def render(click_reset, click_training, cost_factor, risk_factor,
levels_cooling, lr, n_episodes, type_agent):
print("Reset ", click_reset, " - ", reset_clicks.count)
print("Train ", click_training, " - ", train_clicks.count)
if click_reset > reset_clicks.count:
reset_clicks.count = click_reset
env.__init__(levels_cooling=levels_cooling,
risk_factor=risk_factor,
cost_factor=cost_factor,
keep_cooling=True)
elif click_training > train_clicks.count:
train_clicks.count = click_training
env_temp, agent, rewards = run_n_episodes(env,
n_episodes=n_episodes,
lr=lr,
type_agent=type_agent)
utils.plot_average_running_rewards(
rewards, "C:/Users/talvesdacosta/Desktop/results.png")
# os.system("start "+"C:/Users/talvesdacosta/Desktop/results.png")
env.cooling = env_temp.cooling
else:
env.risk_factor = risk_factor
env.cost_factor = cost_factor
return env.render(with_plotly=True)
r,
np.expand_dims(s_next,axis=0),
np.expand_dims(s_before,axis=0),
done)
# Go to the next state
s_before = s
s = s_next
# If the episode is terminated
if done:
print("Episode {}/{} finished after {} timesteps - epsilon : {:.2}".format(i_episode+1,N_EPISODES,i_step,agent.epsilon))
break
#-----------------------------------------
# Store the rewards
rewards.append(i_step)
# Training
agent.train(batch_size = BATCH_SIZE)
# Plot the average running rewards
utils.plot_average_running_rewards(rewards)