Update README.md
Browse files
README.md
CHANGED
@@ -25,13 +25,47 @@ model-index:
|
|
25 |
This is a trained model of a **PPO** agent playing **LunarLander-v2**
|
26 |
using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3).
|
27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
## Usage (with Stable-baselines3)
|
29 |
-
TODO: Add your code
|
30 |
|
31 |
|
32 |
```python
|
33 |
-
from stable_baselines3 import
|
34 |
from huggingface_sb3 import load_from_hub
|
35 |
|
36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
```
|
|
|
25 |
This is a trained model of a **PPO** agent playing **LunarLander-v2**
|
26 |
using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3).
|
27 |
|
28 |
+
## Training
|
29 |
+
|
30 |
+
```python
|
31 |
+
from stable_baselines3 import PPO
|
32 |
+
from stable_baselines3.common.env_util import make_vec_env
|
33 |
+
|
34 |
+
env = make_vec_env("LunarLander-v2", n_envs=16)
|
35 |
+
model = PPO('MlpPolicy',
|
36 |
+
env=env,
|
37 |
+
n_steps=1024,
|
38 |
+
batch_size=64,
|
39 |
+
n_epochs=4,
|
40 |
+
gamma=0.999,
|
41 |
+
gae_lambda=0.98,
|
42 |
+
ent_coef=0.01,
|
43 |
+
verbose=1)
|
44 |
+
model.learn(total_timesteps=10000000, progress_bar=True)
|
45 |
+
```
|
46 |
+
|
47 |
+
|
48 |
## Usage (with Stable-baselines3)
|
|
|
49 |
|
50 |
|
51 |
```python
|
52 |
+
from stable_baselines3 import PPO
|
53 |
from huggingface_sb3 import load_from_hub
|
54 |
|
55 |
+
repo_id = "zhuqi/PPO_LunarLander-v2_steps10M" # The repo_id
|
56 |
+
filename = "PPO_LunarLander-v2_steps10000000.zip" # The model filename.zip
|
57 |
+
|
58 |
+
# When the model was trained on Python 3.8 the pickle protocol is 5
|
59 |
+
# But Python 3.6, 3.7 use protocol 4
|
60 |
+
# In order to get compatibility we need to:
|
61 |
+
# 1. Install pickle5 (we done it at the beginning of the colab)
|
62 |
+
# 2. Create a custom empty object we pass as parameter to PPO.load()
|
63 |
+
custom_objects = {
|
64 |
+
"learning_rate": 0.0,
|
65 |
+
"lr_schedule": lambda _: 0.0,
|
66 |
+
"clip_range": lambda _: 0.0,
|
67 |
+
}
|
68 |
+
|
69 |
+
checkpoint = load_from_hub(repo_id, filename)
|
70 |
+
model = PPO.load(checkpoint, custom_objects=custom_objects, print_system_info=True)
|
71 |
```
|