Hi guys, since I am done with implementing the AWS autoscaler, I would like to share some pain points that I encountered in the process with the hope that they can be documented to help other users:
Nvidia Deep Learning VM AMI is outed in default value for parameter ami_id of the aws autoscheduler task -> Had to update it to ami-04c0416d6bd8e4b1f Parameter extra_configurations is not documented. One should respected "KeyName":"Val" (sensitive to ", should not be ') parameter ebs_volume_type could be updated to "gp3" Parameter ebs_device_name MUST be changed to /dev/sda1 . Default value: /dev/xdva doesn't work