Incorrect distribution in the dataset

#2
by Evgeniy - opened

In wmt19-ru-en, the same data is used in the test and validation. why didn't you write about it anywhere?

this problem also occurs in the Chinese (wmt19-zh-en) dataset, possibly in others

Thanks for your reporting. But we directly use the dataset from hugging face, maybe you can raise an issue on their github.

StevenTang changed discussion status to closed

Sign up or log in to comment