Hi,
Anyone else seeing fine-tuning jobs never leave the queue?
Created 4 DPO jobs last night (method=dpo, starting from existing ft:gpt-4.1-mini-2025-04-14 models). All 4 validated their files fine, last event on each is “Files validated, moving job to queued state” and 12 hours later none has moved to running.
What is odd is that the exact same setup worked three weeks ago and completed normally.
I’m going to guess you actually used your ft models that you recently had success with, not being caught with this current status of the endpoint:
Creating fine-tuning jobs is no longer available to organizations that have not run inference on a fine-tuned model in the past 60 days.
It used to be days of waiting for fine tunes in 2023. It is not unexpected that OpenAI could put fine-tuning training at even lower queue priority than batch inference, since the jobs don’t have any time commitment. I don’t have any DPO that wouldn’t be merely a quick test to clutter the models endpoint, where you already have done a workload-sized test.
Hello, thanks for the answer ! Indeed, I’m not concerned about the current status of the endpoint. The jobs eventually started after around 15 hours. I was just surprised as i never had to wait before. Thanks again.