several commit in this PR (hunkim#80)

* modify code for compatability with keras 2.x.

* change requirement Keras 1.2.2 to 2.0.2.

* fix Travis CI Build Error(First try)

* Fix Travis CI Build Error(Second try).

* Fix Travis CI Build Error(Third try).

* Fix Travis CI Build Error(Forth try).

.travis.yml

@@ -1,12 +1,15 @@
 # code below is taken from https://github.com/fchollet/keras/blob/master/.travis.yml
 sudo: required
 dist: trusty
+env:
+  - LD_PRELOAD="/usr/lib/libtcmalloc.so.4"
 language: python
 python: # Only two versions for now
   - "2.7"
   - "3.5"
 # command to install dependencies
-# command to install dependencies
+before_install:
+  - sudo apt-get install google-perftools
 install: "pip install -r requirements.txt"
 
 script:
@@ -17,7 +20,7 @@ script:
   - sed -i -- 's/nb_epoch = 12/nb_epoch = 1/g' *lab-*.py # change range to 1 for quick testing
   - sed -i -- 's/plot(/# plot (/g' *lab-*.py # change range to 1 for quick testing
   - sed -i -- 's/from keras.utils.visualize_util import plot/#/g' *lab-*.py
-  - python lab-07-2-learning_rate_and_evaluation.py # run this first to download the MNIST file
+  - python lab-07-4-mnist_introduction.py # run this first to download the MNIST file
   # run all python files in parallel, http://stackoverflow.com/questions/5015316
   # The log length has exceeded the limit of 4 MB (this usually means that the test suite is raising the same exception over and over).
   #- ls lab-*.py|xargs -n 1 -P 1 python > /dev/null

Keras/klab-02-1-linear_regression.py

@@ -13,7 +13,7 @@
 # prints summary of the model to the terminal
 model.summary()
 
-model.fit(x_train, y_train, nb_epoch=1000)
+model.fit(x_train, y_train, epochs=1000)
 
 y_predict = model.predict(np.array([4]))
 print(y_predict)

Keras/klab-04-1-multi_input_linear_regression.py

@@ -7,11 +7,11 @@
 y_data = [[152.], [185.], [180.], [196.], [142.]]
 
 model = Sequential()
-model.add(Dense(output_dim=1, input_dim=3))
+model.add(Dense(input_dim=3, units=1))
 model.add(Activation('linear'))
 
-model.compile(loss='mse', optimizer='rmsprop',  lr=1e-10)
-model.fit(x_data, y_data, nb_epoch=1000)
+model.compile(loss='mse', optimizer='rmsprop', lr=1e-10)
+model.fit(x_data, y_data, epochs=1000)
 
 y_predict = model.predict(np.array([[95., 100., 80]]))
 print(y_predict)

Keras/klab-04-2-multi_input_linear_regression.py

@@ -7,7 +7,7 @@
 y_data = [[152.], [185.], [180.], [196.], [142.]]
 
 model = Sequential()
-model.add(Dense(input_dim=3, output_dim=1))
+model.add(Dense(input_dim=3, units=1))
 
 model.compile(loss='mse', optimizer='rmsprop')
 model.fit(x_data, y_data)

Keras/klab-04-3-file_input_linear_regression.py

@@ -9,7 +9,7 @@
 print("y_data", y_data)
 
 model = Sequential()
-model.add(Dense(input_dim=3, output_dim=1))
+model.add(Dense(input_dim=3, units=1))
 
 model.compile(loss='mse', optimizer='rmsprop')
 model.fit(x_data, y_data, nb_epoch=2000)

Keras/klab-04-4-stock_linear_regression.py

@@ -3,6 +3,8 @@
 import numpy as np
 from sklearn.preprocessing import MinMaxScaler
 
+import matplotlib.pyplot as plt
+
 xy = np.loadtxt('data-02-stock_daily.csv', delimiter=',')
 
 # very important. It does not work without it.
@@ -24,10 +26,10 @@
 print(x_data[0], "->", y_data[0])
 
 model = Sequential()
-model.add(Dense(input_dim=4, output_dim=1))
+model.add(Dense(input_dim=4, units=1))
 
 model.compile(loss='mse', optimizer='sgd', metrics=['mse'])
-model.fit(x_data, y_data, nb_epoch=100)
+model.fit(x_data, y_data, epochs=100)
 
 test = x_data[10].reshape(-1, 4)
 print("y=", y_data[10], "prediction=", model.predict(test))
@@ -38,19 +40,18 @@
 # ---------------------------
 # Test
 # split to train and testing
-import matplotlib.pyplot as plt
 
 train_size = int(len(x_data) * 0.7)
 test_size = len(x_data) - train_size
 x_train, x_test = x_data[0:train_size], x_data[train_size:len(x_data)]
 y_train, y_test = y_data[0:train_size], y_data[train_size:len(y_data)]
 
 model = Sequential()
-model.add(Dense(input_dim=4, output_dim=1))
+model.add(Dense(input_dim=4, units=1))
 model.compile(loss='mse', optimizer='sgd', metrics=['mse'])
 
 # Train a model
-model.fit(x_train, y_train, nb_epoch=200)
+model.fit(x_train, y_train, epochs=200)
 
 # evaluate
 results = model.evaluate(x_test, y_test, verbose=1)

Keras/klab-05-1-logistic_regression.py

@@ -11,7 +11,7 @@
 model.compile(loss='binary_crossentropy', optimizer='sgd', lr=0.1)
 
 model.summary()
-model.fit(x_data, y_data, nb_epoch=2000)
+model.fit(x_data, y_data, epochs=2000)
 
 print("2,1", model.predict_classes(np.array([[2, 1]])))
 print("6,5", model.predict_classes(np.array([[6, 5]])))

Keras/klab-06-1-softmax.py

@@ -21,7 +21,7 @@
               optimizer='sgd',
               metrics=['accuracy'])
 
-history = model.fit(x_data, y_data, nb_epoch=1000)
+history = model.fit(x_data, y_data, epochs=1000)
 
 print(model.predict_classes(np.array([[1, 2, 1, 1]])))
 print(model.predict_classes(np.array([[1, 2, 5, 6]])))

Keras/klab-06-2-softmax_zoo.py

@@ -23,7 +23,7 @@
               optimizer='sgd',
               metrics=['accuracy'])
 
-history = model.fit(x_data, y_one_hot, nb_epoch=1000)
+history = model.fit(x_data, y_one_hot, epochs=1000)
 
 # Let's see if we can predict
 pred = model.predict_classes(x_data)

Keras/klab-09-1-xor.py

@@ -9,7 +9,7 @@
 model.compile(loss='binary_crossentropy', optimizer='sgd',
               lr=0.1, metrics=['accuracy'])
 model.summary()
-model.fit(x_data, y_data, nb_epoch=50000)
+model.fit(x_data, y_data, epochs=50000)
 
 score = model.evaluate(x_data, y_data, verbose=0)
 print('Test score:', score[0])

Keras/klab-09-2-xor-nn.py

@@ -1,6 +1,5 @@
 from keras.models import Sequential
 from keras.layers import Dense
-from keras.optimizers import SGD
 
 x_data = [[0., 0.], [0., 1.], [1., 0.], [1., 1.]]
 y_data = [[0.], [1.], [1.], [0.]]

Keras/klab-11-1-cnn_mnist.py

@@ -7,7 +7,6 @@
 
 from __future__ import print_function
 import numpy as np
-np.random.seed(1337)  # for reproducibility
 
 from keras.datasets import mnist
 from keras.models import Sequential
@@ -16,6 +15,8 @@
 from keras.utils import np_utils
 from keras import backend as K
 
+np.random.seed(1337)  # for reproducibility
+
 batch_size = 128
 nb_classes = 10
 nb_epoch = 12

Keras/klab-12-1-rnn_hello_char.py

@@ -7,8 +7,8 @@
 
 # brew install graphviz
 # pip3 install graphviz
-# pip3 install pydot
-from keras.utils.visualize_util import plot
+# pip3 install pydot-ng
+from keras.utils.vis_utils import plot_model
 
 # sample text
 sample = "hihello"
@@ -21,37 +21,38 @@
 
 data_dim = len(char_set)
 timesteps = len(y_str)
-nb_classes = len(char_set)
+num_classes = len(char_set)
 
 print(x_str, y_str)
 
 x = [char_dic[c] for c in x_str]  # char to index
 y = [char_dic[c] for c in y_str]  # char to index
 
 # One-hot encoding
-x = np_utils.to_categorical(x, nb_classes=nb_classes)
+x = np_utils.to_categorical(x, num_classes=num_classes)
 # reshape X to be [samples, time steps, features]
 x = np.reshape(x, (-1, len(x), data_dim))
 print(x.shape)
 
 # One-hot encoding
-y = np_utils.to_categorical(y, nb_classes=nb_classes)
+y = np_utils.to_categorical(y, num_classes=num_classes)
 # time steps
 y = np.reshape(y, (-1, len(y), data_dim))
 print(y.shape)
 
 model = Sequential()
-model.add(LSTM(nb_classes, input_shape=(
+model.add(LSTM(num_classes, input_shape=(
     timesteps, data_dim), return_sequences=True))
-model.add(TimeDistributed(Dense(nb_classes)))
+model.add(TimeDistributed(Dense(num_classes)))
 model.add(Activation('softmax'))
 model.summary()
 # Store model graph in png
-# plot(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
+# (Error occurs on in python interactive shell)
+plot_model(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
 
 model.compile(loss='categorical_crossentropy',
               optimizer='rmsprop', metrics=['accuracy'])
-model.fit(x, y, nb_epoch=1)
+model.fit(x, y, epochs=1)
 
 predictions = model.predict(x, verbose=0)
 for i, prediction in enumerate(predictions):

Keras/klab-12-1-softmax_hello_char.py

@@ -7,8 +7,8 @@
 
 # brew install graphviz
 # pip3 install graphviz
-# pip3 install pydot
-from keras.utils.visualize_util import plot
+# pip3 install pydot-ng
+from keras.utils.vis_utils import plot_model
 
 # sample text
 sample = "hihello"
@@ -21,36 +21,37 @@
 
 data_dim = len(char_set)
 timesteps = len(y_str)
-nb_classes = len(char_set)
+num_classes = len(char_set)
 
 print(x_str, y_str)
 
 x = [char_dic[c] for c in x_str]  # char to index
 y = [char_dic[c] for c in y_str]  # char to index
 
 # One-hot encoding
-x = np_utils.to_categorical(x, nb_classes=nb_classes)
+x = np_utils.to_categorical(x, num_classes=num_classes)
 # reshape X to be [samples, time steps, features]
 x = np.reshape(x, (-1, len(x), data_dim))
 print(x.shape)
 
 # One-hot encoding
-y = np_utils.to_categorical(y, nb_classes=nb_classes)
+y = np_utils.to_categorical(y, num_classes=num_classes)
 # time steps
 y = np.reshape(y, (-1, len(y), data_dim))
 print(y.shape)
 
 model = Sequential()
-model.add(Dense(nb_classes, input_shape=(
+model.add(Dense(num_classes, input_shape=(
     timesteps, data_dim)))
 model.add(Activation('softmax'))
 model.summary()
 # Store model graph in png
-# plot(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
+# (Error occurs on in python interactive shell)
+plot_model(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
 
 model.compile(loss='categorical_crossentropy',
               optimizer='rmsprop', metrics=['accuracy'])
-model.fit(x, y, nb_epoch=100)
+model.fit(x, y, epochs=100)
 
 predictions = model.predict(x, verbose=0)
 for i, prediction in enumerate(predictions):

Keras/klab-12-2-rnn_long_char.py

@@ -7,8 +7,8 @@
 
 # brew install graphviz
 # pip3 install graphviz
-# pip3 install pydot
-from keras.utils.visualize_util import plot
+# pip3 install pydot-ng
+from keras.utils.vis_utils import plot_model
 
 # sample sentence
 sentence = ("if you want to build a ship, don't drum up people together to "
@@ -20,7 +20,7 @@
 
 data_dim = len(char_set)
 seq_length = timesteps = 10
-nb_classes = len(char_set)
+num_classes = len(char_set)
 
 dataX = []
 dataY = []
@@ -36,32 +36,33 @@
     dataY.append(y)
 
 # One-hot encoding
-dataX = np_utils.to_categorical(dataX, nb_classes=nb_classes)
+dataX = np_utils.to_categorical(dataX, num_classes=num_classes)
 # reshape X to be [samples, time steps, features]
 dataX = np.reshape(dataX, (-1, seq_length, data_dim))
 print(dataX.shape)
 
 # One-hot encoding
-dataY = np_utils.to_categorical(dataY, nb_classes=nb_classes)
+dataY = np_utils.to_categorical(dataY, num_classes=num_classes)
 # time steps
 dataY = np.reshape(dataY, (-1, seq_length, data_dim))
 print(dataY.shape)
 
 model = Sequential()
-model.add(LSTM(nb_classes, input_shape=(
+model.add(LSTM(num_classes, input_shape=(
     timesteps, data_dim), return_sequences=True))
-model.add(LSTM(nb_classes, return_sequences=True))
-model.add(TimeDistributed(Dense(nb_classes)))
+model.add(LSTM(num_classes, return_sequences=True))
+model.add(TimeDistributed(Dense(num_classes)))
 
 model.add(Activation('softmax'))
 model.summary()
 
 # Store model graph in png
-# plot(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
+# (Error occurs on in python interactive shell)
+plot_model(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
 
 model.compile(loss='categorical_crossentropy',
               optimizer='rmsprop', metrics=['accuracy'])
-model.fit(dataX, dataY, nb_epoch=1000)
+model.fit(dataX, dataY, epochs=1000)
 
 predictions = model.predict(dataX, verbose=0)
 for i, prediction in enumerate(predictions):

Keras/klab-12-3-rnn_stock_prediction.py

@@ -7,8 +7,8 @@
 
 # brew install graphviz
 # pip3 install graphviz
-# pip3 install pydot
-from keras.utils.visualize_util import plot
+# pip3 install pydot-ng
+from keras.utils.vis_utils import plot_model
 
 import matplotlib.pyplot as plt
 
@@ -52,10 +52,11 @@
 model.summary()
 
 # Store model graph in png
-# plot(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
+# (Error occurs on in python interactive shell)
+plot_model(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
 
 print(trainX.shape, trainY.shape)
-model.fit(trainX, trainY, nb_epoch=200)
+model.fit(trainX, trainY, epochs=200)
 
 # make predictions
 testPredict = model.predict(testX)

Keras/klab-12-4-rnn_deep_prediction.py

@@ -8,8 +8,8 @@
 
 # brew install graphviz
 # pip3 install graphviz
-# pip3 install pydot
-from keras.utils.visualize_util import plot
+# pip3 install pydot-ng
+from keras.utils.vis_utils import plot_model
 
 import matplotlib.pyplot as plt
 
@@ -54,12 +54,13 @@
 model.summary()
 
 # Store model graph in png
-# plot(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
+# (Error occurs on in python interactive shell)
+plot_model(model, to_file=os.path.basename(__file__) + '.png', show_shapes=True)
 
 model.compile(loss='mean_squared_error', optimizer='adam')
 
 print(trainX.shape, trainY.shape)
-model.fit(trainX, trainY, nb_epoch=200)
+model.fit(trainX, trainY, epochs=200)
 
 # make predictions
 testPredict = model.predict(testX)