def train_model(x_train, x_valid, y_train, y_valid, model_name, args):
'''
Train the model
'''
model = build_model(args)
model.compile(loss='mean_squared_error',
optimizer=Adam(lr=args.learning_rate))
train_generator = Generator(x_train, y_train, True, args)
validation_generator = Generator(x_valid, y_valid, False, args)
pass
model.save(model_name)
def train_model(x_train, x_valid, y_train, y_valid, model_name, args):
'''
Train the model
'''
model = build_model(args)
pass
train_generator = Generator(x_train, y_train, True, args)
validation_generator = Generator(x_valid, y_valid, False, args)
model.fit_generator(train_generator, validation_data=\
validation_generator, epochs=\
args.nb_epoch, use_multiprocessing=\
False, max_queue_size=\
10, workers=\
4)
model.save(model_name)
def train_model(x_train, x_valid, y_train, y_valid, model_name, args):
'''
Train the model
'''
model = build_model(args)
model.compile(loss='mean_squared_error', optimizer=Adam(lr=args.learning_rate))
train_generator = Generator(x_train, y_train, True, args)
validation_generator = Generator(x_valid, y_valid, False, args)
model.fit_generator(train_generator, validation_data=\
validation_generator, epochs=\
args.nb_epoch, use_multiprocessing=\
False, max_queue_size=\
10, workers=\
4)
model.save(model_name)
def train_model(model, args, x_train, x_valid, y_train, y_valid):
"""
Train the model
"""
checkpoint = ModelCheckpoint(
'self-driving-car-train' + str(args.train_num) + '-{epoch:03d}.h5',
monitor='val_loss',
verbose=0,
save_best_only=args.
save_best_only, # save the model only if the val_loss gets low
mode='auto')
model.compile(loss='mean_squared_error',
optimizer=Adam(lr=args.learning_rate))
# shuffle the data because they are sequential; should help over-fitting towards certain parts of the track only
x_train, y_train = shuffle(x_train, y_train, random_state=0)
x_valid, y_valid = shuffle(x_valid, y_valid, random_state=0)
# data for training are augmented, data for validation are not
train_generator = Generator(x_train, y_train, True, args)
validation_generator = Generator(x_valid, y_valid, False, args)
history = model.fit_generator(train_generator,
validation_data=validation_generator,
epochs=args.nb_epoch,
callbacks=[checkpoint],
verbose=1)
# summarize history for loss
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.title('training' + str(args.train_num))
plt.legend(['train', 'val'], loc='upper left')
# plt.show()
plt.savefig('history-training' + str(args.train_num) + '.png')
# save the last model anyway (might not be the best)
model.save("models/model-train" + str(args.train_num) + "-final.h5")
def train_model(model, args, X_train, X_valid, y_train, y_valid):
"""
Train the model
"""
os.makedirs('trained_models', exist_ok=True)
checkpoint = ModelCheckpoint(
'trained_models/self-driving-car-{epoch:03d}.h5',
monitor='val_loss',
verbose=0,
save_best_only=args.save_best_only,
mode='auto')
model.compile(loss='mean_squared_error',
optimizer=Adam(lr=args.learning_rate))
train_generator = Generator(X_train, y_train, True, args)
validation_generator = Generator(X_valid, y_valid, False, args)
# model.fit_generator(train_generator,
# validation_data=validation_generator,
# epochs=args.nb_epoch,
# use_multiprocessing=False,
# max_queue_size=10,
# workers=4,
# callbacks=[checkpoint],
# verbose=1,
# steps_per_epoch=X_train.shape[0],
# validation_steps=1
# )
model.fit_generator(batch_generator(args.data_dir, X_train, y_train,
args.batch_size, True),
args.samples_per_epoch,
args.nb_epoch,
max_q_size=1,
validation_data=batch_generator(
args.data_dir, X_valid, y_valid, args.batch_size,
False),
nb_val_samples=len(X_valid),
callbacks=[checkpoint],
verbose=1)
def train_model(model, args, X_train, X_valid, y_train, y_valid):
"""
Train the model
"""
checkpoint = ModelCheckpoint(args.model + '-{epoch:03d}.h5',
monitor='val_loss',
verbose=0,
save_best_only=args.save_best_only,
mode='auto')
train_generator = Generator(X_train, y_train, True, args)
validation_generator = Generator(X_valid, y_valid, False, args)
model.fit_generator(train_generator,
validation_data=validation_generator,
epochs=args.nb_epoch,
use_multiprocessing=False,
max_queue_size=10,
workers=4,
callbacks=[checkpoint],
verbose=1)
def train_model(model, args, X_train, X_valid, y_train, y_valid):
"""
Train the model
"""
checkpoint = ModelCheckpoint('self-driving-car-{epoch:03d}.h5',
monitor='val_loss',
verbose=0,
save_best_only=args.save_best_only,
mode='auto')
model.compile(loss='mean_squared_error',
optimizer=Adam(lr=args.learning_rate))
train_generator = Generator(X_train, y_train, True, args)
validation_generator = Generator(X_valid, y_valid, False, args)
model.fit_generator(train_generator,
validation_data=validation_generator,
epochs=args.nb_epoch,
use_multiprocessing=True,
max_queue_size=10,
workers=4,
callbacks=[checkpoint],
verbose=1)
def get_embed_helper(model, args, X, y, mode, layer_name, dtheta=0):
intermediate_layer_model = Model(
inputs=model.input,
outputs=(model.get_layer(layer_name).output, model.output))
validation_generator = Generator(X, y, False, args, mode, dtheta)
intermediate_output = intermediate_layer_model.predict_generator(
validation_generator, callbacks=None, verbose=1)
embed = intermediate_output[0]
y_pred = intermediate_output[1].squeeze()
# discard the last few examples
l1_dist = np.abs(y_pred - y[:len(y_pred)])
import sklearn
mse = sklearn.metrics.mean_squared_error(y[:len(y_pred)], y_pred)
# print(mode, 'mean error :', np.mean(l1_dist), '% > 0.1', np.mean(l1_dist>0.1))
return embed, y_pred, l1_dist