This project attempts to reduce the boilerplate code when launching simple azureml jobs from command line.
It aims to:
- Be object oriented to enable end users to extend the SDK for their custom workflows, trying to reuse AzureML SDK V2 native classes as much as possible
- Be config driven (thanks to jsonargparse) to improve reusability of objects and code
- Be, as it name states, easy to use (hence focused in simple usecases)
- Unify through wrappers the different flows and runs in Azure through a common interface
You can check more info at the documentation site
Simply run
pip install ez-azml
The main entrypoint is the CLI command. The following command will show you the help
ez-azml -h
You can run an example with (if you have cloned the repo)
ez-azml --config configs/command/main.yaml run
PYTHONPATH=$PYTHONPATH:. ez-azml --config configs/pipeline/main.yaml run
(PYTHONPATH
to point to the examples dir. This is not needed if your pipeline is available in any package as an importable function)