You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When I use the dataset wmt14en-de ,I got the bleu score:24.5,which is just like the paper's score,
but when I use the same way to train the model with Wmt17 zh-en,the bleu score is only 7.0.
When I use the dataset wmt14en-de ,I got the bleu score:24.5,which is just like the paper's score,
but when I use the same way to train the model with Wmt17 zh-en,the bleu score is only 7.0.
the dataset Wmt17 zh-en:
http://data.statmt.org/wmt17/translation-task/training-parallel-nc-v12.tgz",
["training/news-commentary-v12.zh-en.en",
"training/news-commentary-v12.zh-en.zh"]]]
why
how can I do ?
The text was updated successfully, but these errors were encountered: