No GPU for training error

General
1

Hello I am getting this message when training in the model zoo:

  return torch._C._cuda_getDeviceCount() > 0
Using CPU

I believe it’s a Cuda version conflict:

sam@terminus:~/hailo_model_zoo/training/yolov5$ nvidia-smi
Fri May  9 10:41:56 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 575.51.03              Driver Version: 576.28         CUDA Version: 12.9     |
|-----------------------------------------+------------------------+----------------------+
| 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 3060        On  |   00000000:01:00.0  On |                  N/A |
|  0%   37C    P8             19W /  170W |     749MiB /  12288MiB |      4%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+
sam@terminus:~/hailo_model_zoo/training/yolov5$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2020 NVIDIA Corporation
Built on Wed_Jul_22_19:09:09_PDT_2020
Cuda compilation tools, release 11.0, V11.0.221
Build cuda_11.0_bu.TC445_37.28845127_0

I am using WSL2 Ubuntu-20.04 with Cuda 11.0.
Do I also need to install an older driver on the Windows host?

2

The Dataflow Compiler usage on Windows using WSL2, does not currently support GPU. You can find this in the Hailo Dataflow Compiler User Guide.

I recommend using the Hailo AI Software Suite Docker on Ubuntu 20.0.4 or 22.04 directly. This provides the easiest installation and upgrades.

3

Thanks for replying, if I dual boot Linux and Windows instead of using WSL2, will that fix this problem?

4

Yes, we work with dual boot on our machines as well.

5

Do you dual boot off the same SSD or would you recommend separate ones?

6

I have systems configured with both approaches, and both work well.

On the machine with two SSDs, I physically removed the Windows SSD during the initial Linux installation to prevent any changes to its boot sector. I now switch between operating systems using the UEFI/BIOS boot menu. Otherwise, the system will always boot into the last selected OS. I can also remove one of the SSDs and the OS on the other will just boot fine.

On the machine with a single SSD, I rely on the GRUB boot menu to select the operating system at startup.

7

thank you for your help

8

I’m dual booting linux now, I’ll see how the training goes tonight.
btw I noticed this on the model zoo readme, is it out of date?

image
image688×175 6.07 KB