There's a very simple API available. When you know the name of the model you want to use
and/or you know the source and target languages, you just
input_text to the model/language endpoint.
See the interactive api doc, where you can explore the
endpoints and get example curl commands. When inspecting an endpoint you have to click the
Try it out button and execute to see the curl commands and get the response (this
lets you play with the parameters if the endpoint has any).
You can also browse through the actual data structures of the api using hal format viewers such as the hal browser
There is a limitation on the size of the payload (file/request); this is currently 102400B
- Martin Popel
- CUBBITT models, en<->cs, en<->fr, en<->pl as desribed in
Popel, M., Tomkova, M., Tomek, J. et al. Transforming machine translation: a deep learning system reaches news translation quality comparable to human professionals. Nat Commun 11, 4381 (2020). https://doi.org/10.1038/s41467-020-18073-9
As of Jan 19 2022, the en<->cs translation uses the upgraded CUBBITT document-level models. The advanced mode (and the API) still offers the old sentence-level models.
- Shantipriya Parida
- en->hi model
- Dušan Variš
- en<->ru, en<->de models