Not sure what to set for "graphics_device_id"

Autonomous Machines Robotics - Isaac Isaac Gym
1

This is with regards to GitHub - NVIDIA-Omniverse/IsaacGymEnvs: Isaac Gym Reinforcement Learning Environments
I am also using IsaacGym Preview 3.
rl_device and sim_device works for me,
but for graphics_device_id I do not have any idea how to go about setting this, so far it’s been defaulting to be cuda:0 GPU in a 4-GPU workstation regardless of what I parse.

I have also tried lspci giving me

d8:00.0 VGA compatible controller: NVIDIA Corporation Device 2206 (rev a1)

and then doing something like

python train.py task=ShadowHand headless=True graphics_device_id=d8:00.0

Would really appreciate some help!

2

Hi @piggy4eva ,

Graphics devices go through Vulkan rather than CUDA. You need to use vulkaninfo to figure out the id for your device. Here’s the note from the README:

graphics_device_id=GRAHPICS_DEVICE_ID - Which Vulkan graphics device ID to use for rendering. Defaults to 0. Note - this may be different from CUDA device ID, and does not follow PyTorch-like device syntax.

Take care,
-Gav