We're looking at setting up on-demand cloud GPU training, but have hit a blocker: we'd like to use On-Demand EC2 instances (as described here None ), but it seems like the only options supported at the moment by the AWS Autoscaler are Spot and Reserved instance types?
Am I missing something? Is there a roadmap to support On-Demand instances? Is there a workaround?
Hi @<1752139558343938048:profile|ScatteredLizard17> , the two of supported types of instances by the ClearML autoscaler are on demand/spot instances, nothing to do with reserved ones
Okay, maybe the documentation is wrong or just very confusing - this page says "Use Spot Instance - Select to use a spot instance. Otherwise, a reserved instance is used." None
