Open AI Finetunning stuck in status "Running"

Hi Patrick Ragawidya Kristianto

I understand that your fine-tuning job is stuck in "Running" status since hours, there could be several reasons for this issue. Here are some potential causes and steps you can take:

Reviewing the job status in the Fine-tuning section of the Azure AI Studio portal. Fine-tuning jobs are sometimes queued due to high demand or limited resources. Try refreshing the portal to check for any recent updates.

If the job continues to hang with no progress, consider canceling and resubmitting it. Since you've already canceled a previous long-running job, resubmitting with validated data and parameters might help rule out edge issues.

Azure’s compute resources for fine-tuning are shared across tenants and may not be immediately available. Check the Azure Status Page to rule out regional delays or service incidents that might be impacting availability.

Also you can refer Check the status of your custom model,

Troubleshooting for Azure OpenAI fine-tuning.

I Hope this helps. Do let me know if you have any further queries.

Thank you!

Hi Patrick Ragawidya Kristianto

Did you get any chance to check the response.

Thank you!

For now the process is completed successfully. However, sometimes it remains in the state shown in the log for up to 24 hours. What could be causing this delay because there is nothing on the log indicated the process get problems?

Hi Patrick Ragawidya Kristianto

The delay in fine-tuning completion where the process remains in a "running" state for up to 24 hours is expected behavior in Azure OpenAI and does not indicate a problem. This occurs due to shared compute resource management, where final status updates are queued across multiple tenants. Background processes such as resource cleanup, final model validation, and system synchronization contribute to this delay.
While waiting, users can check the Azure Portal for updates, verify resource availability in their region, and review logs for unexpected errors. This delay does not impact the quality of the fine-tuned model, and no action is required during this period. Let me know if you need further clarification!

Thanks