Azure Machine Learning Jobs
Jobs are a key component of Azure Machine Learning. They provide a flexible way to run your machine learning workloads. This document describes how to use jobs in Azure Machine Learning.
Definition
Jobs are the fundamental unit of execution in Azure Machine Learning. They represent a discrete set of steps or tasks that are executed in sequence or concurrently. Think of them as independent processes or containers for your machine learning code.
Types of Jobs
Azure Machine Learning supports several types of jobs, offering flexibility for different scenarios:
- Compute Jobs: These jobs run on compute resources—either on-demand compute or a virtual machine.
- Managed Jobs: These jobs simplify the orchestration of complex workflows and manage dependencies between steps.
- On-Demand Compute Jobs: These jobs run on compute resources that are provisioned on-demand.
Example
The following is an example of a Managed Job:
# Define the Job Compute
job_compute = {
type: "AzureMachineLearning",
image: "mcr.microsoft.com/azureml/managed-compute/python:3.9",
min_count: 1,
max_count: 1,
vm_size: "Standard_DS3_v2",
# ... other configuration options ...
}
# Define the Job
job = {
name: "MyJob",
compute: job_compute,
code_filepath: "my_notebook.ipynb"
}