Estimator Framework Related Packages when specifying conda dependencies file

[phonosync]

I find it more convenient to specify additional conda dependencies in a conda_env.yaml file rather then using conda_packages in the Estimator constructor.
The documentation warns me for the conda_dependencies_file parameter: ...If specified, Azure ML will not install any framework related packages. But I could not find a list of these "framework related packages", which are needed, for example, to mount datastores properly. It turns out, including azureml-defaults in the pip dependencies section of the conda yaml files gives me all the functionality I need, but there might be other packages that I missed but would be necessary for other use cases.
Therefore, I would suggest to add a link in the documentation of the conda_dependencies_file parameter to a complete list of these "framework related packages".

---

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: 0db16bca-3797-0285-3ef2-99de84e4a0be
• Version Independent ID: cba716ef-395b-c8f9-2cbc-967bbc338c87
• Content: azureml.train.estimator.Estimator class - Azure Machine Learning Python
• Content Source: AzureML-Docset/stable/docs-ref-autogen/azureml-train-core/azureml.train.estimator.Estimator.yml
• Service: machine-learning
• Sub-service: core
• GitHub Login: @j-martens
• Microsoft Alias: jmartens

[RohitMungi-MSFT]

@phonosync Thanks for the feedback. @mx-iao Is it possible to add links to conda_dependencies_file as requested and list all the suggested framework related packages?

[jpe316]

Estimators are being deprecated - the recommended path will be to use ScriptRunConfig and Environments to specify these. We will follow up with a new example, thank you!