Metadata-Version: 2.1
Name: irisml-tasks-azureml
Version: 0.0.1
Summary: Utility tools to use IrisML on AzureML
Home-page: https://github.com/microsoft/irisml-tasks-azureml
Author: irisdev
License: MIT
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE

# irisml-tasks-azureml

This package is a part of IrisML pipeline. It provides utility scripts and tasks to work with AzureML.

## Installation
```bash
pip install irisml-tasks-azureml
```

## Available commands
```bash
irisml_run_aml [-h] [--env ENV] [--cache_url CACHE_URL] [--no_cache] [--include_local_tasks [INCLUDE_LOCAL_TASKS]] [--custom_packages CUSTOM_PACKAGES [CUSTOM_PACKAGES ...]]
                      [--extra_index_url EXTRA_INDEX_URL] [--no_wait] [--compute_target COMPUTE_TARGET] [--subscription_id SUBSCRIPTION_ID] [--workspace WORKSPACE] [--experiment EXPERIMENT]
                      [--base_docker_image BASE_DOCKER_IMAGE] [--base_docker_image_registry BASE_DOCKER_IMAGE_REGISTRY]
                      job_filepath
```

You can run an experiment on a remote AzureML node. If --include_local_tasks option is used, python scripts in the current directory or the specified directory will be sent to AzureML and be loaded as custom tasks.

Example
```bash
irisml_run_aml irisml/docs/example/mobilenetv2_mnist_training.json -p irisml-tasks-torchvision irisml-tasks-training --compute_target <cluster_name> --subscription_id <subscription_id> --workspace <workspacename>
```

## Tasks
### run_azureml_child
Submit a new AzureML job as a child of the current run. Raises an exception if the current environment was not on AzureML.
