Helpers

Utility functions for working with artifact schemas.

mlopus.artschema.get_schemas(subject)

Parse artifact schema tags from subject.

See also: mlopus.artschema.Tags.parse_subject()

Parameters:

subject (Experiment | Run | Model | ModelVersion) –

Experiment, run, model or model version.

Return type:

Tags

mlopus.artschema.get_schema(subject, alias=None)

Get artifact schema class specification from subject.

Parameters:

• subject (Experiment | Run | Model | ModelVersion) –

  Experiment, run, model or model version.

• alias (Optional[str]) –

  Alias of a previously registered schema for this subject.

  Defaults to default.

Return type:

ClassSpec

mlopus.artschema.load_artifact(subject, path_in_run=None, schema=None, loader_conf=None, skip_reqs_check=False, dry_run=False)

Load artifact of run or model version using schema.

Parameters:

• subject (RunApi | ModelVersionApi) –

  Run or model version with API handle.

• path_in_run (Optional[str]) –

  See mlopus.mlflow.BaseMlflowApi.load_run_artifact.path_in_run

  If subject is a model version, defaults to model name.

• schema (Union[Schema, Type[Schema], str, None]) –

  • Type or instance of Schema

  • Fully qualified name of a Schema class (e.g.: package.module:Class)

  • Alias of a schema previously registered for this run/model version or its parent experiment/model (see mlopus.artschema.Tags).

• loader_conf (Union[TypeVar(L, bound= Loader), dict, None]) –

  See Schema.get_loader.loader

• skip_reqs_check (bool) –

  If schema is specified by alias, ignore the registered package requirement.

  See mlopus.artschema.ClassSpec.load()

• dry_run (bool) –

  See dry_run

Return type:

Union[TypeVar(A, bound= object), Path]

Returns:

• If dry_run is True: A Path to the cached artifact, after being verified.

• Otherwise: An instance of Artifact

mlopus.artschema.log_run_artifact(artifact, run, path_in_run=None, schema=None, dumper_conf=None, skip_reqs_check=False, auto_register=False, keep_the_source=None, allow_duplication=None, use_cache=None)

Publish run artifact using schema.

Parameters:

• artifact (Union[TypeVar(A, bound= object), dict, Path]) –

  See Schema.get_dumper.artifact

• run (RunApi) –

  Run API object.

• path_in_run (Optional[str]) –

  See mlopus.mlflow.BaseMlflowApi.log_run_artifact.path_in_run

• schema (Union[Schema, Type[Schema], str, None]) –

  • Type or instance of Schema

  • Fully qualified name of a Schema class (e.g.: package.module:Class)

  • Alias of a schema previously registered for this run or its parent experiment (see mlopus.artschema.Tags).

• dumper_conf (Union[TypeVar(D, bound= Dumper), dict, None]) –

  See Schema.get_dumper.dumper

• skip_reqs_check (bool) –

  If schema is specified by alias, ignore the registered package requirement.

  See mlopus.artschema.ClassSpec.load()

• auto_register (bool | dict) –

  After a successful artifact publish, register the used schema in the run tags.

  If a non-empty dict is passed, it is used as keyword arguments for Tags.using().

  If the schema was specified by alias, that alias is used by default.

• keep_the_source (Optional[bool]) –

  See keep_the_source

  (the source in this case is a callback, unless artifact is a Path)

• allow_duplication (Optional[bool]) –

  See allow_duplication

• use_cache (Optional[bool]) –

  See use_cache

Return type:

None

mlopus.artschema.log_model_version(artifact, model, run, path_in_run=None, schema=None, dumper_conf=None, skip_reqs_check=False, auto_register=False, keep_the_source=None, allow_duplication=None, use_cache=None, version=None, tags=None)

Log artifact as model version using schema.

Example:

mlflow = mlopus.mlflow.get_api()

version = mlopus.artschema.log_model_version(
    my_artifact,
    schema=MySchema,
    run=mlflow.start_run(...),
    model=mlflow.get_or_create_model(...),
    auto_register={"aliased_as": "foobar"}  # register `MySchema` as `foobar`
)

mlopus.artschema.load_artifact(version, schema="foobar")

Parameters:

• artifact (Union[TypeVar(A, bound= object), dict, Path]) –

  See Schema.get_dumper.artifact

• model (ModelApi) –

  Model API object.

• run (Run | str) –

  Run API object.

• path_in_run (Optional[str]) –

  See mlopus.mlflow.BaseMlflowApi.log_model_version.path_in_run

• schema (Union[Schema, Type[Schema], str, None]) –

  • Type or instance of Schema

  • Fully qualified name of a Schema class (e.g.: package.module:Class)

  • Alias of a schema previously registered for this run or its parent experiment (see mlopus.artschema.Tags).

• dumper_conf (Union[TypeVar(D, bound= Dumper), dict, None]) –

  See Schema.get_dumper.dumper

• skip_reqs_check (bool) –

  If schema is specified by alias, ignore the registered package requirement.

  See mlopus.artschema.ClassSpec.load()

• auto_register (bool | dict) –

  After a successful artifact publish, register the used schema in the new model version tags.

  See also log_run_artifact.auto_register

• keep_the_source (Optional[bool]) –

  See keep_the_source

  (the source in this case is a callback, unless artifact is a Path)

• allow_duplication (Optional[bool]) –

  See allow_duplication

• use_cache (Optional[bool]) –

  See use_cache

• version (Optional[str]) –

  See mlopus.mlflow.BaseMlflowApi.log_model_version.version

• tags (Optional[Mapping]) –

  See mlopus.mlflow.BaseMlflowApi.log_model_version.version

Return type:

ModelVersionApi