README.md

metadata

language:
  - de
  - en
pipeline_tag: translation

mt5-large-nc16-250k-deen

This model is released as part of the work from Are Character-level Translations Worth the Wait? Comparing Character- and Subword-level Models for Machine Translation. It is an mT5 model finetuned on German-->English translation using 250k sentence pairs from the WMT NewsCommentary v16 dataset.

To use the model correctly, you must prepend the prompt with "translate X to Y: ", where X and Y are your source and target languages (e.g. German, English).

NOTE: The decoder_start_token_id is 259 for byt5 models and 250099 for mt5 models, which is different from the default token from google's byt5 and mt5 models (which is 0).