saymrwulf/stable-baselines3
1
0
Fork 0
mirror of https://github.com/saymrwulf/stable-baselines3.git synced 2026-07-06 04:28:49 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Print std reward for evaluation

Browse source
This commit is contained in:
Antonin RAFFIN 2019-12-24 13:12:04 +01:00
parent 57c890f3e9
commit 4a79f7e5a7
2 changed files with 3 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
torchy_baselines/common/evaluation.py
View file

@ -19,4 +19,4 @@ def evaluate_policy(model, env, n_eval_episodes=10, deterministic=True, render=F
env.render()
episode_rewards.append(episode_reward)
return np.mean(episode_rewards), n_steps
return np.mean(episode_rewards), np.std(episode_rewards)

4
torchy_baselines/sac/sac.py
View file

@ -291,10 +291,10 @@ class SAC(BaseRLModel):
if 0 < eval_freq <= timesteps_since_eval and eval_env is not None:
timesteps_since_eval %= eval_freq
sync_envs_normalization(self.env, eval_env)
mean_reward, _ = evaluate_policy(self, eval_env, n_eval_episodes)
mean_reward, std_reward = evaluate_policy(self, eval_env, n_eval_episodes)
evaluations.append(mean_reward)
if self.verbose > 0:
print("Eval num_timesteps={}, mean_reward={:.2f}".format(self.num_timesteps, evaluations[-1]))
print("Eval num_timesteps={}, mean_reward={:.2f}, std_reward={:.2f}".format(self.num_timesteps, mean_reward, std_reward))
print("FPS: {:.2f}".format(self.num_timesteps / (time.time() - self.start_time)))
return self