Skip to content

Change model deployment AzDo pre-built task for Python SDK  #136

@Anaig

Description

@Anaig

The ms-air-aiagility.vss-services-azureml.azureml-model-deploy-task.AMLModelDeploy@0 task in the azdo-ci-build-train is failing for AKS because the compute_target_name parameter is not correctly filled by the Azure CLI command.

Using the Python SDK is providing a more stable and successful experience to deploy the model to a compute target.

Command produced by the task:
az ml model deploy --ct akscomputetarget -n aks-name --model sklearn_model.pkl:1 --ic /__w/1/src/azureml/code/score/inference_config.json --dc /__w/1/src/azureml/code/score/deployment_config_aks.json -g rg-main -w aml-main --overwrite

Error issued:
ERROR: {\'Azure-cli-ml Version\': \'1.0.83\', \'Error\': MlCliError({\'Error\': \'Error parsing --deploy-config-file. Must be valid JSON or YAML file.\', \'Response Content\': TypeError("deploy_configuration() got an unexpected keyword argument \'compute_target_name\

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions