def train():
# load row dataset
with open(config.RAW_DATA_DIR, "r") as f:
data = f.read()
# get tokenized data
tokenized_data = dataset.get_tokenized_data(data)
# split to train and test data
train_data, test_data = dataset.split_to_train_and_test_sets(
tokenized_data)
# get vocabulary and processed train and test data
minimum_freq = 2
train_data_processed, test_data_processed, vocabulary = dataset.preprocess_data(
train_data, test_data, minimum_freq)
print(type(vocabulary))
def train(
dataset,
job_folder,
logger,
video_root_path='/home/miruware/anomaly/anomaly-detection-and-localization/VIDEO_ROOT_PATH'
):
"""Build and train the model
"""
import yaml
import numpy as np
from keras.callbacks import ModelCheckpoint, EarlyStopping
from custom_callback import LossHistory
import matplotlib.pyplot as plt
from keras.utils.io_utils import HDF5Matrix
logger.debug("Loading configs from {}".format(
os.path.join(job_folder, 'config.yml')))
with open(os.path.join(job_folder, 'config.yml'), 'r') as ymlfile:
cfg = yaml.load(ymlfile)
nb_epoch = cfg['epochs']
batch_size = cfg['batch_size']
loss = cfg['cost']
optimizer = cfg['optimizer']
time_length = cfg['time_length']
# shuffle = cfg['shuffle']
# logger.info("Building model of type {} and activation {}".format(model_type, activation))
model = get_model(time_length)
logger.info("Compiling model with {} and {} optimizer".format(
loss, optimizer))
compile_model(model, loss, optimizer)
logger.info("Saving model configuration to {}".format(
os.path.join(job_folder, 'model.yml')))
yaml_string = model.to_yaml()
with open(os.path.join(job_folder, 'model.yml'), 'w') as outfile:
yaml.dump(yaml_string, outfile)
logger.info("Preparing training and testing data")
preprocess_data(logger, dataset, time_length, video_root_path)
data = HDF5Matrix(
os.path.join(video_root_path,
'{0}/{0}_train_t{1}.h5'.format(dataset, time_length)),
'data')
snapshot = ModelCheckpoint(
os.path.join(job_folder,
'model_snapshot_e{epoch:03d}_{val_loss:.6f}.h5'))
earlystop = EarlyStopping(patience=5)
history_log = LossHistory(job_folder=job_folder, logger=logger)
logger.info("Initializing training...")
history = model.fit(data,
data,
batch_size=batch_size,
epochs=nb_epoch,
validation_split=0.15,
shuffle='batch',
callbacks=[snapshot, earlystop, history_log])
logger.info("Training completed!")
np.save(os.path.join(job_folder, 'train_profile.npy'), history.history)
n_epoch = len(history.history['loss'])
logger.info("Plotting training profile for {} epochs".format(n_epoch))
plt.plot(range(1, n_epoch + 1),
history.history['val_loss'],
'g-',
label='Val Loss')
plt.plot(range(1, n_epoch + 1),
history.history['loss'],
'g--',
label='Training Loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.savefig(os.path.join(job_folder, 'train_val_loss.png'))
(1 - hidden_outputs)), np.transpose(train_data))
def query(self, inputs_list):
query_inputs = np.array(inputs_list, ndmin=2).T
hidden_inputs = np.dot(self.weights_input_to_hidden, query_inputs)
hidden_outputs = activate(hidden_inputs)
final_inputs = np.dot(self.weights_hidden_to_output, hidden_outputs)
final_outputs = activate(final_inputs)
return final_outputs
print('Начало работы\n')
training_data, target = preprocess_data('latin_alphabet/', 260, 1024)
print('Входные данные: ', training_data, '\n')
print('Выходные данные: ', target, '\n')
network = Network(1024, 512, 26)
print("\nПроверка нейронной сети до тренировки на букве 'a'\n")
result = network.query(training_data[2])
for j in range(len(result)):
print(ALPHABET.get(j) + ") ", result[j])
print("\n")
for e in range(100):
print('Эпоха: ', e + 1)
for i in range(len(training_data)):
inputs = training_data[i]
targets = np.zeros(26) + 0.01