Sagemaker Ground Truth Automated Data Labeling Confusion

I had created an API for automated data labeling workflow in AWS sagemaker for the bounding box (object detection) task, in which set the maxconcurrenttask at a time to 220 images. I had roughly about 2000 images. Once I finished hand label those 220 datas, I perceived the task disappear for the worker, hence I assume it then perform the active learning part. But in the output location that I specify in s3, only the active_learning_info.json in the activelearning\ folder was present. I then checked the immediate file that have the output.manifest, in which in the human-annotated was all yes. Thus I'm kinda confuse on if the active learning part is actually deploy or not since the job is being sent back to the worker again and the labeled data is not increase after the first iteration/batch? Furthemore, I observe that a specific EC2 instance is needed for the process, so do I have to create that instance my own for the active learning to process or it will automatically uses that instance and sent the pricing later?

Thank you