_MODULES = ${GPU_MODULES}

52
Makefile.generic
View File

@ -20,12 +20,56 @@
## - make dist-allmodels
%-allmodels:
for l in ${ALL_LANG_PAIRS}; do \
if [ `find ${WORKHOME}/$$l -name '*.${PRE_SRC}-${PRE_TRG}.*.best-perplexity.npz' | wc -l` -gt 0 ]; then \
${MAKE} SRCLANGS="`echo $$l | cut -f1 -d'-' | sed 's/\\+/ /g'`" \
TRGLANGS="`echo $$l | cut -f2 -d'-' | sed 's/\\+/ /g'`" ${@:-allmodels=}; \
fi \
m=`find ${WORKHOME}/$$l -maxdepth 1 -name '*.best-perplexity.npz' -printf "%f\n"`; \
for i in $$m; do \
s=`echo $$l | cut -f1 -d'-' | sed 's/\\+/ /g'`; \
t=`echo $$l | cut -f2 -d'-' | sed 's/\\+/ /g'`; \
d=`echo $$i | cut -f1 -d.`; \
x=`echo $$i | cut -f2 -d. | cut -f1 -d-`; \
y=`echo $$i | cut -f2 -d. | cut -f2 -d-`; \
v=`echo $$i | cut -f3 -d.`; \
echo "model = $$i"; \
echo "dataset = $$d"; \
echo "src-lang = $$s"; \
echo "trg-lang = $$t"; \
echo "pre-src = $$x"; \
echo "pre-trg = $$y"; \
echo "type = $$v"; \
${MAKE} \
SRCLANGS="$$s" TRGLANGS="$$t" \
DATASET=$$d \
PRE_SRC=$$x PRE_TRG=$$y \
MODELTYPE=$$v ${@:-allmodels=}; \
done \
done
## OLD: doesn't work for different model variants
##
# %-allmodels:
# for l in ${ALL_LANG_PAIRS}; do \
# if [ `find ${WORKHOME}/$$l -name '*.${PRE_SRC}-${PRE_TRG}.*.best-perplexity.npz' | wc -l` -gt 0 ]; then \
# ${MAKE} SRCLANGS="`echo $$l | cut -f1 -d'-' | sed 's/\\+/ /g'`" \
# TRGLANGS="`echo $$l | cut -f2 -d'-' | sed 's/\\+/ /g'`" ${@:-allmodels=}; \
# fi \
# done
listallmodels:
@m=`find ${WORKDIR} -maxdepth 1 -name '*.best-perplexity.npz' -printf "%f\n"`; \
for i in $$m; do \
d=`echo $$i | cut -f1 -d.`; \
s=`echo $$i | cut -f2 -d. | cut -f1 -d-`; \
t=`echo $$i | cut -f2 -d. | cut -f1 -d-`; \
v=`echo $$i | cut -f3 -d.`; \
echo "model = $$i"; \
echo "dataset = $$d"; \
echo "pre-src = $$s"; \
echo "pre-trg = $$t"; \
echo "type = $$v"; \
done
## only bilingual models
%-allbilingual:
for l in ${ALL_BILINGUAL_MODELS}; do \

2
Makefile.slurm
View File

@ -31,7 +31,7 @@ ifdef EMAIL
endif
echo '#SBATCH -n 1' >> $@
echo '#SBATCH -N 1' >> $@
echo '#SBATCH -p gpu' >> $@
echo '#SBATCH -p ${HPC_GPUQUEUE}' >> $@
ifeq (${shell hostname --domain},bullx)
echo '#SBATCH --account=${CSCPROJECT}' >> $@
echo '#SBATCH --gres=gpu:${GPU}:${NR_GPUS},nvme:${HPC_DISK}' >> $@

2
finetune/Makefile
View File

@ -55,7 +55,7 @@ MODELINDEX = ${OBJECTSTORAGE}/${MODELCONTAINER}/index.txt
# BASEMODELNAME = ${patsubst %.zip,%,${notdir ${BASEMODELZIP}}}
BASEMODELHOME = ${OBJECTSTORAGE}/${MODELCONTAINER}/${LANGPAIR}
BASEMODELZIP = ${lastword ${sort ${notdir ${shell grep '${LANGPAIR}/opus-.*\.zip' model-index.txt}}}}
BASEMODELZIP = ${lastword ${sort ${notdir ${shell grep '${LANGPAIR}/${DATASET}-.*\.zip' model-index.txt}}}}
BASEMODELNAME = ${BASEMODELZIP:.zip=}
TUNED_MODEL = ${LANGPAIR}/${MODEL}/model/${BASEMODELNAME}_${MODEL}.transformer.model

28
models/de-en/README.md
View File

@ -54,3 +54,31 @@
| newstest2019-deen.de.en | 39.6 | 0.637 |
| Tatoeba.de.en | 55.1 | 0.704 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/de-en/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/de-en/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/de-en/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 |

15
models/en-bcl/README.md
View File

@ -43,3 +43,18 @@
|-----------------------|-------|-------|
| JW300.en.bcl | 53.8 | 0.719 |
# opus+bt-2020-02-26.zip
* dataset: opus+bt
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus+bt-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/en-bcl/opus+bt-2020-02-26.zip)
* test set translations: [opus+bt-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-bcl/opus+bt-2020-02-26.test.txt)
* test set scores: [opus+bt-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-bcl/opus+bt-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| JW300.en.bcl | 54.3 | 0.722 |

27
models/en-de/README.md
View File

@ -52,3 +52,30 @@
| newstest2019-ende.en.de | 41.6 | 0.655 |
| Tatoeba.en.de | 46.8 | 0.655 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/en-de/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-de/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-de/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| newssyscomb2009.en.de | 23.5 | 0.540 |
| news-test2008.en.de | 23.5 | 0.529 |
| newstest2009.en.de | 22.3 | 0.530 |
| newstest2010.en.de | 24.9 | 0.544 |
| newstest2011.en.de | 22.5 | 0.524 |
| newstest2012.en.de | 23.0 | 0.525 |
| newstest2013.en.de | 26.9 | 0.553 |
| newstest2015-ende.en.de | 31.1 | 0.594 |
| newstest2016-ende.en.de | 37.0 | 0.636 |
| newstest2017-ende.en.de | 29.9 | 0.586 |
| newstest2018-ende.en.de | 45.2 | 0.690 |
| newstest2019-ende.en.de | 40.9 | 0.654 |
| Tatoeba.en.de | 47.3 | 0.664 |

23
models/en-fi/README.md
View File

@ -67,3 +67,26 @@
| newstestB2017-enfi.en.fi | 22.3 | 0.560 |
| Tatoeba.en.fi | 41.1 | 0.645 |
# opus+bt-2020-02-26.zip
* dataset: opus+bt
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus+bt-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/en-fi/opus+bt-2020-02-26.zip)
* test set translations: [opus+bt-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-fi/opus+bt-2020-02-26.test.txt)
* test set scores: [opus+bt-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-fi/opus+bt-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| newsdev2015-enfi.en.fi | 23.5 | 0.572 |
| newstest2015-enfi.en.fi | 25.7 | 0.586 |
| newstest2016-enfi.en.fi | 26.4 | 0.594 |
| newstest2017-enfi.en.fi | 29.8 | 0.624 |
| newstest2018-enfi.en.fi | 19.8 | 0.545 |
| newstest2019-enfi.en.fi | 23.7 | 0.562 |
| newstestB2016-enfi.en.fi | 21.3 | 0.558 |
| newstestB2017-enfi.en.fi | 24.9 | 0.584 |
| Tatoeba.en.fi | 41.4 | 0.650 |

24
models/en-fr/README.md
View File

@ -22,3 +22,27 @@
| newstest2013.en.fr | 24.0 | 0.525 |
| Tatoeba.en.fr | 56.7 | 0.682 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/en-fr/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-fr/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-fr/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| newsdiscussdev2015-enfr.en.fr | 33.8 | 0.602 |
| newsdiscusstest2015-enfr.en.fr | 40.0 | 0.643 |
| newssyscomb2009.en.fr | 29.8 | 0.584 |
| news-test2008.en.fr | 27.5 | 0.554 |
| newstest2009.en.fr | 29.4 | 0.577 |
| newstest2010.en.fr | 32.7 | 0.596 |
| newstest2011.en.fr | 34.3 | 0.611 |
| newstest2012.en.fr | 31.8 | 0.592 |
| newstest2013.en.fr | 33.2 | 0.589 |
| Tatoeba.en.fr | 50.5 | 0.672 |

15
models/en-sv/README.md
View File

@ -13,3 +13,18 @@
|-----------------------|-------|-------|
| Tatoeba.en.sv | 60.4 | 0.733 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/en-sv/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-sv/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-sv/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba.en.sv | 60.1 | 0.736 |

15
models/en-tl/README.md Normal file
View File

@ -0,0 +1,15 @@
# opus+bt-2020-02-26.zip
* dataset: opus+bt
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus+bt-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/en-tl/opus+bt-2020-02-26.zip)
* test set translations: [opus+bt-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-tl/opus+bt-2020-02-26.test.txt)
* test set scores: [opus+bt-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-tl/opus+bt-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba.en.tl | 26.6 | 0.577 |

15
models/fi-fr/README.md
View File

@ -28,3 +28,18 @@
|-----------------------|-------|-------|
| Tatoeba.fi.fr | 49.9 | 0.657 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/fi-fr/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/fi-fr/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/fi-fr/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba.fi.fr | 50.7 | 0.670 |

15
models/fi-nl/README.md
View File

@ -28,3 +28,18 @@
|-----------------------|-------|-------|
| JW300.fi.nl | 30.2 | 0.544 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/fi-nl/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/fi-nl/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/fi-nl/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| JW300.fi.nl | 30.5 | 0.557 |

25
models/fr-en/README.md
View File

@ -23,3 +23,28 @@
| newstest2014-fren.fr.en | 35.4 | 0.607 |
| Tatoeba.fr.en | 60.9 | 0.726 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/fr-en/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/fr-en/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/fr-en/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| newsdiscussdev2015-enfr.fr.en | 33.1 | 0.580 |
| newsdiscusstest2015-enfr.fr.en | 38.7 | 0.614 |
| newssyscomb2009.fr.en | 30.3 | 0.569 |
| news-test2008.fr.en | 26.2 | 0.542 |
| newstest2009.fr.en | 30.2 | 0.570 |
| newstest2010.fr.en | 32.2 | 0.590 |
| newstest2011.fr.en | 33.0 | 0.597 |
| newstest2012.fr.en | 32.8 | 0.591 |
| newstest2013.fr.en | 33.9 | 0.591 |
| newstest2014-fren.fr.en | 37.8 | 0.633 |
| Tatoeba.fr.en | 57.5 | 0.720 |

15
models/nl-fi/README.md
View File

@ -13,3 +13,18 @@
|-----------------------|-------|-------|
| JW300.nl.fi | 28.1 | 0.553 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/nl-fi/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/nl-fi/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/nl-fi/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| JW300.nl.fi | 28.6 | 0.569 |

23
models/ru-en/README.md
View File

@ -90,3 +90,26 @@
| newstest2019-ruen.ru.en | 31.5 | 0.577 |
| Tatoeba.ru.en | 60.8 | 0.734 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/ru-en/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/ru-en/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/ru-en/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| newstest2012.ru.en | 34.8 | 0.603 |
| newstest2013.ru.en | 27.9 | 0.545 |
| newstest2014-ruen.ru.en | 31.9 | 0.591 |
| newstest2015-enru.ru.en | 30.4 | 0.568 |
| newstest2016-enru.ru.en | 30.1 | 0.565 |
| newstest2017-enru.ru.en | 33.4 | 0.593 |
| newstest2018-enru.ru.en | 29.6 | 0.565 |
| newstest2019-ruen.ru.en | 31.4 | 0.576 |
| Tatoeba.ru.en | 61.1 | 0.736 |

15
models/sv-en/README.md
View File

@ -43,3 +43,18 @@
|-----------------------|-------|-------|
| Tatoeba.sv.en | 65.3 | 0.769 |
# opus-2020-02-26.zip
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/sv-en/opus-2020-02-26.zip)
* test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/sv-en/opus-2020-02-26.test.txt)
* test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/sv-en/opus-2020-02-26.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba.sv.en | 64.5 | 0.763 |