Resume a fine-tuning job
Available only to organizations enabled for fine-tuning; any other organization receives 400 invalid_request on every fine-tuning operation. Starts the next attempt of a failed or cancelled job from its latest checkpoint’s training state, optimizer included, or from the beginning when it saved none. The job is running again and its checkpoints stay as they are; its metric points and rollouts from that step on are deleted for the new attempt to write again. A run that crashes or stops resumes by itself; this is for a job that settled. A queued or running job is returned unchanged; a succeeded job answers 409 conflict, and one whose last run is still being stopped 409 fine_tuning_job_stopping, to retry shortly.
Authorizations
Headers
Response
OK
queued until a worker first starts it; a restarted or resumed job is running with progress.phase queued while it waits for a worker. succeeded is final; failed and cancelled are settled, and managedAgentsResumeFineTuningJob runs them again from their latest checkpoint, or from the beginning when they saved none.
queued, running, succeeded, failed, cancelled