_, _, errors = learner.get_targets([(0, (state, action, reward,
new_state, done))])
learner.add_memory(errors[0],
(state, action, reward, new_state, done))
state = new_state
if done or step == max_steps - 1:
score_list.append(step)
break
# Decay the epsilon
if actor.epsilon > actor.epsilon_min:
actor.epsilon *= actor.epsilon_decay
print("Episode: {}, Score: {}/{}".format(
e,
sum(score_list) / len(score_list), max_steps - 1))
scores.append(sum(score_list) / len(score_list))
x = range(len(scores))
y = scores
learner.replay()
plt.scatter(x[-1], y[-1])
plt.pause(0.05)
if e % 25 == 0:
learner.update_target()
learner.save_model('./models/inverted_pendulum_v0.2.h5')
