The GET Status request returns one of four values in the status field of its success response body. From my understanding, running means training is still happening, failed suggests an error prevented training from producing a usable model, and complete means the training finished (either automatically or due to an interrupt) and has produced a model that can be used with GET Prediction.
I cannot find a use for cancelled as a response, as DELETE will remove the model entirely, so GET Status will return 404 job_not_found, and DELETE Stop Training will produce a usable model that may be (depending on implementation and timing) identical to an uninterrupted model, so should return complete.
Can we either clearly define the use of cancelled or remove it from the protocol?
The GET Status request returns one of four values in the status field of its success response body. From my understanding, running means training is still happening, failed suggests an error prevented training from producing a usable model, and complete means the training finished (either automatically or due to an interrupt) and has produced a model that can be used with GET Prediction.
I cannot find a use for cancelled as a response, as DELETE will remove the model entirely, so GET Status will return 404 job_not_found, and DELETE Stop Training will produce a usable model that may be (depending on implementation and timing) identical to an uninterrupted model, so should return complete.
Can we either clearly define the use of cancelled or remove it from the protocol?