mlos_bench.config.schemas.config_schemas

A simple class for describing where to find different json config schemas and validating configs against them.

Used by the Launcher and ConfigPersistenceService to validate configs on load.

Notes

• See mlos_bench/config/schemas/README.md for additional documentation in the source tree.

• See mlos_bench/config/README.md for additional config examples in the source tree.

Attributes

CONFIG_SCHEMA_DIR	The local directory where all config schemas shipped as a part of the
SCHEMA_STORE	Static SchemaStore instance used for storing and retrieving schemas for
VALIDATION_ENV_FLAG	The special environment flag to set to skip schema validation when "true".

Classes

ConfigSchema	An enum to help describe schema types and help validate configs against them.
SchemaStore	A simple class for storing schemas and subschemas for the validator to

Module Contents

class mlos_bench.config.schemas.config_schemas.ConfigSchema(*args, **kwds)

Bases: enum.Enum

An enum to help describe schema types and help validate configs against them.

validate(config: dict) → None

Validates the given config against this schema.

Parameters:

config (dict) – The config to validate.

Raises:

• jsonschema.exceptions.ValidationError – On validation failure.

• jsonschema.exceptions.SchemaError – On schema loading error.

Return type:

None

CLI = b''

Json config schema for mlos_bench CLI configuration.

See also

mlos_bench.config

documentation on the configuration system.

mlos_bench.launcher.Launcher

class is responsible for processing the CLI args.

ENVIRONMENT = b''

Json config schema for environments.

GLOBALS = b''

Json config schema for global variables.

OPTIMIZER = b''

Json config schema for optimizers.

SCHEDULER = b''

Json config schema for schedulers.

SERVICE = b''

Json config schema for services.

STORAGE = b''

Json config schema for storage instances.

TUNABLE_PARAMS = b''

Json config schema for tunables instances.

TUNABLE_VALUES = b''

Json config schema for values of TunableGroups instances.

These can be used to specify the values of the tunables for a given experiment using the OneShotOptimizer for instance.

UNIFIED = b''

Combined global json schema use to validate any mlos_bench config file (e.g., *.mlos.jsonc files).

See also

<https://www.schemastore.org/>

property schema: dict

Gets the schema object for this type.

Return type:

dict

class mlos_bench.config.schemas.config_schemas.SchemaStore

Bases: collections.abc.Mapping

A simple class for storing schemas and subschemas for the validator to reference.

__getitem__(key: str) → dict

Gets the schema object for the given key.

Parameters:

key (str)

Return type:

dict

__iter__() → collections.abc.Iterator

Return type:

collections.abc.Iterator

__len__() → int

Return type:

int

property registry: referencing.Registry

Returns a Registry object with all the schemas loaded.

Return type:

referencing.Registry

mlos_bench.config.schemas.config_schemas.CONFIG_SCHEMA_DIR = b''

The local directory where all config schemas shipped as a part of the mlos_bench module are stored.

mlos_bench.config.schemas.config_schemas.SCHEMA_STORE

Static SchemaStore instance used for storing and retrieving schemas for config validation.

mlos_bench.config.schemas.config_schemas.VALIDATION_ENV_FLAG = 'MLOS_BENCH_SKIP_SCHEMA_VALIDATION'

The special environment flag to set to skip schema validation when “true”.

Useful for local development when you’re making a lot of changes to the config or adding new classes that aren’t in the main repo yet.