[Azure] DP-100 자격증 Build AI solutions with Azure Machine Learning - Part 1

출처 : Microsoft Learn

• Azure Machine Learning SDK
  • Scalable, on-demand compute for machine learning workloads
  • Data storage and connectivity to ingest data from a wide range sources
  • ML workflow orchestration to automate model training, deployment, and management processes
  • Model registration and management to track multiple versions of models and the data on which they were trained
  • Metrics and monitoring for training experiments, datasets, and published services
  • Model deployment for real-time and batch inferencing

 

출처 : Microsoft Learn

• Azure Machine Learning Workspaces - A context for the experiments, data, compute targets, and other assets associated with a machine learning workload.
  • Creating a Workspace with Python and Azure Command Line Interface (CLI)

---Python---
from azureml.core import Workspace
    
    ws = Workspace.create(name='aml-workspace', 
                      subscription_id='123456-abc-123...',
                      resource_group='aml-resources',
                      create_resource_group=True,
                      location='eastus'
                     )
                     
---CLI---             
az ml workspace create -w 'aml-workspace' -g 'aml-resources'

 

• Logging - From each run of the experiment, and be able to retrieve them easily from each run.
  • Azure ML SDK: Metric - run.log(name, val)
  • Python printing/logging: Log - print(val) or logging.info(message)
  • OpenCensus Python: Log - logger.addHandler(AzureLogHandler()) or logging.log(message)
• ScriptRunConfig - To run a script-based experiemnt that trains a machine learning model.
• Script Arguments - Use a library such as "argparse" to set the regularization rate hyperparameter for an algorithm used to train a model.
• Registering a Model - To track multiple versions of a model, and retrieve models for inferencing (prodicting label values from new data). Specify a name, description, tags, framework, framework version, custom properties, and other useful metadata.
• Datastore - Abstractions for cloud data sources. To add a datastore to your workspace, you can register it using the graphical interface in Azure Machine Learning studio.
  • Azure Storage (blob and file containers)
  • Azure Data Lake stores
  • Azure SQL Database
  • Azure Databricks file system
• Tubular Dataset - Provide the easiest way to consume structured data as a Pandas dataframe.

출처 : Microsoft Learn

• Environment for the script - Include all packages on which the script depends.
• Compute Target - Environment will be deployed and the script run
  • Local Compute - Pysical workstation or a virtual machine such as an Azure Machine Learning compute instance.
  • Compute Clusters - For experiment workloads with high scalability requirements.
  • Attached Compute - If Azure-based compute environment for data science is already used, attach it to your Azure Machine Learning workspace.