# train discriminator
trainer.run_discriminator_one_step(data_i)
# Visualizations
if iter_counter.needs_printing():
losses = trainer.get_latest_losses()
visualizer.print_current_errors(epoch, iter_counter.epoch_iter,
losses, iter_counter.time_per_iter)
visualizer.plot_current_errors(losses,
iter_counter.total_steps_so_far)
if iter_counter.needs_displaying():
visuals = OrderedDict([('input_label', data_i['label']),
('synthesized_image',
trainer.get_latest_generated()),
('real_image', data_i['image'])])
visualizer.display_current_results(visuals, epoch,
iter_counter.total_steps_so_far)
if iter_counter.needs_saving():
print('saving the latest model (epoch %d, total_steps %d)' %
(epoch, iter_counter.total_steps_so_far))
trainer.save('latest')
iter_counter.record_current_iter()
trainer.update_learning_rate(epoch)
iter_counter.record_epoch_end()
if epoch % opt.save_epoch_freq == 0 or \
epoch == iter_counter.total_epochs:
# create tool for visualization
visualizer = Visualizer(opt)
for epoch in iter_counter.training_epochs():
iter_counter.record_epoch_start(epoch)
for i, data in enumerate(dataloader, start=iter_counter.epoch_iter):
iter_counter.record_one_iter()
trainer.g_losses, trainer.d_losses = {}, {}
CT, MR = data['CT'].squeeze(1), data['MR'].squeeze(1)
if opt.D_steps_per_G:
trainer.run_generator_one_step(CT, MR)
trainer.run_discriminator_one_step(CT, MR)
data = {**data, **trainer.get_latest_generated()}
# Visualizations
if iter_counter.needs_printing():
losses = trainer.get_latest_losses()
visualizer.print_current_errors(epoch, iter_counter.epoch_iter,
losses, iter_counter.time_per_iter)
visualizer.plot_current_errors(
losses, iter_counter.total_steps_so_far)
if iter_counter.needs_displaying():
visuals = OrderedDict([**data])
visualizer.display_current_results(
visuals, epoch, iter_counter.total_steps_so_far)
if iter_counter.needs_saving():
