Changing the Default Model for Watson Language Translation

Question

I'm trying to set the default model to a trained custom model with Watson Language Translation, so when I translate for a given language pair, I don't need to specify my model.

I've tried passing default_model as parameter in my curl statement (all parameters correctly filled):

curl --user apikey:{mykey} --request POST --form forced_glossary=@{mytmx} "{myurl}/v3/models?version=2018-09-28&base_model_id=es-en&name={myname}&default_model=true"

This still creates the new model with "default_model": false.

I've read through the documentation for training and the API documentation, but I don't see anything specifically pertaining to this issue.

Can anyone tell me what I am missing here?

Answer 1

If you take a closer look at the API documentation you will notice that default_model is always output, never input. ie. You can't promote a custom model to a default_model. Principally because it won't have all the required translations for a default_model.

So if you do specify it as input it will be ignored. Which means that to use the custom model in a translation you do have to specify it as model_id.