Closed rdednl closed 8 months ago
I found that the issue is related to the fact that the SimulationContext
in the environment is not closed properly in the environment, and it's variable SimulationContext._sim_context_initialized
remains True
across instances, which stops it from initializing a new stage after having it cleared.
SimulationContext
provides a clear_instance()
which solves the issue.
Thank you for bringing up the issue and also sending a fix for it. This has been accounted for in the latest release.
Question
Not sure if this is the expected behavior so I'm posting it as a question.
A process which is doing robot learning with multiple sequential training procedures, will create an environment, do whatever it has to do, and then close the environment and launch a new environment. Unfortunately from what I've been trying, this is not currently possible in Orbit, as when doing
env.close()
, and thengym.make()
,Isaac will throw the following error:
[19,623ms] [Error] [carb.physx.python] Replicator already registered for given stage.
Here is a minimum working example to reproduce the error:
This is the stack trace that it produces: