Hello. Does anyone have experience using CUDA in WSL 2 on Windows 11?
Is it normal to receive such messages?
~/.local/bin$ python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
2022-06-15 12:26:38.641299: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:38.663494: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:38.663891: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:38.664215: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2022-06-15 12:26:38.665784: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:38.666145: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:38.666447: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:39.086758: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:39.087183: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:39.087209: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1616] Could not identify NUMA node of platform GPU id 0, defaulting to 0. Your kernel may not have been built with NUMA support.
2022-06-15 12:26:39.087578: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:961] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2022-06-15 12:26:39.087640: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1532] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 1629 MB memory: -> device: 0, name: NVIDIA GeForce RTX 3050 Laptop GPU, pci bus id: 0000:01:00.0, compute capability: 8.6
tf.Tensor(49.4563, shape=(), dtype=float32)
Also be sure that your tensorflow is compatible with your CUDA.
That is why I didn’t intall the last CUDA just 11.1 and 11.2
to be sure they work almost with everything unless you really need something
new not implemented in earlier version.
Nvidia Guys: include the PATH story to your manual
Good afternoon. Thanks for the detailed answer. Based on your experience, I need to install an earlier version of CUDA. I understand correctly that it is also necessary to install an earlier version of the nvidia studio drivers?
No, NO, No, Leave the drivers they are installed from windows and now are included in microsoft update.
DRIVER IS NOT THE PROBLEM WITH WSL . NVDIA DRIVERS ARE INCLUDED
in WINDOWS UPDATE
this question already answered many times here.
ALL you need to follow the manual to install CUDA TOOLS KIT
I installed this more than one year ago, the important thing DON’T install
Linux Drivers they will not work and shadow the correct windows drivers.
Don’t Install DRIVERS, Your Correct question should be
How to install nvcc? Nvidia compiler
The Normal Way in the manual DON’T INSTALL ANY LINUX drivers
then once you installed nvcc check it with
nvcc -V
if it works then everything shall be ok otherwise never any CUDA will
compile.
IF you need more nvc++, THEN INSTALLL HPC WITHOUT ANY LINUX DRIVERS
Good Luck
Was this ever resolved? I also get the same errors. Using latest drivers and CUDA 11.7. TF2 appears to be working fine with GPU accel, but getting that annoying warning.
Get:1 file:/var/cuda-repo-ubuntu2004-11-7-local InRelease [1575 B]
Get:1 file:/var/cuda-repo-ubuntu2004-11-7-local InRelease [1575 B]
Err:1 file:/var/cuda-repo-ubuntu2004-11-7-local InRelease
The following signatures couldn't be verified because the public key is not available: NO_PUBKEY 5AAE466D15CCF53C
Hit:2 http://archive.ubuntu.com/ubuntu focal InRelease
Hit:3 http://archive.ubuntu.com/ubuntu focal-updates InRelease
Hit:4 http://security.ubuntu.com/ubuntu focal-security InRelease
Hit:5 https://download.docker.com/linux/ubuntu focal InRelease
Hit:6 http://archive.ubuntu.com/ubuntu focal-backports InRelease
Hit:7 https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64 InRelease
Reading package lists... Done
W: GPG error: file:/var/cuda-repo-ubuntu2004-11-7-local InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY 5AAE466D15CCF53C
E: The repository 'file:/var/cuda-repo-ubuntu2004-11-7-local InRelease' is not signed.
N: Updating from such a repository can't be done securely, and is therefore disabled by default.
N: See apt-secure(8) manpage for repository creation and user configuration details.
I use WSL2 Ubuntu-20.04 Kernel version: 5.10.102.1-microsoft-standard-WSL2
$ sudo docker run --gpus all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -gpu -benchmark
docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].
ERRO[0000] error waiting for container: context canceled
I tried following this steps from the link you provided, which is similar to comment. But, I still get the warnings. Did you do anything specifically different?