import time
from options.train_options import TrainOptions
from data.dataloader import CreateDataLoader
from util.visualizer import Visualizer
from models import create_model
def main():
opt = TrainOptions().parse()
data_loader = CreateDataLoader(opt)
dataset_size = len(data_loader) * opt.batch_size
visualizer = Visualizer(opt)
model = create_model(opt)
start_epoch = model.start_epoch
total_steps = start_epoch*dataset_size
for epoch in range(start_epoch+1, opt.niter+opt.niter_decay+1):
epoch_start_time = time.time()
model.update_lr()
save_result = True
for i, data in enumerate(data_loader):
iter_start_time = time.time()
total_steps += opt.batch_size
epoch_iter = total_steps - dataset_size * (epoch - 1)
model.prepare_data(data)
model.update_model()
if save_result or total_steps % opt.display_freq == 0:
save_result = save_result or total_steps % opt.update_html_freq == 0
visualizer.display_current_results(model.get_current_visuals(), epoch, ncols=1, save_result=save_result)
save_result = False
if total_steps % opt.print_freq == 0:
errors = model.get_current_errors()
t = (time.time() - iter_start_time) / opt.batch_size
visualizer.print_current_errors(epoch, epoch_iter, errors, t)
if opt.display_id > 0:
visualizer.plot_current_errors(epoch, float(epoch_iter)/dataset_size, opt, errors)
print('epoch {} cost dime {}'.format(epoch,time.time()-epoch_start_time))
model.save_ckpt(epoch)
model.save_generator('latest')
if epoch % opt.save_epoch_freq == 0:
print('saving the generator at the end of epoch {}, iters {}'.format(epoch, total_steps))
model.save_generator(epoch)
if __name__ == '__main__':
main()
