RainerGa's picture
Update README.md
787fede
metadata
license: apache-2.0
language:
  - de

This is only a Copy of the Work of OpenAssistant and the user timdettmers

Target of this trainingdata is finetuning only in German language.

File openassistant_origfile_with_lang_informations.txt is the full trainingdata. Every line starts with Language Informations. You can easily filter with:

cat openassistant_origfile_with_lang_informations.txt | grep ^de | sed s/^de,//g > openassistant_best_replies_de_train.jsonl

Replace ^de with the language you are interessted in.

For language detection I use python "langdetect". The simple python Script "detect_language.py" is part of this dataset but NOT needed for finetuning a model!

Languages

OpenAssistant Conversations only German language:

Languages with over 1000 messages

  • German: 5279