Views
No views yet
1# https://stackoverflow.com/questions/72483775/stable-baselines3-ppo-how-to-change-clip-range-parameter-during-training
2def lrsched():
3 def reallr(progress):
4 lr = 0.003
5 if progress < 0.85:
6 lr = 0.0005
7 if progress < 0.66:
8 lr = 0.00025
9 if progress < 0.33:
10 lr = 0.0001
11 return lr
12 return reallr
13model = PPO('MlpPolicy', env, n_steps = 512, batch_size = 64, n_epochs = 4, gamma = 0.999, gae_lambda = 0.98, ent_coef = 0.01, verbose=1, learning_rate=lrsched())
14
15model.learn(total_timesteps=3000000)
16
17model_name = "ppo-LunarLander-v2"
18model.save(model_name)
19
20eval_env = Monitor(gym.make("LunarLander-v2"))
21
22# Evaluate the model with 10 evaluation episodes and deterministic=True
23mean_reward, std_reward = evaluate_policy(model, eval_env, n_eval_episodes=10, deterministic=True)
24
25# Print the results
26print(f"mean_reward={mean_reward:.2f} +/- {std_reward}")
27
28# mean_reward=300.79 +/- 21.633168219199078