mjx.put_data() and mjx.put_model() takes lot of time to resolve about mujoco HOT 2 OPEN

Hi @AnxSnip , see here for more info about the performance characteristics of MJX: https://mujoco.readthedocs.io/en/stable/mjx.html#mjx-the-sharp-bits

Seems like you are trying to simulate a bunch of boxes? Try to fiddle with some of these parameters (NB: max contacts, the solver iters):

The timestep seems really high, but I notice you use RK4. Consider lowering the timestep and using Newton for MJX. Hope that helps.

from mujoco.

Hello @btaba, thank you for your quick response.
I am currently trying to do some supervised learning on the stability of a bunch of boxes in a container under acceleration force (as if they were in a truck)
My goal would be to be automatically load a setup from my database to mujoco and then get the state before and after the simulation to compute the stability of the model. The boxes will be under force for around 10 to 30s. But I need it to be able to do fast simulation.
The simulation with this xml takes 10s to complete, but I wanted it to be faster so I looked into MJX but even after 30min it is still not finishing.
I choose these parameters because it seemed to me that they were close to what I expected to happen in real life. But because it is the first time I use physics simulation, I don't really know the meaning of each parameter
I tried changing the values of the parameters you provided but It behave weirdly: some blocks are flying, or exploding without any apparent reason.
If you have any lead or reading that would help me, it will be really kind.
Thank you!

from mujoco.