New error with 'az ml model deploy' in 1.0.81

[adiun]

Extension name (the extension in question)

"extensionType": "whl",
"name": "azure-cli-ml",
"version": "1.0.81"

Description of issue (in as much detail as possible)

When running the following model deployment command, with azure-cli-ml 1.0.81 this is now throwing an error. This did not throw an error in azure-cli-ml 1.0.79.

az ml model deploy -n quote-clf --inference-config-file src/inferconfig.json --auth-enabled true --deploy-config-file src/deployconfig_stab.json --compute-target aks-aml-stab -m quote_clf:3 -m quote_tfidf:3 --workspace-name aml-stab --resource-group aml-stab --overwrite -v

Error:

Creating image
Running.............................................................................
Succeeded
ERROR: {'Azure-cli-ml Version': '1.0.81', 'Error': AttributeError("'bool' object has no attribute 'name'",)}
Image creation operation finished for image quote-clf:17, operation "Succeeded"
##[error]Script failed with error: Error: The process '/bin/bash' failed with exit code 1

src/inferconfig.json:

{
    "condaFile": "environment.yaml",
    "enableGpu": false,
    "entryScript": "entry_script.py",
    "extraDockerfileSteps": "extra_dockerfile_steps",
    "runtime": "python",
    "sourceDirectory": "src"
}

src/deployconfig_stab.json:

{
    "autoScaler": {
        "autoscaleEnabled": true,
        "minReplicas": 2,
        "maxReplicas": 4,
        "refreshPeriodInSeconds": 1,
        "targetUtilization": 70
    },
    "authEnabled": true,
    "computeType": "aks",
    "containerResourceRequirements": {
        "cpu": 1,
        "memoryInGB": 1
    },
    "enableAppInsights": true,
    "maxConcurrentRequestsPerContainer": 3,
    "maxQueueWaitMs": 5000,
    "scoringTimeoutMs": 120000
}

src/environment.yaml:

name: quote
channels:
  - defaults
  - conda-forge
dependencies:
  - pip=19.2.2
  - scikit-learn=0.20.3
  - spacy=2.1.7
  - pip:
      - azureml-defaults>=1.0.60
      - https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-2.1.0/en_core_web_sm-2.1.0.tar.gz

[rsethur]

@adiun: I ran into the same error. As a intermediate fix use the yml format for inference config - it worked for me. Template for inference_config.yml: https://pastebin.com/CzX5pViy

[gogowings]

We have located the root cause and are working on releasing a fix.

[gogowings]

A related comment: the bug is only applicable in the deployment which doesn't use Azure Machine Learning Environment. We encourage the usage of Environment based deployment which is faster and more reliable.

[gogowings]

@adiun, in your case, using extraDockerfileSteps in InferenceConfig file makes the deployment goes to the non-Environment route. You can consider switch to Environment base deployment documented at: https://docs.microsoft.com/en-us/azure/machine-learning/how-to-use-environments#enable-docker

[adiun]

@rsethur Thanks! @gogowings Thanks for the updates and the tip about using extraDockerFileSteps. I received this guidance separately as well re. deployment performance so will definitely look into switching to that.