def setup_mocker(self, mocker):
mocker.patch.object(ConfigParser,
"save_dir",
return_value=dict2obj({"mkdir": lambda a: a}))
mocker.patch.object(ConfigParser,
"log_dir",
return_value=dict2obj({"mkdir": lambda a: a}))
mocker.patch.object(parse_config, "write_json")
mocker.patch.object(parse_config, "setup_logging")
def test_from_args_load_resume_config(self, mocker):
args = self.args
self.setup_mocker(mocker)
with mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device": None,
"resume": "tests/config.json",
"config": None
})):
config = ConfigParser.from_args(args)
assert config.config['name'] == 'Mnist_LeNet_Resume'
args.parse_args.assert_called_once()
with mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device":
None,
"resume":
"tests/config.json",
"config":
"./config.json"
})):
config = ConfigParser.from_args(args)
# If both resume and config parameter set, the config will override the configs in resume one.
assert config.config['name'] == 'Mnist_LeNet'
args.parse_args.assert_called_once()
with mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device":
"1,2",
"resume":
None,
"config":
"tests/config.json"
})):
config = ConfigParser.from_args(args)
# If both resume and config parameter set, the config will override the configs in resume one.
assert config.config['name'] == 'Mnist_LeNet_Resume'
assert os.environ["CUDA_VISIBLE_DEVICES"] == "1,2"
args.parse_args.assert_called_once()
assert ConfigParser.save_dir.mkdir.call_count == 3
assert ConfigParser.log_dir.mkdir.call_count == 3
def setup_class(self, ):
args = {}
def parse_args():
return {}
args['parse_args'] = parse_args
self.args = dict2obj(args)
def test_from_args_empty_config(self, mocker):
with pytest.raises(AssertionError) as excinfo:
args = self.args
mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device": None,
"resume": None,
"config": None
}))
config = ConfigParser.from_args(args)
assert "Configuration file need to be specified. Add '-c config.json', for example." in str(
excinfo.value)
def test_init_transform(self, mocker):
args = self.args
self.setup_mocker(mocker)
with mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device": None,
"resume": "tests/config.json",
"config": None
})):
config = ConfigParser.from_args(args)
mytransfm = config.init_transform()
assert len(mytransfm.transforms) == 5
def test_init_obj_without_args(self, mocker):
args = self.args
self.setup_mocker(mocker)
with mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device": None,
"resume": "tests/config.json",
"config": None
})):
config = ConfigParser.from_args(args)
config.config['tests'] = {}
config.config['tests']['type'] = "FakeClass"
obj = config.init_obj('tests')
assert obj.test_func()
assert obj.test_func_with_param1() is None
def main(job_dir, data_dir, num_gpus, use_distortion_for_training,
log_device_placement, num_intra_threads, **hparams):
# Session configuration.
sess_config = tf.compat.v1.ConfigProto(
allow_soft_placement=True,
log_device_placement=log_device_placement,
intra_op_parallelism_threads=num_intra_threads,
gpu_options=tf.compat.v1.GPUOptions(force_gpu_compatible=True))
config = utils.RunConfig(session_config=sess_config, model_dir=job_dir)
if hparams['eval']:
config = config.replace(save_checkpoints_steps=hparams['eval_freq'])
train_input_fn = functools.partial(
input_fn,
data_dir,
subset='train',
batch_size=hparams['train_batch_size'],
dataset=hparams['dataset'],
use_distortion_for_training=use_distortion_for_training)
eval_input_fn = functools.partial(input_fn,
data_dir,
subset='test',
dataset=hparams['dataset'],
batch_size=hparams['eval_batch_size'])
train_steps = hparams['train_steps']
eval_steps = 2000 // hparams['eval_batch_size']
train_spec = tf.estimator.TrainSpec(train_input_fn, max_steps=train_steps)
eval_spec = tf.estimator.EvalSpec(input_fn=eval_input_fn,
steps=eval_steps,
start_delay_secs=0,
throttle_secs=0)
classifier = tf.estimator.Estimator(model_fn=get_model_fn,
config=config,
params=utils.dict2obj(**hparams))
# Create experiment.
tf.estimator.train_and_evaluate(estimator=classifier,
train_spec=train_spec,
eval_spec=eval_spec)
def test_init_ftn_with_arg(self, mocker):
args = self.args
self.setup_mocker(mocker)
with mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device": None,
"resume": "tests/config.json",
"config": None
})):
config = ConfigParser.from_args(args)
config.config['tests'] = ["fake_func1", "fake_func2"]
ftns = config.init_ftn("tests", param1="param1")
for ftn in ftns:
assert ftn() == "param1"
def test_init_obj_with_lib(self, mocker):
args = self.args
self.setup_mocker(mocker)
mocker.patch.object(torch.optim, "SGD")
with mocker.patch.object(args,
"parse_args",
return_value=dict2obj({
"device": None,
"resume": "tests/config.json",
"config": None
})):
config = ConfigParser.from_args(args)
config.config['optimizer'] = {}
config.config['optimizer']['type'] = "SGD"
obj = config.init_obj('optimizer', torch.optim, {})
obj()
assert obj.call_count == 1