Compute Instance, Attached Compute in ML Studio issues

Question

Compute Instance, Attached Compute in ML Studio issues

Carmine Basile 0

I am not able to attach a Virtual Machine I created with GPU to My ML workplace in ML Studio. I honestly don't understand why things have to be so difficult in Azure.

It is telling me to modify firewall, to set up private IP, Storage in the VN, I am loosing my patience, how without any python code can I bloody attach the VM on the ML studio so I can run some ML automated and see if things is of any use? It's 3 days I spent to learn about Azure, with no success at all

Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-04-25T06:57:43.3533333+00:00

Hi Carmine Basile,

I understand how frustrating it is to spend so much time just trying to attach a virtual machine to Azure ML Studio. The main reason you're facing these issues is because Azure has strict networking and security rules, especially when the virtual machine (VM) is created outside of the ML Studio. When a VM is inside a Virtual Network (VNet), Azure requires things like private IP addresses, storage accounts in the same network, firewall rule changes, and sometimes private endpoints to allow the VM to communicate with the ML workspace. All of this is meant for advanced security in enterprise environments, but it can feel overwhelming.

The easiest and recommended solution is to create the VM (called a Compute Instance) directly from inside ML Studio. When you do that, Azure automatically sets everything up correctly in the background, and you won’t need to deal with any network settings, IPs, or firewalls. It just works. This way, you can quickly start using the VM for training, AutoML, or notebooks without extra setup. I truly understand your frustration, and I recommend this simpler route to avoid further delays.

I hope this information helps. Thank you!
Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-04-28T05:38:55.67+00:00

Hi Carmine Basile,
Did you get any chance to check the response. Thank you!
Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-04-29T07:53:32.6366667+00:00

Hi Carmine Basile,
Just following up to see if you had a chance to review the above response. Thank you!
Carmine Basile 0 Reputation points

2025-04-30T10:14:05.3333333+00:00

Dear Pavankumar, thanks for your reply. I followed your suggestion and slowly managing to train my first model and do adjusments. Can I please ask if you can assist with another issue I am encountering: I cannot find in my Azure ML Designer the component: "Execute a Python Script". Can you explain why and how I can make this work? Many thanks
Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-05-02T07:29:46.4766667+00:00

Hi Carmine Basile,

If you’re unable to find the "Execute Python Script" component in Azure ML Designer, it’s likely because this component is only available in the classic Azure Machine Learning Studio (Studio Classic), not in the newer Azure Machine Learning Designer under Azure Machine Learning Service. The newer Designer focuses more on low-code/no-code pipelines with prebuilt modules and doesn’t directly support executing arbitrary Python scripts within a pipeline component. To achieve similar functionality, you would need to create a custom Python module by writing and registering your script as a custom component, or alternatively, switch to using notebooks or Python SDK for more flexibility. If your goal is quick experimentation with Python code, using a compute instance with Jupyter notebooks inside Azure ML might be an easier path.

Your answer

Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-04-25T06:57:43.3533333+00:00

Hi Carmine Basile,

I understand how frustrating it is to spend so much time just trying to attach a virtual machine to Azure ML Studio. The main reason you're facing these issues is because Azure has strict networking and security rules, especially when the virtual machine (VM) is created outside of the ML Studio. When a VM is inside a Virtual Network (VNet), Azure requires things like private IP addresses, storage accounts in the same network, firewall rule changes, and sometimes private endpoints to allow the VM to communicate with the ML workspace. All of this is meant for advanced security in enterprise environments, but it can feel overwhelming.

The easiest and recommended solution is to create the VM (called a Compute Instance) directly from inside ML Studio. When you do that, Azure automatically sets everything up correctly in the background, and you won’t need to deal with any network settings, IPs, or firewalls. It just works. This way, you can quickly start using the VM for training, AutoML, or notebooks without extra setup. I truly understand your frustration, and I recommend this simpler route to avoid further delays.

I hope this information helps. Thank you!
Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-04-28T05:38:55.67+00:00

Hi Carmine Basile,
Did you get any chance to check the response. Thank you!
Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-04-29T07:53:32.6366667+00:00

Hi Carmine Basile,
Just following up to see if you had a chance to review the above response. Thank you!
Carmine Basile 0 Reputation points

2025-04-30T10:14:05.3333333+00:00

Dear Pavankumar, thanks for your reply. I followed your suggestion and slowly managing to train my first model and do adjusments. Can I please ask if you can assist with another issue I am encountering: I cannot find in my Azure ML Designer the component: "Execute a Python Script". Can you explain why and how I can make this work? Many thanks
Pavankumar Purilla 7,015 Reputation points Microsoft External Staff

2025-05-02T07:29:46.4766667+00:00

Hi Carmine Basile,

If you’re unable to find the "Execute Python Script" component in Azure ML Designer, it’s likely because this component is only available in the classic Azure Machine Learning Studio (Studio Classic), not in the newer Azure Machine Learning Designer under Azure Machine Learning Service. The newer Designer focuses more on low-code/no-code pipelines with prebuilt modules and doesn’t directly support executing arbitrary Python scripts within a pipeline component. To achieve similar functionality, you would need to create a custom Python module by writing and registering your script as a custom component, or alternatively, switch to using notebooks or Python SDK for more flexibility. If your goal is quick experimentation with Python code, using a compute instance with Jupyter notebooks inside Azure ML might be an easier path.

Share via

Compute Instance, Attached Compute in ML Studio issues

Your answer