lunny/keras
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix autoencoder serialization

Browse Source
This commit is contained in:
fchollet 2015-08-18 16:49:13 -07:00
parent 5e9579aeac
commit 78feed7fa9
2 changed files with 15 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
keras/utils/layer_utils.py
View File

@ -5,7 +5,7 @@ import theano
import copy import copy
from ..layers.advanced_activations import LeakyReLU, PReLU from ..layers.advanced_activations import LeakyReLU, PReLU
from ..layers.core import Dense, Merge, Dropout, Activation, Reshape, Flatten, RepeatVector, Layer from ..layers.core import Dense, Merge, Dropout, Activation, Reshape, Flatten, RepeatVector, Layer, AutoEncoder
from ..layers.core import ActivityRegularization, TimeDistributedDense, AutoEncoder, MaxoutDense from ..layers.core import ActivityRegularization, TimeDistributedDense, AutoEncoder, MaxoutDense
from ..layers.convolutional import Convolution1D, Convolution2D, MaxPooling1D, MaxPooling2D, ZeroPadding2D from ..layers.convolutional import Convolution1D, Convolution2D, MaxPooling1D, MaxPooling2D, ZeroPadding2D
from ..layers.embeddings import Embedding, WordContextProduct from ..layers.embeddings import Embedding, WordContextProduct
@ -58,6 +58,14 @@ def container_from_config(original_layer_dict):
graph_layer.add_output(**output) graph_layer.add_output(**output)
return graph_layer return graph_layer
elif name == 'AutoEncoder':
kwargs = {'encoder': container_from_config(layer_dict.get('encoder_config')),
'decoder': container_from_config(layer_dict.get('decoder_config'))}
for kwarg in ['output_reconstruction', 'weights']:
if kwarg in layer_dict:
kwargs[kwarg] = layer_dict[kwarg]
return AutoEncoder(**kwargs)
else: else:
layer_dict.pop('name') layer_dict.pop('name')

8
tests/manual/check_autoencoder.py
View File

@ -1,7 +1,7 @@
from __future__ import absolute_import from __future__ import absolute_import
from __future__ import print_function from __future__ import print_function
from keras.datasets import mnist from keras.datasets import mnist
from keras.models import Sequential from keras.models import Sequential, model_from_config
from keras.layers.core import AutoEncoder, Dense, Activation, TimeDistributedDense, Flatten from keras.layers.core import AutoEncoder, Dense, Activation, TimeDistributedDense, Flatten
from keras.layers.recurrent import LSTM from keras.layers.recurrent import LSTM
from keras.layers.embeddings import Embedding from keras.layers.embeddings import Embedding
@ -57,6 +57,7 @@ print('\nclassical_score:', classical_score)
# autoencoder model test # # autoencoder model test #
########################## ##########################
def build_lstm_autoencoder(autoencoder, X_train, X_test): def build_lstm_autoencoder(autoencoder, X_train, X_test):
X_train = X_train[:, np.newaxis, :] X_train = X_train[:, np.newaxis, :]
X_test = X_test[:, np.newaxis, :] X_test = X_test[:, np.newaxis, :]
@ -95,7 +96,6 @@ for autoencoder_type in ['classical', 'lstm']:
print("Error: unknown autoencoder type!") print("Error: unknown autoencoder type!")
exit(-1) exit(-1)
autoencoder.get_config(verbose=1)
autoencoder.compile(loss='mean_squared_error', optimizer='adam') autoencoder.compile(loss='mean_squared_error', optimizer='adam')
# Do NOT use validation data with return output_reconstruction=True # Do NOT use validation data with return output_reconstruction=True
autoencoder.fit(X_train, X_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=1) autoencoder.fit(X_train, X_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=1)
@ -128,3 +128,7 @@ for autoencoder_type in ['classical', 'lstm']:
print('Loss change:', (score[0] - classical_score[0])/classical_score[0], '%') print('Loss change:', (score[0] - classical_score[0])/classical_score[0], '%')
print('Accuracy change:', (score[1] - classical_score[1])/classical_score[1], '%') print('Accuracy change:', (score[1] - classical_score[1])/classical_score[1], '%')
# check serialization
config = autoencoder.get_config(verbose=1)
autoencoder = model_from_config(config)