zhengyinan-air / fisor Goto Github PK

[ICLR 2024] The official implementation of "Safe Offline Reinforcement Learning with Feasibility-Guided Diffusion Model"

Home Page: https://zhengyinan-air.github.io/FISOR/

Python 75.47% Shell 0.68% CSS 4.89% HTML 18.96%

diffusion-models hamilton-jacobi-reachability jax offline-reinforcement-learning safe-reinforcement-learning imitation-learning reinforcement-learning

fisor's People

Contributors

Stargazers

Watchers

Forkers

air-di facebear-ljx

fisor's Issues

Actor issue

Hi, I trained the model and used the same parameters in the train_offline.py for env: PointRobot. But I find the actor performs well at some points but does not show feasible policy at other points. Why does this happen? and how should I elaborate on my results.

Best,
Jiahui

Why the time-variable t need to be input into FourierFeatures class？

Hi!
In 'https://github.com/ZhengYinan-AIR/FISOR/blob/master/jaxrl5/networks/diffusion.py', the time variable t were input into FourierFeatures and then the mlp to be a time embediing. I wonder what is it for doing so. It seems that the fourier features is one of the high frequency features, if so, why such a high frequency feature need to be extracted and then be emdded?

Recommend Projects