Bug in differentiable physics tutorial notebook #2741
Description
Intro
Hi!
I am a (graduate student) at the University of Groningen, I want to use MuJoCo for my research on locomotion via differentiable simulation. However, when trying to run the differentiable physics tutorial notebook linked in the official README.md
My setup
I simply opened the tutorial jupyter notebook in google colab, set the "runtime type" to "T4 GPU" and pressed "run all".
What's happening? What did you expect?
I get the following error in the Imitation Learning via FoPG's cell:
/usr/local/lib/python3.11/dist-packages/brax/io/mjcf.py:480: UserWarning: Brax System, piplines and environments are not actively being maintained. Please see MJX for a well maintained JAX-based physics engine: https://github.com/google-deepmind/mujoco/tree/main/mjx. For a host of environments that use MJX, see: https://github.com/google-deepmind/mujoco_playground.
warnings.warn(
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
[/usr/local/lib/python3.11/dist-packages/mujoco/mjx/_src/types.py](https://localhost:8080/#) in __getattr__(self, name)
1135 impl_instsance = object.__getattribute__(self, '_impl')
-> 1136 val = getattr(impl_instsance, name)
1137 warnings.warn(
AttributeError: 'DataJAX' object has no attribute 'pos'
During handling of the above exception, another exception occurred:
AttributeError Traceback (most recent call last)
15 frames
[... skipping hidden 13 frame]
[... skipping hidden 23 frame]
[... skipping hidden 10 frame]
[... skipping hidden 7 frame]
[/usr/local/lib/python3.11/dist-packages/mujoco/mjx/_src/types.py](https://localhost:8080/#) in __getattr__(self, name)
1143 except AttributeError:
1144 # raise the standard exception
-> 1145 raise AttributeError( # pylint: disable=raise-missing-from
1146 f"'{type(self).__name__}' object has no attribute '{name}'"
1147 )
AttributeError: 'State' object has no attribute 'pos'
If I manually specify python package versions from the time of the last edit of the notebook, i.e.:
!pip install mujoco==3.2.4
!pip install mujoco_mjx==3.2.4
!pip install brax==0.11.0
then the error disappears. However, on the recent versions of mujoco (3.3.3 or 3.3.4) the notebook does not run.
Steps for reproduction
- Open notebook
- Select a gpu runtime
- I had to change
jax.lax.Precision.HIGHto"high"in the lineconfig.update('jax_default_matmul_precision', jax.lax.Precision.HIGH)in cell Import MuJoCo, MJX, and Brax to avoid aValueError: new enum value must be None or in ['default', 'high' ...error. - "Run all"
Minimal model for reproduction
N/A
Code required for reproduction
No changes other than step 3. in "Steps for reproduction" were done to the notebook.
Confirmations
- I searched the latest documentation thoroughly before posting.
- I searched previous Issues and Discussions, I am certain this has not been raised before.