Edit

Upgrade AutoML to SDK v2

  • Article

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

In SDK v1, AutoML was primarily configured and run using the AutoMLConfig class. In SDK v2, this class has been converted to an AutoML job. Although there are some differences in the configuration options, by and large, naming & functionality has been preserved in V2.

This article gives a comparison of scenario(s) in SDK v1 and SDK v2.

Submit AutoML run

  • SDK v1: Below is a sample AutoML classification task. For the entire code, check out our examples repo.

    # Imports

import azureml.core
from azureml.core.experiment import Experiment
from azureml.core.workspace import Workspace
from azureml.core.dataset import Dataset
from azureml.train.automl import AutoMLConfig
from azureml.train.automl.run import AutoMLRun   

# Load tabular dataset
data = "<url_to_data>"
dataset = Dataset.Tabular.from_delimited_files(data)
training_data, validation_data = dataset.random_split(percentage=0.8, seed=223)
label_column_name = "Class"

# Configure Auto ML settings
automl_settings = {
    "n_cross_validations": 3,
    "primary_metric": "average_precision_score_weighted",
    "enable_early_stopping": True,
    "max_concurrent_iterations": 2,  
    "experiment_timeout_hours": 0.25,  
    "verbosity": logging.INFO,
}

# Put together an AutoML job constructor
automl_config = AutoMLConfig(
    task="classification",
    debug_log="automl_errors.log",
    compute_target=compute_target,
    training_data=training_data,
    label_column_name=label_column_name,
    **automl_settings,
)

# Submit run
remote_run = experiment.submit(automl_config, show_output=False)
azureml_url = remote_run.get_portal_url()
print(azureml_url)

  • SDK v2: Below is a sample AutoML classification task. For the entire code, check out our examples repo.

    # Imports
from azure.ai.ml import automl, Input, MLClient

from azure.ai.ml.constants import AssetTypes
from azure.ai.ml.automl import (
    classification,
    ClassificationPrimaryMetrics,
    ClassificationModels,
)


# Create MLTables for training dataset
# Note that AutoML Job can also take in tabular data
my_training_data_input = Input(
    type=AssetTypes.MLTABLE, path="./data/training-mltable-folder"
)

# Create the AutoML classification job with the related factory-function.
classification_job = automl.classification(
    compute="<compute_name>",
    experiment_name="<exp_name?",
    training_data=my_training_data_input,
    target_column_name="<name_of_target_column>",
    primary_metric="accuracy",
    n_cross_validations=5,
    enable_model_explainability=True,
    tags={"my_custom_tag": "My custom value"},
)

# Limits are all optional
classification_job.set_limits(
    timeout_minutes=600,
    trial_timeout_minutes=20,
    max_trials=5,
    max_concurrent_trials = 4,
    max_cores_per_trial= 1,
    enable_early_termination=True,
)

# Training properties are optional
classification_job.set_training(
    blocked_training_algorithms=["LogisticRegression"],
    enable_onnx_compatible_models=True,
)

# Submit the AutoML job
returned_job = ml_client.jobs.create_or_update(classification_job)  
returned_job

Mapping of key functionality in SDK v1 and SDK v2

Functionality in SDK v1 Rough mapping in SDK v2
Method/API in SDK v1 (use links to ref docs) Method/API in SDK v2 (use links to ref docs)

Next steps

For more information, see: