Upgrade script run to SDK v2

In SDK v2, "experiments" and "runs" are consolidated into jobs.

A job has a type. Most jobs are command jobs that run a command, like python main.py. What runs in a job is agnostic to any programming language, so you can run bash scripts, invoke python interpreters, run a bunch of curl commands, or anything else.

To upgrade, change your code for submitting jobs to SDK v2. You don't need to upgrade what you run within the job to SDK v2. However, remove any code specific to Azure Machine Learning from your model training scripts. This separation allows for an easier transition between local and cloud and is considered best practice for mature MLOps. In practice, this means removing azureml.* lines of code. Replace model logging and tracking code with MLflow. For more details, see how to use MLflow in v2.

This article gives a comparison of scenarios in SDK v1 and SDK v2.

Submit a script run

• SDK v1

  from azureml.core import Workspace, Experiment, Environment, ScriptRunConfig
  
  # connect to the workspace
  ws = Workspace.from_config()
  
  # define and configure the experiment
  experiment = Experiment(workspace=ws, name='day1-experiment-train')
  config = ScriptRunConfig(source_directory='./src',
                              script='train.py',
                              compute_target='cpu-cluster')
  
  # set up pytorch environment
  env = Environment.from_conda_specification(
      name='pytorch-env',
      file_path='pytorch-env.yml')
  config.run_config.environment = env
  
  run = experiment.submit(config)
  
  aml_url = run.get_portal_url()
  print(aml_url)
  

• SDK v2

  #import required libraries
  from azure.ai.ml import MLClient, command
  from azure.ai.ml.entities import Environment
  from azure.identity import DefaultAzureCredential
  
  #connect to the workspace
  ml_client = MLClient.from_config(credential=DefaultAzureCredential())
  
  # set up pytorch environment
  env = Environment(
      image="mcr.microsoft.com/azureml/openmpi4.1.0-ubuntu22.04",
      conda_file="pytorch-env.yml",
      name="pytorch-env"
  )
  
  # define the command
  command_job = command(
      code="./src",
      command="python train.py",
      environment=env,
      compute="cpu-cluster",
  )
  
  returned_job = ml_client.jobs.create_or_update(command_job)
  print(f"Job submitted. View in studio: {returned_job.studio_url}")
  

Mapping of key functionality in v1 and v2

Next steps

For more information, see:

• V1 - Experiment
• V2 - Command Job
• Train models with the Azure Machine Learning Python SDK v2