hyperasによるハイパーパラメーター探索時のエラー

前提・実現したいこと

kerasでCNNを構築して、ハイパーパラメータをhyperasを用いて探索していようとしています。
ちなみに学習、検証データはデータ拡張をしたいので、ImageDataGenerator.flow_from_directoryを使ってデータを持ってきています。

発生している問題・エラーメッセージ

>>> Imports:
#coding=utf-8

try:
    import os
except:
    pass

try:
    import keras
except:
    pass

try:
    from keras import layers
except:
    pass

try:
    from keras import models
except:
    pass

try:
    from keras import optimizers
except:
    pass

try:
    from keras.preprocessing.image import ImageDataGenerator
except:
    pass

try:
    from keras.utils import plot_model
except:
    pass

try:
    from hyperopt import Trials, STATUS_OK, tpe
except:
    pass

try:
    from hyperas import optim
except:
    pass

try:
    from hyperas.distributions import choice, uniform
except:
    pass

Traceback (most recent call last):
  File "model.py", line 78, in <module>
    trials=Trials())
  File "/home/yukawa/.pyenv/versions/anaconda3-5.2.0/lib/python3.6/site-packages/hyperas/optim.py", line 69, in minimize
    keep_temp=keep_temp)
  File "/home/yukawa/.pyenv/versions/anaconda3-5.2.0/lib/python3.6/site-packages/hyperas/optim.py", line 98, in base_minimizer
    model_str = get_hyperopt_model_string(model, data, functions, notebook_name, verbose, stack)
  File "/home/yukawa/.pyenv/versions/anaconda3-5.2.0/lib/python3.6/site-packages/hyperas/optim.py", line 191, in get_hyperopt_model_string
    parts = hyperparameter_names(model_string)
  File "/home/yukawa/.pyenv/versions/anaconda3-5.2.0/lib/python3.6/site-packages/hyperas/optim.py", line 258, in hyperparameter_names
    parts.append(parts[-1])
IndexError: list index out of range

該当のソースコード

python
1import os
2import keras
3from keras import layers
4from keras import models
5from keras import optimizers
6from keras.preprocessing.image import ImageDataGenerator
7from keras.utils import plot_model
8from hyperopt import Trials, STATUS_OK, tpe
9from hyperas import optim
10from hyperas.distributions import choice, uniform
11def model(train_generator, valid_generator):
12	model = models.Sequential()
13	if {{choice(['normarize','no_normarize'])}} == 'normarize':
14		model.add(layers.Conv2D({{choice([32, 64, 128])}}, (3, 3), activation='relu', input_shape=(150, 150, 1)))
15		model.add(layers.normalization.BatchNormalization())
16		model.add(layers.MaxPooling2D((2, 2)))
17		model.add(layers.Conv2D({{choice([64, 128, 256])}}, (3, 3), activation='relu'))
18		model.add(layers.normalization.BatchNormalization())
19		model.add(layers.MaxPooling2D((2, 2)))
20		model.add(layers.Conv2D({{choice([64, 128, 256])}}, (3, 3), activation='relu'))
21		model.add(layers.normalization.BatchNormalization())
22		model.add(layers.Flatten())
23		model.add(layers.Dropout({{uniform(0,1)}}))
24		model.add(layers.Dense({{choice([50, 100, 150])}}, activation='relu'))
25		model.add(layers.normalization.BatchNormalization())
26		model.add(layers.Dense(1, activation='sigmoid'))
27	elif {{choice(['normarize','no_normarize'])}} == 'no_normarize':
28		model.add(layers.Conv2D({{choice([32, 64, 128])}}, (3, 3), activation='relu', input_shape=(150, 150, 1)))
29		model.add(layers.MaxPooling2D((2, 2)))
30		model.add(layers.Conv2D({{choice([64, 128, 256])}}, (3, 3), activation='relu'))
31		model.add(layers.MaxPooling2D((2, 2)))
32		model.add(layers.Conv2D({{choice([64, 128, 256])}}, (3, 3), activation='relu'))
33		model.add(layers.Flatten())
34		model.add(layers.Dropout({{uniform(0,1)}}))
35		model.add(layers.Dense({{choice([50, 100, 150])}}, activation='relu'))
36		model.add(layers.Dense(1, activation='sigmoid'))
37	model.compile(loss='binary_crossentropy',
38			  optimizer=optimizers.Adam(lr=1e-4),
39			  metrics=['acc'])
40	model.fit_generator(train_generator,
41					  steps_per_epoch=64,
42					  epochs=20,
43					  validation_data=valid_generator,
44					  validation_steps=64)
45	val_loss, val_acc = model.evaluate_generator(valid_generator, verbose=0)
46	return {'loss': -val_acc, 'status': STATUS_OK, 'model': model}
47def data():
48	train_dir = 'train'
49	valid_dir = 'valid'
50	train_datagen = ImageDataGenerator(rescale=1./255,
51									   rotation_range=40,
52									   width_shift_range=0.2,
53									   height_shift_range=0.2,
54									   shear_range=0.2,
55									   zoom_range=0.2,
56									   horizontal_flip=True)
57	valid_datagen = ImageDataGenerator(rescale=1./255)
58	train_generator = train_datagen.flow_from_directory(train_dir,
59														target_size=(150, 150),
60														color_mode='grayscale',
61														batch_size=64,
62														class_mode='binary')
63	valid_generator = valid_datagen.flow_from_directory(valid_dir,
64														target_size=(150, 150),
65														color_mode='grayscale',
66														batch_size=64,
67														class_mode='binary')
68	return train_generator, valid_generator
69if __name__ == '__main__':
70	output_dir = 'model_output'
71	if not os.path.exists(output_dir):
72		os.mkdir(output_dir)
73	train_generator, validation_generator = data()
74	best_run, best_model = optim.minimize(model=model,
75										  data=data,
76										  algo=tpe.suggest,
77										  max_evals=100,
78										  trials=Trials())
79	best_model.summary()
80	print(best_run)
81	val_loss, val_acc = best_model.evaluate_generator(valid_generator, verbose=0)
82	print('val_loss : {}'.format(val_loss))
83	print('val_acc : {}'.format(val_acc))
84	best_model.save(output_dir + '/cnn_model.h5')
85	plot_model(best_model, show_shapes=True, to_file=output_dir + '/model.png')
86