def handle_plot_ckpt(do_plot=False):
train_loss = np.array(to_scalar([loss_1, loss_2, loss_3])).mean(0)
info['train_losses'].append(train_loss)
info['train_cnts'].append(train_cnt)
vx, vy, _ = data_loader.validation_data()
vloss_1, vloss_2, vloss_3, vx_d, vz_e_x, vz_q_x, vlatents = forward_pass(
vx, vy)
test_loss = np.array(to_scalar([vloss_1, vloss_2,
vloss_3])).mean(0)
info['test_losses'].append(test_loss)
info['test_cnts'].append(train_cnt)
print(
'examples %010d train loss %03.03f test loss %03.03f' %
(train_cnt, info['train_losses'][-1], info['test_losses'][-1]))
if do_plot:
info['last_plot'] = train_cnt
plot_name = os.path.join(
default_base_savedir,
basename + "_%010dloss.png" % train_cnt)
print('plotting: %s' % plot_name)
n = 3
plot_losses(info['train_cnts'],
info['train_losses'],
info['test_cnts'],
info['test_losses'],
name=plot_name,
rolling_length=n)
def handle_plot_ckpt(do_plot, train_cnt, avg_train_loss):
info['train_losses'].append(avg_train_loss)
info['train_cnts'].append(train_cnt)
test_loss = test_acn(train_cnt,do_plot)
info['test_losses'].append(test_loss)
info['test_cnts'].append(train_cnt)
print('examples %010d train loss %03.03f test loss %03.03f' %(train_cnt,
info['train_losses'][-1], info['test_losses'][-1]))
# plot
if do_plot:
info['last_plot'] = train_cnt
rolling = 3
if len(info['train_losses'])<rolling*3:
rolling = 1
print('adding last loss plot', train_cnt)
plot_name = vae_base_filepath + "_%010dloss.png"%train_cnt
print('plotting loss: %s with %s points'%(plot_name, len(info['train_cnts'])))
plot_losses(info['train_cnts'],
info['train_losses'],
info['test_cnts'],
info['test_losses'], name=plot_name, rolling_length=rolling)