Can it run on a Linux system with A100 GPUs?

[RZFan525]

No description provided.

[xuanlinli17]

Yes. Please follow the instructions in readme and troubleshooting. Though, rendering for the drawer tasks will be slow due to the use of ray tracing.

[RZFan525]

Thank you for your reply. However, I encountered the same error as #7. And, when I install vulkan-utils with sudo apt-get install vulkan-utils, an error appears:
The package vulkan-utils could not be located
I don't have any computers with RTX GPUs, how can I run it?

[RZFan525]

I have followed https://maniskill.readthedocs.io/en/latest/user_guide/getting_started/installation.html#vulkan to add three json files, but it does not work.

[xuanlinli17]

Did you sudo apt update and vulkan-utils is still not found since it's ubuntu 22.04?

Try sudo apt install vulkan-tools

[RZFan525]

Thank you for your reply. I have tried it, and it can be installed successfully. However, the same error has appeared.

And, I found that vulkaninfo works without /usr/share/vulkan/icd.d/nvidia_icd.json, /usr/share/glvnd/egl_vendor.d/10_nvidia.json, and /etc/vulkan/implicit_layer.d/nvidia_layers.json. But, when I follow https://maniskill.readthedocs.io/en/latest/user_guide/getting_started/installation.html#vulkan to manually add these three files, vulkaninfo doesn't work with the error ERROR_OUT_OF_HOST_MEMORY.

Anyway, the following error always appears whether the vulkaninfo can work or not.

[2024-07-28 11:58:15.019] [svulkan2] [error] GLFW error: X11: The DISPLAY environment variable is missing
[2024-07-28 11:58:15.019] [svulkan2] [warning] Continue without GLFW.
Traceback (most recent call last):
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv-OpenVLA/test.py", line 4, in <module>
    env = simpler_env.make('google_robot_pick_coke_can')
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv-OpenVLA/simpler_env/__init__.py", line 78, in make
    env = gym.make(env_name, obs_mode="rgbd", **kwargs)
  File "/cpfs01/user/liupengfei/rzfan/miniconda3/envs/simpler_env/lib/python3.10/site-packages/gymnasium/envs/registration.py", line 802, in make
    env = env_creator(**env_spec_kwargs)
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv/ManiSkill2_real2sim/mani_skill2_real2sim/utils/registration.py", line 92, in make
    env = env_spec.make(**kwargs)
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv/ManiSkill2_real2sim/mani_skill2_real2sim/utils/registration.py", line 34, in make
    return self.cls(**_kwargs)
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv/ManiSkill2_real2sim/mani_skill2_real2sim/envs/custom_scenes/grasp_single_in_scene.py", line 630, in __init__
    super().__init__(**kwargs)
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv/ManiSkill2_real2sim/mani_skill2_real2sim/envs/custom_scenes/grasp_single_in_scene.py", line 540, in __init__
    super().__init__(**kwargs)
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv/ManiSkill2_real2sim/mani_skill2_real2sim/envs/custom_scenes/grasp_single_in_scene.py", line 64, in __init__
    super().__init__(**kwargs)
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv/ManiSkill2_real2sim/mani_skill2_real2sim/envs/custom_scenes/base_env.py", line 134, in __init__
    super().__init__(**kwargs)
  File "/cpfs01/user/liupengfei/rzfan/SimplerEnv/ManiSkill2_real2sim/mani_skill2_real2sim/envs/sapien_env.py", line 107, in __init__
    self._renderer = sapien.SapienRenderer(**renderer_kwargs)
RuntimeError: vk::Instance::enumeratePhysicalDevices: ErrorInitializationFailed

[RZFan525]

I don't know how to run it :(

I have tried three different servers with A100 GPUs which encounter the same error. :(

[xuanlinli17]

Are you setting cuda devices properly? Also ensure that nvidia-driver version is at least above 535. Older nvidia drivers might not work.

You can make a fake display like

tmux new -s 1
sudo X :0 &
[exist tmux ctrl-b]
export DISPLAY=:0

[RZFan525]

Thank you for getting back to me.

The servers I used are in a docker and I changed to another server, which makes it work.

However, I encountered another error which is attributed to the lack of display.

RuntimeError: Create window failed: context is not created with present support

Do you have any suggestions to help me observe the environment and the process of action?

[xuanlinli17]

Inside docker, you might want to port the (fake) display (e.g., sudo X :0 &) in the main bash to the docker container

However, the SIMPLER environments shouldn't create a window unless you are visualizing robots using the utility scripts.

[RZFan525]

I'm new in robotics, so I want to visualize the simulation environment to help me understand deeply. Maybe, it's better to output a video.

[xuanlinli17]

The evaluation videos are automatically saved.

[RZFan525]

Thank you!

I can run the scripts scripts/openvla_bridge.sh, but it suddenly reports an error after running for a while.

[xuanlinli17]

If you consecutively create 2 environments in ipython, does it still report an error?

[RZFan525]

When I create 2 environments, it can work. But there is a warning:

[2024-07-31 03:20:06.870] [svulkan2] [warning] A second renderer will share the same internal context with the first one. Arguments passed to constructor will be ignored.

[RZFan525]

i don't know why. But, I also try SimplerEnv-OpenVLA/scripts/openvla_drawer_variant_agg.sh It's successful to output the average success

Thank you!

I can run the scripts scripts/openvla_bridge.sh, but it suddenly reports an error after running for a while.

[RZFan525]

I find that the error appears when the obj_episode_id is 11 in any scripts that define obj-variation-mode as the episode.

[xuanlinli17]

That's strange; episode 11 doesn't introduce new objects.

[RZFan525]

Could you give me some instructions on how to debug? Thank you very much!!

[xuanlinli17]

I actually don't know... and sorry that I don't have much bandwidth at the moment to look closely.

[RZFan525]

Ok. Thank you for your reply.

[xuanlinli17]

Also you might create fake display like sudo X :0 &; export DISPLAY=:0 or xvfb-run -a {script}, to see if it works.

[RZFan525]

Thank you. After trying this command, I found it cannot work. The error is the same. I don't know why.

[COST-97]

Hello:

I don't know how to run it :(

I have tried three different servers with A100 GPUs which encounter the same error. :(

Same error in A100 GPU.
"libGLX_nvidia.so.0" does not exist in the A100.

Does anyone have an updated solution?
Thanks a lot!

[xuanlinli17]

Hello:

I don't know how to run it :(
I have tried three different servers with A100 GPUs which encounter the same error. :(

Same error in A100 GPU. "libGLX_nvidia.so.0" does not exist in the A100.

Does anyone have an updated solution? Thanks a lot!

Could you try the troubleshooting section in readme?