- add of a2c_sb3_panda_reach.ipynb (notebook version)

- removal of hf and wandb logs from code (for security)
- final version of the reamdme (last part added)

README.md

@@ -25,9 +25,20 @@ see [a2c_sb3_cartpole.py](a2c_sb3_cartpole.py)
 
 ### Hugging Face Hub
 
-[Link to the trained model](https://huggingface.co/Thomstr/A2C_CartPole/tree/main)
+[Link to the trained model (cartpole)](https://huggingface.co/Thomstr/A2C_CartPole/tree/main)
 
 ### Weights & Biases
-[Link to the wandb run](https://wandb.ai/thomasdgr-ecole-centrale-de-lyon/cartpole/runs/vh4anh20/workspace?nw=nwuserthomasdgr)
+[Link to the wandb run (cartpole)](https://wandb.ai/thomasdgr-ecole-centrale-de-lyon/cartpole/runs/vh4anh20/workspace?nw=nwuserthomasdgr)
 
 ### Full workflow with panda-gym
+see [a2c_sb3_panda_reach.py](a2c_sb3_panda_reach.py)
+
+As I couldn't make it work on my PC (difficulties to install panda-gym), I've used Google Colab.
+
+see my notebook [here (online)](https://colab.research.google.com/drive/1l03F398QLHHVVqJ-GvRgxA4d-cCocF4K?usp=sharing) 
+or directly [a2c_sb3_panda_reach.ipynb](a2c_sb3_panda_reach.ipynb)
+
+
+[Link to the trained model (panda reach)](https://huggingface.co/Thomstr/A2C_PandaReach/tree/main)
+
+[Link to the wandb run (panda reach)](https://wandb.ai/thomasdgr-ecole-centrale-de-lyon/pandareach/runs/y39cy9ws?nw=nwuserthomasdgr)

a2c_sb3_cartpole.py

@@ -21,7 +21,7 @@ if __name__ == "__main__":
         "total_timesteps": 25000,
         "env_name": "CartPole-v1",
     }
-    wandb.login(key='4ac81e81b051a56ebfc528b579021cfc9ed1e5dc')
+    wandb.login(key='xxxxxxx')
     run = wandb.init(
         project="cartpole",
         config=config,

a2c_sb3_panda_reach.py

-import gym
+import gymnasium as gym
 import panda_gym
 from stable_baselines3 import A2C
 from stable_baselines3.common.monitor import Monitor
 from stable_baselines3.common.vec_env import DummyVecEnv
 import wandb
 from wandb.integration.sb3 import WandbCallback
+from huggingface_hub import login
+from huggingface_sb3 import push_to_hub
 
 
+if __name__ == "__main__":
     config = {
         "policy_type": "MultiInputPolicy",
         "total_timesteps": 500000,
         "env_name": "PandaReachJointsDense-v3",
     }
 
+    wandb.login(key='xxx')
+
     run = wandb.init(
         project="pandareach",
         config=config,
@@ -21,13 +26,7 @@ run = wandb.init(
         save_code=True,
     )
 
-def make_env():
     env = gym.make(config["env_name"])
-    env = Monitor(env)  # record stats such as returns
-    return env
-
-env = DummyVecEnv([make_env])
-env = gym.make("PandaReachJointsDense-v3")
     model = A2C(config["policy_type"], env, verbose=1, tensorboard_log=f"runs/{run.id}")
     model.learn(
         total_timesteps=config["total_timesteps"],
@@ -37,7 +36,9 @@ model.learn(
 
     run.finish()
 
-login(token="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")
+    model.save("a2c_pandareach")
+
+    login(token="xxx")
     push_to_hub(
         repo_id="Thomstr/A2C_PandaReach",
         filename="a2c_pandareach.zip",