google-deepmind / google-deepmind/mujoco_playground
Madrona-MJX
Open
question
- Dominant language
- Python
- Stars
- 2.2k
- Forks
- 359
- Avg merge
- 7d 3h
- Merged PRs (30d)
- 1
Description
Hello, it's a really hot project and thanks for open-source this project
I tried some training on your open-source envs&codes but i got a issue
I wanna train vision-based PPO policy (in your env : PandaPickCube (--vision = True)),
But the brax, jax version dependencies are quite confused between mujoco_playground and Madrona MJX
so I wonder which version of brax, jax, you use in local computer to train that env.
thanks a lot
------
what i found
Brax 0.12.4 + JAX 0.5.3: error in cuSolver
JAX 0.6.0+: can't serve some API in Madrona-MJX(they said in github)
JAX 0.5.x: Dependency gap between Brax and CUDA (12.xx)
Contributor guide
Assessment
This issue has not been assessed yet.