marian-nmt / marian-examples

Examples, tutorials and use cases for Marian, including our WMT-2017/18 baselines.
Other
78 stars 34 forks source link

Double data? #6

Closed duyvuleo closed 6 years ago

duyvuleo commented 6 years ago

Hi Marcin,

May I need you to clarify something?

''' if [ ! -e "data/all.bpe.en" ] then cat data/corpus.bpe.en data/corpus.bpe.en data/news.2016.bpe.en > data/all.bpe.en cat data/corpus.bpe.de data/corpus.bpe.de data/news.2016.bpe.de > data/all.bpe.de fi '''

Why do you need to double the parallel data (corpus.bpe.{en,de})?

Thanks!

duyvuleo commented 6 years ago

I understood this. It is to fit with sample size of synthetic data.