def load_dict_data(dataset_path, is_linear, shape):
if is_linear:
train, test = load_data_vec(dataset_path)
else:
train, test = load_data_img(dataset_path, shape)
train['data'] = np.asarray(train['data'], dtype='float32')
test['data'] = np.asarray(test['data'] , dtype='float32')
train['data'] = train['data'] / 255
test['data'] = test['data'] / 255
return train, test
sys.exit(1)
dataset_path = sys.argv[1]
measure_type = sys.argv[2]
output_folder = sys.argv[3]
if not output_folder.endswith('/'):
output_folder += '/'
if not os.path.exists(output_folder):
os.mkdir(output_folder)
if not measure_type in ['cos', 'euc']:
sys.stderr.write('Error: measure_type must be in [cos, euc]\n')
sys.exit(1)
train, test = load_data_vec(dataset_path)
assert train['data'].shape[1] == test['data'].shape[1]
# train['data'] = normalize_data(train['data'])
# test['data'] = normalize_data(test['data'])
train['data'] = np.asarray(train['data'], dtype='float64')
test['data'] = np.asarray(test['data'], dtype='float64')
print train['data'].shape
print test['data'].shape
sample_num, dim = test['data'].shape
batch_len = 500
batch_num = sample_num / batch_len
engine = Engine(train, test, measure_type)
# engine = FileNameEngine(train, test, measure_type)