How can I use multiple GPUs on WSL2?

windows
1

Hello,

I try to use multiple GPUs (RTX 2080Ti *2) with torch.distributed and pytorch-lightning on WSL2 (windows subsystem for linux).
But I receiving following error:

NCCL error in: /pytorch/torch/lib/c10d/ProcessGroupNCCL.cpp:518, unhandled system error, NCCL version 2.4.8

I can run same process using single GPU.

Thanks for your great help.

2

Is SLI enabled? What the output of nvidia-smi?

3

Thank you for your reply.
Since I use CUDA on WSL (https://docs.nvidia.com/cuda/wsl-user-guide/index.html#getting-started), Ubuntu (on WSL) cannot use the “nvidia-smi” utility.
nvidia-smi on windows shell output following:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.41       Driver Version: 455.41       CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208... WDDM  | 00000000:1A:00.0  On |                  N/A |
| 41%   50C    P0    59W / 250W |    662MiB / 11264MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce RTX 208... WDDM  | 00000000:68:00.0 Off |                  N/A |
| 27%   35C    P8    20W / 250W |    216MiB / 11264MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |

Does this output mean SLI is enabled?

4

I have the same question, it would be great if we could have the answer for this?

5

@Naruki_Ichihara did you ever figure this out?