prashanthgowni
commited on
Commit
•
c5855e2
1
Parent(s):
6f329c8
Update README.md
Browse files
README.md
CHANGED
@@ -19,6 +19,8 @@ model-index:
|
|
19 |
value: 277.82 +/- 22.28
|
20 |
name: mean_reward
|
21 |
verified: false
|
|
|
|
|
22 |
---
|
23 |
|
24 |
# **PPO** Agent playing **LunarLander-v2**
|
@@ -30,8 +32,41 @@ TODO: Add your code
|
|
30 |
|
31 |
|
32 |
```python
|
33 |
-
from stable_baselines3 import
|
|
|
|
|
|
|
34 |
from huggingface_sb3 import load_from_hub
|
35 |
|
36 |
-
|
37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
value: 277.82 +/- 22.28
|
20 |
name: mean_reward
|
21 |
verified: false
|
22 |
+
language:
|
23 |
+
- en
|
24 |
---
|
25 |
|
26 |
# **PPO** Agent playing **LunarLander-v2**
|
|
|
32 |
|
33 |
|
34 |
```python
|
35 |
+
from stable_baselines3 import PPO
|
36 |
+
from stable_baselines3.common.env_util import make_vec_env
|
37 |
+
from stable_baselines3.common.evaluation import evaluate_policy
|
38 |
+
|
39 |
from huggingface_sb3 import load_from_hub
|
40 |
|
41 |
+
|
42 |
+
# Download the model checkpoint
|
43 |
+
model_checkpoint = load_from_hub("prashanthgowni/ppo-LunarLander-v2", "ppo-LunarLander-v2")
|
44 |
+
# Create a vectorized environment
|
45 |
+
env = make_vec_env("LunarLander-v2", n_envs=1)
|
46 |
+
|
47 |
+
# Load the model
|
48 |
+
model = PPO.load(model_checkpoint, env=env)
|
49 |
+
|
50 |
+
# Evaluate
|
51 |
+
print("Evaluating model")
|
52 |
+
mean_reward, std_reward = evaluate_policy(
|
53 |
+
model,
|
54 |
+
env,
|
55 |
+
n_eval_episodes=30,
|
56 |
+
deterministic=True,
|
57 |
+
)
|
58 |
+
print(f"Mean reward = {mean_reward:.2f} +/- {std_reward}")
|
59 |
+
|
60 |
+
# Start a new episode
|
61 |
+
obs = env.reset()
|
62 |
+
|
63 |
+
try:
|
64 |
+
while True:
|
65 |
+
action, state = model.predict(obs, deterministic=True)
|
66 |
+
obs, reward, done, info = env.step(action)
|
67 |
+
env.render()
|
68 |
+
|
69 |
+
except KeyboardInterrupt:
|
70 |
+
pass
|
71 |
+
|
72 |
+
```
|