opus1m+bt-tuned4eng2fao-2021-04-16.yml

release: eng-gmq/opus1m+bt-tuned4eng2fao-2021-04-16.zip
release-date: 2021-04-16
dataset-name: opus1m+bt-tuned4eng2fao
modeltype: transformer-align
vocabulary:
   source: opus1m+bt-tuned4eng2fao.spm32k-spm32k.vocab.yml
   target: opus1m+bt-tuned4eng2fao.spm32k-spm32k.vocab.yml
pre-processing: normalization + SentencePiece (spm32k,spm32k)
subwords:
   source: spm32k
   target: spm32k
subword-models:
   source: source.spm
   target: target.spm
source-languages:
   - eng
target-languages:
   - fao
use-target-labels:
   - ">>fao<<"
training-data:
   eng-fao: Tatoeba-train (2613) wiki.aa.fao-eng (103717) wikisource.aa.fao-eng (1298) 
validation-data:
   eng-fao: Tatoeba-dev, 1000
   total-size-shuffled: 708
   devset-selected: top 708  lines of Tatoeba-dev.eng-fao.src.shuffled
test-data:
   Tatoeba-test.eng-fao: 294/1933
   Tatoeba-test.eng-multi: 10000/71671
BLEU-scores:
   Tatoeba-test.eng-fao: 18.7
   Tatoeba-test.eng-multi: 44.1
chr-F-scores:
   Tatoeba-test.eng-fao: 0.406
   Tatoeba-test.eng-multi: 0.607