def get_raw_params(self,
partition_entity='memberId',
num_of_lbfgs_iterations=None):
base_training_params = setup_fake_base_training_params(
training_stage=constants.RANDOM_EFFECT)
base_training_params.batch_size = 2
# flatten the params
raw_params = list(base_training_params.__to_argv__())
model_params = setup_fake_raw_model_params(
training_stage=constants.RANDOM_EFFECT)
raw_params.extend(model_params)
raw_params.extend(['--' + constants.MODEL_IDS_DIR, test_dataset_path])
raw_params.extend([
'--' + constants.FEATURE_FILE,
os.path.join(test_dataset_path, fake_feature_file)
])
raw_params.extend(
['--' + constants.PARTITION_ENTITY, partition_entity])
raw_params.extend(['--' + constants.LABEL, 'response'])
raw_params.extend(['--' + constants.L2_REG_WEIGHT, '0.1'])
if num_of_lbfgs_iterations:
raw_params.extend([
'--' + constants.NUM_OF_LBFGS_ITERATIONS,
f'{num_of_lbfgs_iterations}'
])
return base_training_params, raw_params
def get_raw_params(self,
partition_entity='memberId',
num_of_lbfgs_iterations=None,
intercept_only=False):
base_training_params = setup_fake_base_training_params(
training_stage=constants.RANDOM_EFFECT)
base_training_params.batch_size = 2
# flatten the params
raw_params = list(base_training_params.__to_argv__())
model_params = setup_fake_raw_model_params(
training_stage=constants.RANDOM_EFFECT)
raw_params.extend(model_params)
raw_params.extend(['--' + constants.MODEL_IDS_DIR, test_dataset_path])
raw_params.extend([
'--' + constants.FEATURE_FILE,
os.path.join(test_dataset_path, fake_feature_file)
])
raw_params.extend(
['--' + constants.PARTITION_ENTITY, partition_entity])
raw_params.extend(['--' + constants.LABEL_COLUMN_NAME, 'response'])
raw_params.extend(['--' + constants.L2_REG_WEIGHT, '0.1'])
if num_of_lbfgs_iterations:
raw_params.extend([
'--' + constants.NUM_OF_LBFGS_ITERATIONS,
f'{num_of_lbfgs_iterations}'
])
if intercept_only:
feature_bag_index = raw_params.index(f'--{constants.FEATURE_BAG}')
raw_params.pop(feature_bag_index)
raw_params.pop(feature_bag_index)
assert (f'--{constants.FEATURE_BAG}' not in raw_params)
assert ('per_member' not in raw_params)
return base_training_params, raw_params
def setUp(self):
self.task_type = "worker"
self.worker_index = 0
self.num_workers = 5
set_fake_tf_config(task_type=self.task_type,
worker_index=self.worker_index)
self.params = setup_fake_base_training_params()
self.model_params = setup_fake_raw_model_params()
def get_raw_params(self, partition_entity='memberId'):
base_training_params = setup_fake_base_training_params(
training_stage=constants.RANDOM_EFFECT)
base_training_params[constants.BATCH_SIZE] = 2
# flatten the params
raw_params = [
x for key in base_training_params
for x in ['--' + key, str(base_training_params[key])]
]
model_params = setup_fake_raw_model_params(
training_stage=constants.RANDOM_EFFECT)
raw_params.extend(model_params)
raw_params.extend(['--' + constants.MODEL_IDS_DIR, test_dataset_path])
raw_params.extend([
'--' + constants.FEATURE_FILE,
os.path.join(test_dataset_path, fake_feature_file)
])
raw_params.extend(
['--' + constants.PARTITION_ENTITY, partition_entity])
raw_params.extend(['--' + constants.LABEL, 'response'])
raw_params.extend(['--' + constants.L2_REG_WEIGHT, '0.1'])
return base_training_params, raw_params
def setUp(self):
self.params = setup_fake_base_training_params()
self.model_params = setup_fake_raw_model_params()
def setUp(self):
self.model_params = setup_fake_raw_model_params()