search

qgallouedec / panda-gym

Set of robotic environments based on PyBullet physics engine and gymnasium.
MIT License
492 stars 106 forks source link

Documentation Update #50

Closed simoninithomas closed 1 year ago

simoninithomas commented 1 year ago

Hey there 👋 ,

So I tried yesterday panda-gym and I really love the project, my agent needs more training though 😆 : https://huggingface.co/ThomasSimonini/a2c-PandaReachDense-v2

So I've updated the documentation:

I'm going to use panda-gym in addition to classical PyBullet for the A2C unit on the Hugging Face Deep RL Course 🤗 ( normally we use more sample efficient algos when we use robotics but given it's a simulation we will not break anything 😅 )

qgallouedec commented 1 year ago

So I tried yesterday panda-gym and I really love the project, my agent needs more training though 😆 : https://huggingface.co/ThomasSimonini/a2c-PandaReachDense-v2

That's great! If you find satisfactory hyperparameter, feel free to submit them in rl-baselines3-zoo.

  • In SB3 part you call -v3 but since we need to use panda-gym==2.0.0 it's -v2.

Thanks! The documentation was written that way to anticipate the upgrade of sb3 to gym v0.26. But I agree that's it's more clear that way. (I've also fixed the import).

  • I copy the notebook and placed it to examples since the colab link on the readme was not working.

Great, it should solve the issue.

  • I also added the link to the rl-trained-agents models on the HuggingFace Hub.

💯 💯

I'm going to use panda-gym in addition to classical PyBullet for the A2C unit on the Hugging Face Deep RL Course 🤗 ( normally we use more sample efficient algos when we use robotics but given it's a simulation we will not break anything 😅 )

Similarly to my first remark, if you find satisfactory hyperparameter, feel free to submit them in rl-baselines3-zoo.

Your deep RL course is great BTW, I'm glad panda-gym is part of it. If you need any help integrating it, making it work, or anything else, I'm available, feel free to free me.