Nvidia-smi shows 0MB GPU memory utilization for docker processes

Hi,

nvidia-smi shows 0MB as GPU memory utilization for all processes running in docker containers:

+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.28.03              Driver Version: 560.28.03      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 3060        Off |   00000000:01:00.0  On |                  N/A |
| 53%   46C    P0             27W /  170W |    2623MiB /  12288MiB |      4%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
                                                                                         
+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A      1991      G   /usr/lib/xorg/Xorg                            292MiB |
|    0   N/A  N/A      2140      G   /usr/bin/gnome-shell                          212MiB |
|    0   N/A  N/A      4354      G   ...seed-version=20240905-050113.669000        272MiB |
|    0   N/A  N/A      4964      G   ...yOnDemand --variations-seed-version         52MiB |
|    0   N/A  N/A      9484      C   python3                                         0MiB |
|    0   N/A  N/A      9491      C   python3                                         0MiB |
|    0   N/A  N/A      9497      C   python3                                         0MiB |
|    0   N/A  N/A      9525      C   python3                                         0MiB |
+-----------------------------------------------------------------------------------------+

Of course, these docker processes use GPU and are therefore listed by nvidia-smi. Note that total memory utilizations 2623MiB seems to show a correct (or at least probable) value.

System details:

$ uname -a
Linux bluewaters 6.1.0-1028-oem #28-Ubuntu SMP PREEMPT_DYNAMIC Wed Dec 13 15:38:49 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

Looks like a bug. How can I fix it?