epochs = int(params.get('e', 60))
bias_e = float(params.get('bias', 1.3))
logging_helper.debug('epochs={}, batchsize={}, bias_e={}'.format(
epochs, batch_size, bias_e))
img_width, img_height = 512, 512
# train_data_dir = '../pre_data'
# validation_data_dir = '../val_data'
nb_train_samples = 220
nb_validation_samples = 100
#
img_dir = './images/origin-gen'
test_dir = './images/img_test'
samples_num = 5200
model = create_noise_est_model(img_height=img_height,
img_width=img_width,
channel_num=3)
callbacks = [
my_callbacks.NoiseEstCallback(img_dir=test_dir,
height=img_height,
width=img_width,
max_sigma=60)
]
model.fit_generator(generator=noise_img_gen(img_dir=img_dir,
batch_size=batch_size,
sigma_max_value=60),
steps_per_epoch=samples_num // batch_size,
epochs=epochs,
callbacks=callbacks)
model_helper.save_model(model=model,
struct_path='est_model.json',
weight_path='est_model.h5')
feature_train, target_train, pred, mape = zip(* cleaned)
start_time = time()
model.fit(feature_train, target_train)
print("re-training time after cleaning outliers", round(time() - start_time, 3), "s")
# Start prediction using cross validation set
start_time = time()
pred = model.predict(feature_test)
print("predition time", round(time() - start_time, 3), "s")
# Show MAPE score of trained model with cross validation set
score = mape_score(pred, target_test)
print('MAPE score ', score)
# Save model for later use
save_model(model, './model', MODEL_NAME)
from result_helper import save_result
data_file = '../../processed_data/prediction_data/to_predict_features'
test_metadata_list, test_features_list, test_target_list = load_and_split_data(data_file)
test_features_list = feature_selection.transform(test_features_list)
test_predictions = model.predict(test_features_list)
OUTPUT_FOLDER = './output'
if __name__ == "__main__":
# Load and split training data into metadata, features, target
data_file = '../../processed_data/features_data/training_data/features_ready'
metadata_list, features_list, target_list = load_and_split_data(data_file)
# Split data into training set and cross validation set
feature_train, feature_test, target_train, target_test = train_test_split(
features_list,
target_list,
test_size=TEST_SIZE,
random_state=RANDOM_STATE)
# Start training
start_time = time()
model = create_model()
model.fit(feature_train, target_train)
print("training time", round(time() - start_time, 3), "s")
# Start prediction using cross validation set
start_time = time()
pred = model.predict(feature_test)
print("predition time", round(time() - start_time, 3), "s")
# Show MAPE score of trained model with cross validation set
score = mape_score(pred, target_test)
print('MAPE score ', score)
# Save model for later use
save_model(model, './model', MODEL_NAME)
import sys
from model_helper import save_model
from keras.models import Sequential
from keras.layers.core import Dense, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D
model_file = sys.argv[1]
weights_file = sys.argv[2]
model = Sequential()
model.add(Convolution2D(8, 5, 5, input_shape = (1, 28, 28)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2, 2), strides = (2, 2)))
model.add(Convolution2D(16, 5, 5))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (3, 3), strides = (3, 3)))
model.add(Flatten())
model.add(Dense(10))
model.add(Activation('softmax'))
save_model(model, model_file, weights_file)
with tf.Session() as sess:
sess.run(init)
print('Learning...')
for epoch in range(config.learning_epochs):
_, cost = sess.run([optimizer, cost_func],
feed_dict={
x: states,
y: rewards
})
print('Learning for X: Epoch ', epoch, ' of ',
config.learning_epochs, ' with cost of ', cost)
print('saving model X...')
model_helper.save_model(sess, 'tf-model/X', x, pred)
# Learn for O
with open('learn_o.dat', 'rb') as learn_o_file:
states = []
rewards = []
print('Reading games...')
while True:
try:
rows = pickle.load(learn_o_file).tolist()
for row in rows:
states.append(row)
rows = pickle.load(learn_o_file).tolist()
for row in rows:
rewards.append(row)
import sys
from model_helper import save_model
from keras.models import Sequential
from keras.layers.core import Dense, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D
model_file = sys.argv[1]
weights_file = sys.argv[2]
model = Sequential()
model.add(Convolution2D(8, 5, 5, input_shape=(1, 28, 28)))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
model.add(Convolution2D(16, 5, 5))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(3, 3), strides=(3, 3)))
model.add(Flatten())
model.add(Dense(10))
model.add(Activation("softmax"))
save_model(model, model_file, weights_file)
print('### Trainging the model')
stop_accuracy = args.accurecy / 100.0
epoch = mh.train_model(device,
model,
trainloader,
optimizer,
criterion,
validloader,
verbose=True,
stop_accuracy=stop_accuracy)
# Testing Network
print('### Testing the model')
accuracy = mh.test_model(device, model, criterion, testloader)
print(f"Accuracy: {accuracy:.2f}%")
# Saving the model
print('### Saving Model')
filepath = args.save_dir
mh.save_model(model,
arch,
train_dataset,
optimizer,
epoch,
filepath,
nHiddens=nHiddens,
nOutputs=nOutputs,
pDropout=pDropout,
lr=lr)
print('Model Saved to: ', filepath)