run the following from repo root
az login --tenant <tenant>
az account set -s <subscription>
Create a .env
file at root level of the repo (gitignored) as per the sample provided.
Then run the setup script.
bash infra/setup.sh
Verify the resources created in the Azure portal:
- Resource group
- Azure ML workspace and associated resources
- VM and associated resources
update the vm NSG to allow network access from Azure ML workspace
(here, we're using the portal and configuring an allow rule for the Azure ML service tag, i.e. any ML workspace would be allowed to connect to this VM)
source .env
az ml compute attach -f definitions/compute/attached-vm.yml --resource-id $VM_ID --admin-password $VM_PASSWORD
This should result in a new compute visible through CLI:
az ml compute list -o table
-->
Name Compute type State
---------- -------------- ---------
vm-example virtualmachine Succeeded
or portal:
az ml compute create -f definitions/compute/aml-compute-instance.yml
az ml compute create -f definitions/compute/aml-compute-cluster.yml
az ml job create -f definitions/job/simplejob/job.yml
โ Currently seeing error without useful explanation: "UserError: Unable to establish SSH connection". Allowing access on all ports from anywhere on the NSG does not fix it. Perhaps try using ssh key instead of password for the VM.
๐ Expected outcome (obtained running on AML CI):