AzureML v2 CLI Workshop

Setup before the Workshop

1. Have an AzureML workspace (or create one at: https://ml.azure.com/)
2. Install VSCode (get it here: https://code.visualstudio.com/download)
3. Install the AzureML VSCode Extension (get it here: https://marketplace.visualstudio.com/items?itemName=ms-toolsai.vscode-ai)
4. Create a new Compute Instance (to make sure you have the latest changes already on the machine).
   Instructions are here -- no need to change any settings; leave everything to defaults.
5. Once the compute instance has been created, click on the VSCode link
   
   That should take you to VSCode and log you in to your Compute Instance.
6. Start a terminal with Terminal / New Terminal and type the following:
   Make sure conda is working

   conda init bash
   

   
   This will make the CLI output prettier

   az config set core.output=yamlc
   

   
   This will log your CI into Azure the identity you used to log into the CI

   az login --identity
   

   
   Test that you are properly logged in:

   az ml compute list -o table
   

   
   This should give you a list like this:

   Name       Compute type     State    Instance type
   ---------  ---------------  -------  ----------------
   danielsc4  computeinstance  Running  STANDARD_DS12_V2
   

7. Clone this repo:

   git clone https://github.com/danielsc/v2workshop.git
   

8. Open the v2workshop folder in VSCode
   
   
   This will reload the folder in VSCode
9. Last, set your Python interpreter by hitting Shift-Ctrl-P (Windows) / Shift-Cmd-P (Mac) and then typing Python: Select Interpreter.
   
   
   Now select azureml_py310_sdkv2
   
   Then start a new Terminal and install Scikit-Learn, etc.

   pip install -r src/requirements.txt 
   

Troubleshooting

Also see reported issues here: https://github.com/danielsc/v2workshop/issues

1. Cannot find resource group / workspace
   If az ml compute list
   gives you the error resource group could not be found
   Then you need to set your default subscription with

       az account set -s <subscription_id>
   

2. When trying to Python: Select Interpreter no envrionments are show.
   In some cases, hitting the refresh button on the top right has helped.
   

3. When az login --identity does not succeed
   In this case, you will need to log in to the compute instance via jupyter