multi30k dataset only contains test set for French and no Czech data

[bentrevett]

The multi30k dataset provided by torchtext.datasets contains the training/testing/validation sets for English and German, however for French it only contains a test set - even though training and validation sets are available.

It also does not contain any of the Czech data, or any of the 2017 test data (which is not available in Czech).

The official multi30k repository - https://github.com/multi30k/dataset - contains all of the data at https://github.com/multi30k/dataset/tree/master/data/task1/raw, although getting the data from GitHub is slightly more tricky than directly downloading it from the urls torchtext is already using for multi30k.

I'm guessing that adding this can wait until #751 is finished?

[zhangguanheng66]

@bentrevett Thanks. We @akurniawan should check this in the new translation dataset.