Missing features in az ml environment CLIv2 that were present in v1
See original GitHub issueRelated command az ml environment create
Is your feature request related to a problem? Please describe. When creating an environment with a Docker image that is located in a private registry, there is no current way to authenticate with the private registry in the YAML configuration file. Trying to create the image results in an “Unauthorized” error message, as we are not able to pass in credentials.
Describe the solution you’d like In the Azure ML Python SDK v1, there is an option to define a private registry address as well as username and password. Here is the “fastai” demo code:
fastai_env.docker.base_image_registry.address = "myregistry.azurecr.io"
fastai_env.docker.base_image_registry.username = "username"
fastai_env.docker.base_image_registry.password = "password"
Please include options in the Environment YAML configuration (as well as the SDKv2) to match existing features in the Python SDK v1.
Describe alternatives you’ve considered As a work-around, we are currently using the SDK v1 to register the environment that uses a private registry. After the environment has been registered with v1, we can use the CLI v2 to perform training, model deployments, etc.
Additional context
Issue Analytics
- State:
- Created a year ago
- Comments:7 (1 by maintainers)
Top GitHub Comments
Fantastic, thanks!
Unfortunately, the schemas for the credentials are not published. I will work with the team to publish them during upcoming release. Meanwhile sharing the schemas for credentials below:
PAT Credential Configuration
Username Password Credential Configuration
Managed Identity Credential Configuration
Service Principal Credentials Configuration