You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, I'm looking for a way to increase the rendering and simulation speed.
First, I'm trying to train my policy with sim_dt=0.001 and decimation=10, unlike usual examples where sim_dt=0.005 and decimation=4.
This cause my training time very slow since the physics stepping loop in manager_based_rl_env.py takes longer time.
I observed that the main bottleneck is self.scene.write_data_to_sim() and self.scene.update(). Is there any way that I can make this collection/update time faster?
Also, the self.sim.render() takes 0.01 seconds which indicates that the rendering cannot exceed 100Hz by the render function itself. I think this is too slow. Is there any way that I can reduce this rendering time in exchange of rendering quality? Wondering if there's any solution other than just skipping some frames.
Thanks!
Please refer to the bottom related issues: #1157 #379
The text was updated successfully, but these errors were encountered:
Hi, I'm looking for a way to increase the rendering and simulation speed.
First, I'm trying to train my policy with
sim_dt=0.001
anddecimation=10
, unlike usual examples wheresim_dt=0.005
anddecimation=4
.This cause my training time very slow since the physics stepping loop in
manager_based_rl_env.py
takes longer time.I observed that the main bottleneck is
self.scene.write_data_to_sim()
andself.scene.update()
.Is there any way that I can make this collection/update time faster?
Also, the
self.sim.render()
takes 0.01 seconds which indicates that the rendering cannot exceed 100Hz by the render function itself. I think this is too slow. Is there any way that I can reduce this rendering time in exchange of rendering quality? Wondering if there's any solution other than just skipping some frames.Thanks!
Please refer to the bottom related issues:
#1157
#379
The text was updated successfully, but these errors were encountered: