WSL2 Ubuntu22.04 python3.9 CUDA11.8: no CUDA-capable device is detected while calling init

[LongHZ140516]

I am currently using a WSL2 Ubuntu22.04 python3.9 environment with CUDA 11.8 for Genesis experiments. When I run gs.init(backend=gs.cpu), everything works fine, but the FPS is very low.

Therefore, I would like to use CUDA to accelerate the computation. When I set gs.init(backend=gs.cuda), I encountered the following error.

[Genesis] [15:16:25] [INFO] ╭─────────────────────────────────────────────────────────────────────────────────────╮
[Genesis] [15:16:25] [INFO] │┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉ Genesis ┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉┈┉│
[Genesis] [15:16:25] [INFO] ╰─────────────────────────────────────────────────────────────────────────────────────╯
[Genesis] [15:16:26] [INFO] Running on [NVIDIA GeForce RTX 4080 SUPER] with backend gs.cuda. Device memory: 15.99 GB.
[E 12/21/24 15:16:26.157 71086] [cuda_driver.h:operator()@92] CUDA Error CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected while calling init (cuInit)


Traceback (most recent call last):
  File "/home/serein/code/Genesis/hello_genesis.py", line 21, in <module>
    gs.init(backend=gs.cuda)
  File "/home/serein/code/Genesis/genesis/__init__.py", line 97, in init
    ti.init(arch=TI_ARCH[platform][backend], debug=debug, force_scalarize_matrix=True)
  File "/home/serein/anaconda3/envs/genesis/lib/python3.9/site-packages/taichi/lang/misc.py", line 458, in init
    impl.get_runtime().create_program()
  File "/home/serein/anaconda3/envs/genesis/lib/python3.9/site-packages/taichi/lang/impl.py", line 388, in create_program
    self.prog = _ti_core.Program()
RuntimeError: [cuda_driver.h:operator()@92] CUDA Error CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected while calling init (cuInit)
[Genesis] [15:16:26] [INFO] 💤 Exiting Genesis and caching compiled kernels...

Here is my nvidia-smi:

Sat Dec 21 15:05:51 2024
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 555.42.06              Driver Version: 560.94         CUDA Version: 12.6     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 4080 ...    On  |   00000000:01:00.0  On |                  N/A |
|  0%   40C    P5             18W /  340W |    3144MiB /  16376MiB |     18%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |

and nvcc --version:

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:33:58_PDT_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

Additionally, I checked CUDA using PyTorch, and the result was True. However, I still cannot run successfully with gs.cuda.

I also tried changing my CUDA version to 12.1, but the result remained exactly the same as before, with no changes. The result of running python -m taichi diagnose is as follows:

[Taichi] version 1.7.2, llvm 15.0.4, commit 0131dce9, linux, python 3.9.21

*******************************************
**      Taichi Programming Language      **
*******************************************

Docs:   https://docs.taichi-lang.org/
GitHub: https://github.com/taichi-dev/taichi/
Forum:  https://forum.taichi.graphics/

Taichi system diagnose:

python: 3.9.21 (main, Dec 11 2024, 16:24:11) 
[GCC 11.2.0]
system: linux
executable: /home/serein/anaconda3/envs/genesis/bin/python
platform: Linux-5.15.146.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
architecture: 64bit ELF
uname: uname_result(system='Linux', node='Serein', release='5.15.146.1-microsoft-standard-WSL2', version='#1 SMP Thu Jan 11 04:09:03 UTC 2024', machine='x86_64')
locale: en_US.UTF-8
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 22.04.4 LTS
Release:        22.04
Codename:       jammy



import: <module 'taichi' from '/home/serein/anaconda3/envs/genesis/lib/python3.9/site-packages/taichi/__init__.py'>

cpu: True
metal: False
RHI Error: GLFW Error 65543: GLX: Failed to create context: GLXBadFBConfig
opengl: ERROR Command '['/home/serein/anaconda3/envs/genesis/bin/python', '-c', 'import taichi as ti; print("===="); print(ti.lang.misc.is_arch_supported(ti.opengl), end="")']' died with <Signals.SIGSEGV: 11>.
cuda: True
vulkan: False

`glewinfo` not available: [Errno 2] No such file or directory: 'glewinfo'

Sat Dec 21 14:12:15 2024       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 555.42.06              Driver Version: 560.94         CUDA Version: 12.6     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 4080 ...    On  |   00000000:01:00.0  On |                  N/A |
|  0%   41C    P0             37W /  340W |    3479MiB /  16376MiB |      5%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
                                                                                         
+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

[Taichi] version 1.7.2, llvm 15.0.4, commit 0131dce9, linux, python 3.9.21

[Taichi] version 1.7.2, llvm 15.0.4, commit 0131dce9, linux, python 3.9.21
[Taichi] Starting on arch=x64

RHI Error: GLFW Error 65543: GLX: Failed to create context: GLXBadFBConfig
Taichi OpenGL test failed: Command '['/home/serein/anaconda3/envs/genesis/bin/python', '-c', 'import taichi as ti; ti.init(arch=ti.opengl)']' died with <Signals.SIGSEGV: 11>.
[E 12/21/24 14:12:16.540 35656] [cuda_driver.h:operator()@92] CUDA Error CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected while calling init (cuInit)
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/serein/anaconda3/envs/genesis/lib/python3.9/site-packages/taichi/lang/misc.py", line 458, in init
    impl.get_runtime().create_program()
  File "/home/serein/anaconda3/envs/genesis/lib/python3.9/site-packages/taichi/lang/impl.py", line 388, in create_program
    self.prog = _ti_core.Program()
RuntimeError: [cuda_driver.h:operator()@92] CUDA Error CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected while calling init (cuInit)
Taichi CUDA test failed: Command '['/home/serein/anaconda3/envs/genesis/bin/python', '-c', 'import taichi as ti; ti.init(arch=ti.cuda)']' returned non-zero exit status 1.

[Yzhang367]

Maybe you can try to update your OpenGL or reinstall CUDA in 12.6 version.

[yasuhiroinoue]

I encountered the same issue where the error message "no CUDA-capable device is detected" was displayed. In my case, adding the following line to the ~/.bashrc file resolved the problem:

export LD_LIBRARY_PATH=/usr/lib/wsl/lib:$LD_LIBRARY_PATH