azure.ml.component package

Subpackages

azure.ml.component.dsl package

Submodules

azure.ml.component.component module

azure.ml.component.environment module

Contracts for environment override.

class azure.ml.component.environment.Conda(*, pip_requirements_file: str | Path = None, conda_file: str | Path = None, **kwargs)

Bases: object

Conda environment settings for running a component.

Note

The following example shows how to create a conda environment settings using azure.ml.component.environment.Conda class:

# Create from pip requirements file
conda = Conda(pip_requirements_file="<path-to-pip-requirements-file>")
# Create from conda yaml file
conda = Conda(conda_file="<path-to-conda-file>")

For more details of working with conda, see environment in runsettings.

class azure.ml.component.environment.Docker(*, image: str = None, file: str | Path = None, **kwargs)

Bases: object

Docker environment settings for running a component.

Note

The following example shows how to create a docker environment settings using azure.ml.component.environment.Docker class:

# Create from image
docker = Docker(image="mcr.microsoft.com/azureml/openmpi3.1.2-ubuntu18.04")
# Create from dockerfile
docker = Docker(file="<path-to-dockerfile>")

For more details of working with docker, see environment in runsettings.

class azure.ml.component.environment.Environment(*, workspace: Workspace = None, name: str = None, version: str = None, conda_file: str | PathLike = None, file: str | PathLike = None, **kwargs)

Bases: object

Configure Python environment for running components.

An Environment defines Docker, Conda and OS settings. An Environment can be initialized from a curated AML environment (For more information about curated environments, see create and manage reusable environments), or a YAML file which defines environment setting by following environment specs.

Note

The following example shows how to create an environment using azure.ml.component.environment.Environment class:

# Create from a curated environment, suppose we have a workspace as 'ws'
env = Environment(workspace=ws, name="AzureML-Designer")
# Or use a curated environment with a specific version
env = Environment(workspace=ws, name="AzureML-Designer", version="19")

# Create from a YAML file, see below YAML as an example
env = Environment(file="<path-to-YAML-file>")

# Sample environment YAML
docker:
  image:  mcr.microsoft.com/azureml/openmpi3.1.2-ubuntu18.04
conda:
  conda_dependencies:
    name: project_environment
    channels:
    - defaults
    dependencies:
    - python=3.8.10
    - pip=20.0
    - pip:
      - azureml-defaults
      - azureml-dataprep>=1.6
os: Linux

# Create from an conda YAML file, see below YAML as an example
env = Environment(conda_file="<path-to-conda-YAML-file>")

# Sample conda YAML
name: project_environment
channels:
- defaults
dependencies:
- python=3.8.10
- pip=20.0
- pip:
  - azureml-defaults
  - azureml-dataprep>=1.6

Note

If you set some keys in [‘docker’, ‘conda’, ‘os’] and some in [‘name’, ‘version’] at the same time, keys in [‘docker’, ‘conda’, ‘os’] have higher priority.

For more samples of working with environments, see environment in runsettings.

azure.ml.component.pipeline module

azure.ml.component.run module

azure.ml.component.run_settings module

Defines classes for run settings.

class azure.ml.component.run_settings.PipelineRunSettings(display_section: _RunSettingsParamSection, store: _RunSettingsStore, workspace)

Bases: RunSettings

A RunSettings aggregates the run settings of a pipeline.

property continue_on_failed_optional_input: Pipeline level continue run on failed optional input. The default value is true.

property continue_on_step_failure: Whether continue pipeline run when a step run fail.

property default_compute: Pipeline level default compute.

property default_datastore: Pipeline level default datastore.

property force_rerun: Whether enforce to rerun the pipeline.

property priority: PipelineRunSettingsPriority: Pipeline level default job’s priority in Cosmos or compute cluster.

property timeout_seconds: Pipeline level time out in seconds.

class azure.ml.component.run_settings.PipelineRunSettingsPriority(priority_section, workspace)

Bases: object

For strong type intellisense of ‘runsettings.priority.scope’.

property compute_cluster

Pipeline level default job’s priority in compute cluster.

When not specified, system default value will be used.

property scope

Pipeline level default Scope job’s priority in Cosmos.

When not specified, system default value (1000) will be used.

class azure.ml.component.run_settings.RunSettings(display_section: _RunSettingsParamSection, store: _RunSettingsStore, workspace)

Bases: WorkspaceTelemetryMixin

A RunSettings aggregates the run settings of a component.

configure(*args, **kwargs)

Configure the runsettings.

Note that this method will be replaced by a dynamic generated one at runtime with parameters that corresponds to the runsettings of the component.

get_runsettings_for_submit(): Generate pipeline runsetting data and pass it to the backend.

validate(raise_error=False, process_error=None, for_definition_validate=False, definition_validator=None, component_type=None)

Validate RunSettings parameter values for current component.

Parameters:

raise_error (bool) – Whether to raise exceptions on error.
process_error (function) – The error process function to be used.
for_definition_validate (bool) – If the validate is for pipeline definition validate.
definition_validator (azure.ml.component._component_validator.DefinitionValidator) – definition validator
component_type (str) – indicate current node type.

Returns:

The errors found during validation.

Return type:

builtin.list

class azure.ml.component.run_settings.SubPipelineRunSettings(display_section: _RunSettingsParamSection, store: _RunSettingsStore, workspace)

Bases: WorkspaceTelemetryMixin

A RunSettings aggregates the run settings of a sub pipeline.

configure(*args, **kwargs): Do nothing, just for compatibility with runsetting functions.

validate(raise_error=False, process_error=None, for_definition_validate=False, definition_validator=None, component_type=None): Do nothing, just for compatibility with runsetting functions.