def ohe_test(self, scores, episode_numbers):
"""
method runs 100 testing episodes
"""
complete_estimation = agent_score_estimate(self, 10)
if complete_estimation >= self.solved_score:
for i in range(2, 11):
estimation = agent_score_estimate(self, 10)
complete_estimation = complete_estimation + estimation
if (complete_estimation / i) < self.solved_score:
return
else:
return
score = complete_estimation / 10
if score > self.solved_score:
if not self.args.dont_save:
self.agent.save_model_weights("{}-solved.h5" .format(self.name))
combined_graph(scores, episode_numbers, "{}_results" .format(self.name),
[episode_numbers[-1], max(scores)+10], {self.average_rand_score:self.average_rand_score}, scatter=True)
if self.solved_score <= score:
print("[Task was solved after {} episodes with score {}.]" .format(episode_numbers[-1], score))
else:
print("[Task wasn't solved after {} episodes with score {}.]" .format(episode_numbers[-1], score))
print("[SUCCESSFUL RUN]")
K.clear_session()
sys.exit()
def aohe_test(self, scores, episode_numbers):
"""
method runs 100 testing episodes after 100 training episodes
"""
if episode_numbers[-1] == 99:
score = agent_score_estimate(self, 100)
self.agent.save_model_weights("{}-solved.h5" .format(self.name))
combined_graph(scores, episode_numbers, "{}_results" .format(self.name),
[episode_numbers[-1], max(scores)+10], {self.average_rand_score:self.average_rand_score}, scatter=True)
print("[Task was solved after {} episodes with score {}.]" .format(episode_numbers[-1], score))
print("[SUCCESSFUL RUN]")
K.clear_session()
sys.exit()
if task.args.mode == "train":
task, true_scores, episodes_numbers, _, normalized_scores = train(
task, True)
task.agent.save_model_weights("{}-last.h5".format(task.name))
task.agent.save_target_weights("{}-last.h5".format(task.name))
try:
combined_graph(true_scores, episodes_numbers,
"{}_results".format(task.name),
[episodes_numbers[-1] + 10,
max(true_scores) + 10],
{task.average_rand_score: task.average_rand_score},
True)
except ValueError:
err_print("[Not enough values to make graph.]")
elif task.args.mode == "test":
agent_score_estimate(task,
task.args.episodes,
render=False,
show_bar=True)
elif task.args.mode == "render":
agent_score_estimate(task,
task.args.episodes,
render=True,
show_bar=True)
K.clear_session()
print("[SUCCESSFUL RUN]")