OS translate-train translate-test datasets?

[Timoeller]

Hey Patrick,
awesome work you are doing there on getting cool ML also to other languages.
Is there a way you guys open source the automatically translated datasets for translate-train and/or translate-test as well?

[patrick-s-h-lewis]

Hi Timo,

I'm looking into this,

Patrick

[josecannete]

Hi Patrick @patrick-s-h-lewis, any news about this?

Thank you so much!

[patrick-s-h-lewis]

I think we're very close to getting machine-translations open-sourced, hopefully next week.

[patrick-s-h-lewis]

Translate-train and Translate-test now available:

https://github.com/facebookresearch/MLQA#translate-train-and-translate-test-data

[F95GIT]

Hey, i tried to use the translated german version "de_squad-translate-train-train-v1.1.json" but when I am trying to load it it always raises an Error in the read_squad_examples script:
`---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
in
2 examples2 = read_squad_examples(input_file=input_file2,
3 is_training=True,
----> 4 version_2_with_negative=True
5 )

~/Downloads/huggingface-transformers-971c9b0/examples/utils_squad.py in read_squad_examples(input_file, is_training, version_2_with_negative)
145 start_position = None
146 end_position = None
--> 147 orig_answer_text = None
148 is_impossible = False
149 if is_training:

KeyError: 'is_impossible'`

Do you maybe have a solution for my problem ?

Thank you