def test_build_config():
"""
Test build_config and check log_dir setting and checkpoint path verification
"""
config_path = "deepreg/config/unpaired_labeled_ddf.yaml"
log_dir = "test_build_config"
# checkpoint path empty
got_config, got_log_dir = build_config(config_path=config_path,
log_dir=log_dir,
ckpt_path="")
assert isinstance(got_config, dict)
assert got_log_dir == os.path.join("logs", log_dir)
# checkpoint path ends with ckpt
got_config, got_log_dir = build_config(config_path=config_path,
log_dir=log_dir,
ckpt_path="example.ckpt")
assert isinstance(got_config, dict)
assert got_log_dir == os.path.join("logs", log_dir)
# checkpoint path ends with h5
with pytest.raises(ValueError):
build_config(config_path=config_path,
log_dir=log_dir,
ckpt_path="example.h5")
def test_ckpt_path_err(self):
# checkpoint path ends with h5
with pytest.raises(ValueError) as err_info:
build_config(
config_path=self.config_path,
log_root=log_root,
log_dir=self.log_dir,
ckpt_path="example.h5",
)
assert "checkpoint path should end with .ckpt" in str(err_info.value)
def test_build_dataset():
"""
Test build_dataset by checking the output types
"""
# init arguments
config_path = "deepreg/config/unpaired_labeled_ddf.yaml"
log_dir = "test_build_dataset"
ckpt_path = ""
# load config
config, log_dir = build_config(
config_path=config_path, log_dir=log_dir, ckpt_path=ckpt_path
)
# build dataset
data_out_train, data_out_val = build_dataset(
dataset_config=config["dataset"],
preprocess_config=config["train"]["preprocess"],
)
data_loader_train, dataset_train, steps_per_epoch_train = data_out_train
data_loader_val, dataset_val, steps_per_epoch_val = data_out_val
# check output types
assert isinstance(data_loader_train, DataLoader)
assert isinstance(dataset_train, tf.data.Dataset)
assert isinstance(steps_per_epoch_train, int)
assert isinstance(data_loader_val, DataLoader)
assert isinstance(dataset_val, tf.data.Dataset)
assert isinstance(steps_per_epoch_val, int)
def test_max_epochs(self, max_epochs, expected_epochs, expected_save_period):
got_config, _ = build_config(
config_path=self.config_path,
log_root=log_root,
log_dir=self.log_dir,
ckpt_path="",
max_epochs=max_epochs,
)
assert got_config["train"]["epochs"] == expected_epochs
assert got_config["train"]["save_period"] == expected_save_period
def test_ckpt_path(self, ckpt_path):
# check the code can pass
got_config, got_log_dir = build_config(
config_path=self.config_path,
log_root=log_root,
log_dir=self.log_dir,
ckpt_path=ckpt_path,
)
assert isinstance(got_config, dict)
assert got_log_dir == os.path.join(log_root, self.log_dir)
def test_ckpt_path(self, ckpt_path):
# check the code can pass
got_config, got_log_dir, _ = build_config(
config_path=self.config_path,
log_dir=self.log_dir,
exp_name=self.exp_name,
ckpt_path=ckpt_path,
)
assert isinstance(got_config, dict)
assert got_log_dir == os.path.join(self.log_dir, self.exp_name)
def test_build_dataset():
"""
Test build_dataset by checking the output types
"""
# init arguments
config_path = "config/unpaired_labeled_ddf.yaml"
log_dir = "logs"
exp_name = "test_build_dataset"
ckpt_path = ""
# load config
config, _, _ = build_config(config_path=config_path,
log_dir=log_dir,
exp_name=exp_name,
ckpt_path=ckpt_path)
# build dataset
data_loader_train, dataset_train, steps_per_epoch_train = build_dataset(
dataset_config=config["dataset"],
preprocess_config=config["train"]["preprocess"],
split="train",
training=False,
repeat=False,
)
# check output types
assert isinstance(data_loader_train, DataLoader)
assert isinstance(dataset_train, tf.data.Dataset)
assert isinstance(steps_per_epoch_train, int)
# remove valid data
config["dataset"]["valid"]["dir"] = ""
# build dataset
data_loader_valid, dataset_valid, steps_per_epoch_valid = build_dataset(
dataset_config=config["dataset"],
preprocess_config=config["train"]["preprocess"],
split="valid",
training=False,
repeat=False,
)
assert data_loader_valid is None
assert dataset_valid is None
assert steps_per_epoch_valid is None