def trainnetwork(config_file):
#Open configuration file for this network
config = ConfigParser.ConfigParser()
config.read(config_file)
#Set the device on which to perform these computations
device = config.get('General', 'device')
theano.sandbox.cuda.use(device)
if (device != 'cpu'):
theano.config.nvcc.flags='-use=fast=math'
theano.config.allow_gc=False
#------------------------------------------------------------------------------
print '\nOpening Data Files'
#Load the data for training
training_data = LoadData(directory = config.get('Training Data', 'folders').split(','),
data_file_name = config.get('Training Data', 'data_file'),
label_file_name = config.get('Training Data', 'label_file'),
seg_file_name = config.get('Training Data', 'seg_file'))
#------------------------------------------------------------------------------
starttime=time.clock()
#Create the network and trainer
if os.path.exists(config.get('General', 'directory')+config.get('Network', 'weights_folder')):
print 'Loading Network'
network = CNN(weights_folder = config.get('General', 'directory')+config.get('Network', 'weights_folder'),
activation = config.get('Network', 'activation'))
print 'Loading Trainer'
network_trainer = Trainer(network, training_data.get_data(), training_data.get_labels(), training_data.get_segments(),
chunk_size = config.getint('Training', 'chunk_size'),
batch_size = config.getint('Training', 'batch_size'),
cost_func = config.get('Training', 'cost_func'),
learning_method = config.get('Training', 'learning_method'),
learning_rate = config.getfloat('Training', 'learning_rate'),
beta1 = config.getfloat('Training', 'beta1'),
beta2 = config.getfloat('Training', 'beta2'),
damping = config.getfloat('Training', 'damping'),
trainer_folder = config.get('General', 'directory')+config.get('Training', 'trainer_folder'),
log_interval = config.getint('Training', 'log_interval'),
log_folder = config.get('General', 'directory')+config.get('Training', 'log_folder'))
else:
print 'Initializing Network'
network = CNN(num_layers = config.getint('Network', 'num_layers'),
num_filters = config.getint('Network', 'num_filters'),
filter_size = config.getint('Network', 'filter_size'),
activation = config.get('Network', 'activation'))
print 'Initializing Trainer'
network_trainer = Trainer(network, training_data.get_data(), training_data.get_labels(), training_data.get_segments(),
chunk_size = config.getint('Training', 'chunk_size'),
batch_size = config.getint('Training', 'batch_size'),
cost_func = config.get('Training', 'cost_func'),
learning_method = config.get('Training', 'learning_method'),
learning_rate = config.getfloat('Training', 'learning_rate'),
beta1 = config.getfloat('Training', 'beta1'),
beta2 = config.getfloat('Training', 'beta2'),
damping = config.getfloat('Training', 'damping'),
log_interval = config.getint('Training', 'log_interval'),
log_folder = config.get('General', 'directory')+config.get('Training', 'log_folder'))
init_time = time.clock() - starttime
#------------------------------------------------------------------------------
print "Initialization = " + `init_time` + " seconds"
starttime = time.clock()
#Train the network
print 'Training...\n'
train_error = network_trainer.train(config.getint('Training', 'num_epochs'),
config.getboolean('Training', 'early_stop'),
config.getboolean('Training', 'print_updates'))
total_time = time.clock() - starttime
#------------------------------------------------------------------------------
print "Total Time =",total_time,"seconds"
training_data.close()
return train_error
