[Question] I got memory issues with direct RL env.

[sizsJEon]

[INFO]: Starting the simulation. This may take a few seconds. Please wait...
2024-12-04 09:23:08 [27,397ms] [Error] [rtx.neuraylib.plugin] [MDLC:COMPILER]   1.0   MDLC   comp error: http://omniverse-content-production.s3-us-west-2.amazonaws.com/Assets/Isaac/4.1/NVIDIA/Materials/Base/Architecture/Shingles_01.mdl?watch=00007efec5bf1a20(1,1): C100 "mdl" expected
2024-12-04 09:23:08 [27,397ms] [Error] [rtx.neuraylib.plugin] [MDLC:COMPILER]   1.0   MDLC   comp error: http://omniverse-content-production.s3-us-west-2.amazonaws.com/Assets/Isaac/4.1/NVIDIA/Materials/Base/Architecture/Shingles_01.mdl?watch=00007efec5bf1a20(17,4): C100 IDENT expected
2024-12-04 09:23:08 [27,397ms] [Warning] [rtx.neuraylib.plugin] Loading MdlModule to DB (rtx_scope0) failed: ::Z6Ahttp_3A_2F::Z58omniverse_2Dcontent_2Dproduction_2Es3_2Dus_2Dwest_2D2_2Eamazonaws_2Ecom::Assets::Isaac::ZCD_34_2E1::NVIDIA::Materials::Base::Architecture::Shingles_01
2024-12-04 09:23:08 [27,398ms] [Warning] [rtx.neuraylib.plugin] Loading transaction committed (this thread). MdlModule is NOT in the DB (rtx_scope0): ::Z6Ahttp_3A_2F::Z58omniverse_2Dcontent_2Dproduction_2Es3_2Dus_2Dwest_2D2_2Eamazonaws_2Ecom::Assets::Isaac::ZCD_34_2E1::NVIDIA::Materials::Base::Architecture::Shingles_01
[INFO]: Time taken for simulation start : 64.168856 seconds
[INFO] Event Manager:  <EventManager> contains 2 active terms.

while these error and warnings are rose,
my RAM memory consumption rise to 80GB.

I tried to solve this problems, but currently failing again and again.

any advices will be helpful.

[sizsJEon]

accessing
mdl_path="{NVIDIA_NUCLEUS_DIR}/Materials/Base/Architecture/Shingles_01.mdl",
rise those errors.
it is odd that I can access those address in browser and download it.

but still have cpu-memory leakage.

[sizsJEon]

it seems to be train with Terrain(self.cfg.terrain.terrain_type == "generator") and recording video result memory leakage
(--headless --video --video_length 100 --video_interval 500)
without video recording option, RAM stay in 8GB.
anybody ever experienced a situation like this?

[RandomOakForest]

Thanks for posting this. You are using Assets/Isaac/4.1 when they should be 4.2. Upgrading to Isaac Sim 4.2 should resolve the issue.

[Duuuc]

it seems to be train with Terrain(self.cfg.terrain.terrain_type == "generator") and recording video result memory leakage (--headless --video --video_length 100 --video_interval 500) without video recording option, RAM stay in 8GB. anybody ever experienced a situation like this?

I experienced the same. I also observed it with a manager based env (I ran the command ./isaaclab.sh -p source/standalone/workflows/rsl_rl/train.py --task Isaac-Reach-Franka-v0 --headless --video --video_length 1000)