def train_model(self, x_train, y_train, x_val, y_val):
self.model.compile(loss='binary_crossentropy',
optimizer='adam',
metrics=['accuracy'])
self.model.summary()
# File used to save model checkpoints
model_filename = 'chargram-cnn.{0:03d}.hdf5'
if self.last_finished_epoch > 0:
self.model = load_model(
model_filename.format(self.last_finished_epoch - 1))
elif self.initial_weights_file is not None:
self.model = load_model(self.initial_weights_file)
print('Fitting model')
self.model.fit(x_train,
y_train,
batch_size=self.batch_size,
epochs=self.epochs,
validation_data=(x_val, y_val),
callbacks=[
keras_utils.ModelSaveCallback(model_filename),
keras_utils.TqdmProgressCallback()
],
verbose=1,
initial_epoch=self.last_finished_epoch or 0)
#### fill `last_finished_epoch` with your latest finished epoch
# from keras.models import load_model
# s = reset_tf_session()
# last_finished_epoch = 7
# model = load_model(model_filename.format(last_finished_epoch))
# fit model
model.fit(
x_train2,
y_train2, # prepared data
batch_size=BATCH_SIZE,
epochs=EPOCHS,
callbacks=[
keras.callbacks.LearningRateScheduler(lr_scheduler),
LrHistory(),
keras_utils.TqdmProgressCallback(),
keras_utils.ModelSaveCallback(model_filename)
],
validation_data=(x_test2, y_test2),
shuffle=True,
verbose=0,
initial_epoch=last_finished_epoch or 0)
# save weights to file
model.save_weights("weights.h5")
# load weights from file (can call without model.fit)
model.load_weights("weights.h5")
"""# Evaluate model"""
# make test predictions
"""Meld them together into one model:"""
s = reset_tf_session()
encoder, decoder = build_pca_autoencoder(IMG_SHAPE, code_size=32)
inp = L.Input(IMG_SHAPE)
code = encoder(inp)
reconstruction = decoder(code)
autoencoder = keras.models.Model(inputs=inp, outputs=reconstruction)
autoencoder.compile(optimizer='adamax', loss='mse')
autoencoder.fit(x=X_train, y=X_train, epochs=15,
validation_data=[X_test, X_test],
callbacks=[keras_utils.TqdmProgressCallback()],
verbose=0)
def visualize(img,encoder,decoder):
"""Draws original, encoded and decoded images"""
code = encoder.predict(img[None])[0] # img[None] is the same as img[np.newaxis, :]
reco = decoder.predict(code[None])[0]
plt.subplot(1,3,1)
plt.title("Original")
show_image(img)
plt.subplot(1,3,2)
plt.title("Code")
plt.imshow(code.reshape([code.shape[-1]//2,-1]))
