loss=tf.nn.softmax_cross_entropy_with_logits,
run_eagerly=True)
'''
MODEL 5 Input --> Dense Layer --> Projection --> Output
'''
model = models.Sequential()
model.add(layers.Dense(32, input_shape=(300, )))
model.add(layer=ProjectVectorLayer(dimension=32))
model.compile(optimizer='SGD',
loss=tf.nn.softmax_cross_entropy_with_logits,
run_eagerly=True)
'''
TRAINING THE MODEL
'''
'''
# Checkpoint to save weights at lowest validation loss
checkpoint_filepath = '/tmp/checkpoint'
model_checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(
filepath=checkpoint_filepath,
save_weights_only=True,
monitor='val_loss',
mode='min',
save_best_only=True)
early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=300)
num_epochs = 250
history = model.fit(x_train, y_train, epochs=num_epochs, batch_size=32, validation_data=(x_val, y_val), callbacks=[model_checkpoint_callback])
# print model structure diagram
print(classifier.summary())
# Compiling the CNN
classifier.compile(adam(lr=.0001),
loss='categorical_crossentropy',
metrics=['accuracy'])
cp_callback = create_callback(checkpoint_path, 1)
hist = classifier.fit_generator(
train_generator,
epochs=num_epoch,
steps_per_epoch=num_train_samples // batch_size,
validation_data=validation_generator,
validation_steps=num_validate_samples // batch_size,
class_weight='auto',
callbacks=[cp_callback])
# Plot training and validation accuracy
plot_accuracy(hist)
# Plot training and validation loss
plot_loss(hist)
# Print test set accuracy and loss values
scores = classifier.evaluate_generator(test_generator,
num_test_samples / batch_size)
print("loss: {}, accuracy: {}".format(scores[0], scores[1]))
num_epochs = 1000
history = model.fit(x_train,
y_train,
epochs=num_epochs,
batch_size=batch_size,
validation_data=(x_val, y_val),
callbacks=[model_checkpoint_callback])
# Loading the Best Weights
model.load_weights(checkpoint_filepath)
# Save this Model
model.save_weights('verb_prediction_from_dependent_model')
# Training History
history_dict = history.history
loss_values = history_dict['loss']
val_loss_values = history_dict['val_loss']
num_epochs = len(val_loss_values)
# Plot the loss
fun.plot_loss(num_epochs, loss_values, val_loss_values)
# Calculate Perplexity
model.load_weights('verb_prediction_from_dependent_model') # Load the model
perplexity = fun.calculate_perplexity(x_val, y_val, model)
print("Perplexity: " + str(perplexity))
# print model structure diagram
print(model.summary())
# Transfer Learning
print("\nPerforming Transfer Learning")
# Compiling the model
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
# Fit the Transfer Learning model to the data from the generators
history = model.fit_generator(train_generator,
epochs=num_epoch,
steps_per_epoch=num_train_samples // batch_size,
validation_data=validation_generator,
validation_steps=num_validate_samples //
batch_size,
class_weight='auto',
shuffle=True)
# Plot training and validation accuracy
plot_accuracy(history)
# Plot training and validation loss
plot_loss(history)
# Print test set accuracy and loss values
scores = model.evaluate_generator(test_generator,
num_test_samples / batch_size)
print("loss: {}, accuracy: {}".format(scores[0], scores[1]))