araffin
/

dqn-LunarLander-v2

Reinforcement Learning

stable-baselines3

deep-reinforcement-learning

Model card Files Files and versions Community

araffin commited on May 5, 2022

Commit

fa50406

•

1 Parent(s): dd1f7df

Update code

Files changed (1) hide show

README.md +8 -8

README.md CHANGED Viewed

@@ -32,7 +32,7 @@ from stable_baselines3.common.callbacks import EvalCallback
 # Create the environment
 env_id = "LunarLander-v2"
-n_envs = 1
 env = make_vec_env(env_id, n_envs=n_envs)
 # Create the evaluation envs
@@ -58,14 +58,14 @@ model = DQN(
     env,
     learning_starts=0,
     batch_size=128,
-    buffer_size=50000,
-    learning_rate=1e-3,
     target_update_interval=250,
-    train_freq=4,
-    gradient_steps=-1,
-    # Explore for 20_000 timesteps
-    exploration_fraction=0.04,
-    exploration_final_eps=0.1,
     policy_kwargs=dict(net_arch=[256, 256]),
     verbose=1,
 )

 # Create the environment
 env_id = "LunarLander-v2"
+n_envs = 8
 env = make_vec_env(env_id, n_envs=n_envs)
 # Create the evaluation envs
     env,
     learning_starts=0,
     batch_size=128,
+    buffer_size=100000,
+    learning_rate=7e-4,
     target_update_interval=250,
+    train_freq=1,
+    gradient_steps=4,
+    # Explore for 40_000 timesteps
+    exploration_fraction=0.08,
+    exploration_final_eps=0.05,
     policy_kwargs=dict(net_arch=[256, 256]),
     verbose=1,
 )