import warnings
from alibi_detect.saving.schemas import (
DETECTOR_CONFIGS, DETECTOR_CONFIGS_RESOLVED)
from alibi_detect.version import __version__
[docs]def validate_config(cfg: dict, resolved: bool = False) -> dict:
"""
Validates a detector config dict by passing the dict to the detector's pydantic model schema.
Parameters
----------
cfg
The detector config dict.
resolved
Whether the config is resolved or not. For example, if resolved=True, `x_ref` is expected to be a
np.ndarray, wheras if resolved=False, `x_ref` is expected to be a str.
Returns
-------
The validated config dict, with missing fields set to their default values.
"""
# Get detector name and meta
if 'name' in cfg:
detector_name = cfg['name']
else:
raise ValueError('`name` missing from config.toml.')
# Validate detector specific config
if detector_name in DETECTOR_CONFIGS.keys():
if resolved:
cfg = DETECTOR_CONFIGS_RESOLVED[detector_name](**cfg).dict()
else:
cfg = DETECTOR_CONFIGS[detector_name](**cfg).dict()
else:
raise ValueError(f'Loading the {detector_name} detector from a config.toml is not yet supported.')
# Get meta data
meta = cfg.get('meta')
meta = {} if meta is None else meta # Needed because pydantic sets meta=None if it is missing from the config
version_warning = meta.get('version_warning', False)
version = meta.get('version', None)
# Raise warning if config file already contains a version_warning
if version_warning:
warnings.warn('The config file appears to be have been generated from a detector which may have been '
'loaded with a version mismatch. This may lead to breaking code or invalid results.')
# check version
if version is not None and version != __version__:
warnings.warn(f'Config is from version {version} but current version is '
f'{__version__}. This may lead to breaking code or invalid results.')
cfg['meta'].update({'version_warning': True})
return cfg