ML-Assisted Labeling – Inference Job Not Triggering After Successful Training Runs

Question

Issue Description:

I am experiencing an issue with ML-assisted data labeling in Azure ML Studio where the Inference job is not being triggered after multiple successful Training runs.

Steps Taken:

1. Created an ML-assisted labeling project for object detection.
2. Manually labeled the required number of samples to start training ( 300 samples).
3. Completed multiple Training runs successfully – no errors.
4. Expected the Inference run to start automatically, but it never begins.
5. Checked that the compute resources are available and correctly assigned for inference.
6. Manually labeled additional samples to trigger new Training runs, but Inference still doesn’t start.

Current Status:

• Training experiments are completing successfully.

• Inference experiment status remains “Not Started.”

• Pre-labeled tasks count stays at 0.

Question:

• How do I manually trigger the Inference run if it does not start automatically?

• Is there a setting or action required to force the system to begin inference?

• What troubleshooting steps can I take to resolve this issue?

I have almost 3,000 documents to label, and manual labeling defeats the purpose of ML-assisted labeling. Any guidance on how to get the inference process working would be greatly appreciated!