def compose(
config_file: Optional[str] = None,
overrides: List[str] = [],
strict: Optional[bool] = None,
) -> DictConfig:
"""
:param config_file: optional config file to load
:param overrides: list of overrides for config file
:param strict: optionally override the default strict mode
:return: the composed config
"""
assert (
GlobalHydra().is_initialized()
), "GlobalHydra is not initialized, use @hydra.main() or call hydra.experimental.initialize() first"
gh = GlobalHydra.instance()
assert gh.hydra is not None
cfg = gh.hydra.compose_config(config_file=config_file,
overrides=overrides,
strict=strict)
assert isinstance(cfg, DictConfig)
if "hydra" in cfg:
del cfg["hydra"]
return cfg
def test_initialize() -> None:
try:
assert not GlobalHydra().is_initialized()
hydra.experimental.initialize(config_dir=None, strict=True)
assert GlobalHydra().is_initialized()
finally:
GlobalHydra().clear()
def test_initialize_hydra():
try:
assert not GlobalHydra().is_initialized()
hydra.experimental.initialize_hydra(task_name="task",
search_path_dir=None,
strict=True)
assert GlobalHydra().is_initialized()
finally:
GlobalHydra().clear()
def test_config_installed(hydra_global_context): # noqa: F811
"""
Tests that color options are available for both hydra/hydra_logging and hydra/job_logging
"""
with hydra_global_context(config_dir="../hydra_plugins/hydra_colorlog/conf"):
config_loader = GlobalHydra().hydra.config_loader
assert "colorlog" in config_loader.get_group_options("hydra/job_logging")
assert "colorlog" in config_loader.get_group_options("hydra/hydra_logging")
def get_hydra():
global_hydra = GlobalHydra()
if not global_hydra.is_initialized():
return Hydra.create_main_hydra_file_or_module(
calling_file=__file__,
calling_module=None,
config_dir="configs",
strict=False,
)
else:
return global_hydra.hydra
def test_initialize_with_config_dir() -> None:
try:
assert not GlobalHydra().is_initialized()
hydra.experimental.initialize(config_dir="../hydra/test_utils/configs",
strict=True)
assert GlobalHydra().is_initialized()
config_search_path = (
GlobalHydra.instance().hydra.config_loader.config_search_path)
idx = config_search_path.find_first_match(
SearchPath(provider="main", search_path=None))
assert idx != -1
finally:
GlobalHydra().clear()
def compose(config_file=None, overrides=[], strict=None):
"""
:param config_file: optional config file to load
:param overrides: list of overrides for config file
:param strict: optionally override the default strict mode
:return: the composed config
"""
assert (
GlobalHydra().is_initialized()
), "GlobalHydra is not initialized, use @hydra.main() or call hydra.experimental.initialize() first"
cfg = GlobalHydra().hydra.compose_config(config_file=config_file,
overrides=overrides,
strict=strict)
if "hydra" in cfg:
del cfg["hydra"]
return cfg
def test_config_installed(
hydra_global_context: TGlobalHydraContext, # noqa: F811
) -> None:
"""
Tests that color options are available for both hydra/hydra_logging and hydra/job_logging
"""
with hydra_global_context(
config_dir="../hydra_plugins/hydra_colorlog/conf"):
gh = GlobalHydra.instance()
assert gh.hydra is not None
config_loader = gh.hydra.config_loader
assert "colorlog" in config_loader.get_group_options(
"hydra/job_logging")
assert "colorlog" in config_loader.get_group_options(
"hydra/hydra_logging")
def __enter__(self):
try:
config_dir, config_file = split_config_path(self.config_path)
hydra = Hydra.create_main_hydra_file_or_module(
calling_file=self.calling_file,
calling_module=self.calling_module,
config_dir=config_dir,
strict=self.strict,
)
self.hydra = hydra
self.temp_dir = tempfile.mkdtemp()
overrides = copy.deepcopy(self.overrides)
overrides.append("hydra.run.dir={}".format(self.temp_dir))
self.job_ret = self.hydra.run(
config_file=config_file, task_function=self, overrides=overrides,
)
strip_node(self.job_ret.cfg, "hydra.run.dir")
return self
finally:
GlobalHydra().clear()
def __enter__(self) -> "SweepTaskFunction":
self.temp_dir = tempfile.mkdtemp()
overrides = copy.deepcopy(self.overrides)
assert overrides is not None
overrides.append("hydra.sweep.dir={}".format(self.temp_dir))
try:
config_dir, config_file = split_config_path(self.config_path)
hydra = Hydra.create_main_hydra_file_or_module(
calling_file=self.calling_file,
calling_module=self.calling_module,
config_dir=config_dir,
strict=self.strict,
)
self.returns = hydra.multirun(
config_file=config_file, task_function=self, overrides=overrides
)
finally:
GlobalHydra().clear()
return self
def __enter__(self):
self.temp_dir = tempfile.mkdtemp()
overrides = copy.deepcopy(self.overrides)
overrides.append("hydra.sweep.dir={}".format(self.temp_dir))
try:
config_dir, config_file = split_config_path(self.config_path)
hydra = Hydra.create_main_hydra_file_or_module(
calling_file=self.calling_file,
calling_module=self.calling_module,
config_dir=config_dir,
strict=self.strict,
)
self.returns = hydra.multirun(
config_file=config_file, task_function=self, overrides=overrides
)
flat = [item for sublist in self.returns for item in sublist]
for ret in flat:
strip_node(ret.cfg, "hydra.sweep.dir")
finally:
GlobalHydra().clear()
return self
from hydra._internal.hydra import GlobalHydra
from hydra.experimental import compose as hydra_compose
from hydra.experimental import initialize as hydra_init
from PIL import Image
from meta_blocks.experiment.eval import evaluate
from meta_blocks.experiment.train import train
logger = logging.getLogger(__name__)
AVAILABLE_METHODS = ("maml", "fomaml", "reptile", "proto")
AVAILABLE_SETTINGS = ("classic_supervised", "self_supervised")
# Initialize hydra.
# TODO: is there a way to check if hydra is initialized using public API?
if not GlobalHydra().is_initialized():
hydra_init(config_dir="conf", strict=False)
@pytest.mark.parametrize("adaptation_method", AVAILABLE_METHODS)
@pytest.mark.parametrize("experiment_setting", AVAILABLE_SETTINGS)
def test_omniglot_integration(adaptation_method, experiment_setting):
def generate_dummy_miniimagenet_data(dir_path):
"""Generates dummy data that imitates mini-ImageNet structure.
Mini-ImageNet is too heavy for integration testing, so we generate
synthetic data (dummy images) that satisfy the mini-ImageNet spec.
"""
num_dummy_categories = 20
num_dummy_img_per_category = 10
img_height, img_width = 84, 84
def __exit__(self, exc_type, exc_val, exc_tb):
GlobalHydra().clear()
def __exit__(self, exc_type, exc_val, exc_tb) -> None: # type: ignore
GlobalHydra().clear()
