[Bug Report] The training shut down for no reasons with 50 series card

[81578823]

The training shut down for no reasons with 50 series card [bug].

Describe the bug

The training shuts down for unknown reasons.

[INFO]: Simulation is stopped. Shutting down the app.

Steps to reproduce

python scripts/reinforcement_learning/rsl_rl/train.py --task=Isaac-Velocity-Rough-Anymal-B-v0 --headless

and

python scripts/reinforcement_learning/rsl_rl/train.py --task=Isaac-Velocity-Rough-Anymal-C-v0 --headless

System Info

• Isaac Sim Version: 4.5.0
• OS: Ubuntu 20.04
• GPU: RTX5080
• CUDA:12.8
• GPU Driver: 570.144
• torch: 2.7.0

[81578823]

[81578823]

Sometimes, it just gets stuck and does not move on

[RandomOakForest]

Thank you for posting this, the team will review it.

[baijubaiju233]

Can you use the 50-series graphics cards for training? But when you open Isaac Sim 4.5 with a 50-series card, doesn't it blur objects? I think these two aren't compatible. I assumed that 50-series cards could only be used with Isaac Sim 5.0 paired with Isaac Lab 2.2. I’m really sorry for bringing up a topic unrelated to the issue, but I’m eager to know the answer.

[kellyguo11]

We have not been able to reproduce the initial issue, could you give the 5.0 Isaac Sim branch a try and see if you still observe it there?

Isaac Sim 4.5 will be able to run with 50-series cards, but there is a rendering issue that disables the denoiser, so rendered results will look very noisy. This has been fixed in Isaac Sim 5.0.

[81578823]

@kellyguo11 Yes, now I am using 5.1.0 Isaacsim, 2.3.0 Isaaclab, 570 nvidia-driver, however, the interruption still happens without any error information or warning

[81578823]

I have figured it out finally!!! The interruption occurs since I use intel_pstate driver for my cpu(Intel ultra 7-265K), which is the newest CPU hardware having compatibility issues with ubuntu.

The solutions are as follows:

• sudo gedit /etc/default/grub
• Change the line GRUB_CMDLINE_LINUX_DEFAULT="quiet splash" to GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_pstate=disable" and save.
• sudo update-grub
• sudo reboot

After these operations, you can see the cpu driver has changed from intel_pstate to acpi-cpufreq according to cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_driver. The interruption does not occur any more.

ps: Do not forget to set CPU scaling_governor as schedutil, which will be better