Docker container for compute instance in machine learning

Question

Hi Everyone,

I'm new to the cloud environment and the Microsoft Azure so incase this question doesn't belong here, my apologies in advance.

I was wondering if it's possible to create a docker container for a python environment within compute instance?

I'm working my way towards getting DP-100 certification. The other day I was trying to follow a tutorial here https://github.com/MicrosoftLearning/mslearn-dp100/blob/main/07%20-%20Work%20with%20Compute.ipynb
There's a section in there 'Create a compute cluster' in which a docker configuration is specified within ScriptRunConfig. I tried running in this code on a compute instance instead of a cluster and got the following error

The code works perfectly fine if I use a compute cluster instead of a compute instance

Answer 1

@Kamal Tanwar Thanks for the question. Can you please add more details about the docker container that you are trying.

Please follow the document to train a model by using custom docker image. The base images are available on Github.
You can also use these Docker images as base images for your custom Azure ML Environments. If you specify any conda dependencies in your Environment, the extra dependencies are installed on top of the dependencies in the Docker image.

Please follow Prebuilt Docker images for inference (before Inference-optimized curated images). This provides with a faster and simpler deployment experience. These prebuilt Docker Images come with popular machine learning frameworks and python packages.