George Yang

icon indicating a website link george614.github.io icon indicating an email [email protected]

icon company Writer icon location AI Engineer @Writer. Research interests include AI/ML, continual learning, active inference, reinforcement learning, GenAI, AR/VR, computer vision.

Results 5 issues of George Yang

[feature request] allowing user play / control

1 comment

Can we have a function that allows user to play as an agent in the environment? It allows for demonstration data collection etc.

Data collection script doesn't work, API change in robosuite

Hi Brent, I'm using the data collection script simulate_door.py, but it returns an error indicating that the PandaIKControler was not found. I looked up the major versions of robosuite API,...

Unable to open file (file signature not found) Errror

Hi Brent, I'm trying to run the train_door task with one of the filters. But it seems that the data file needs to be generated before training? Below is the...

[Feature Request] Continual Learning and Evaluation

Hi Danijar, I've been a big fan of your works. Can you consider adding functionalities for supporting continual reinforcement learning and evaluation setup in the future? I think it's a...

Pong results do not match paper

5 comment

Hi Danijar, Thanks for sharing this amazing repo and creating a robust model-based RL algorithm! I've been playing with the replay buffer and trying to reproduce some of the results....