I’m following the NVIDIA Container Toolkit user guide one step at a time (Ubuntu 20 LTS). However, I have quickly run into some things that don’t seem right (https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/user-guide.html).
If I execute
docker run --rm --gpus all nvidia/cuda:10.0-base nvidia-smi
I get:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.47.03 Driver Version: 510.47.03 CUDA Version: 11.6 |
...
Notice that it shows CUDA Version 11.6 (the version installed on the host). Shouldn’t it show 10.0?
Also, if I copy-paste from the guide and execute:
sudo docker run --rm --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all nvidia/cuda:10.0-base nvidia-smi
I get exactly the same output (full output below). Shouldn’t it show me that the GPU is disabled (or not visible)?
Thanks!
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.47.03 Driver Version: 510.47.03 CUDA Version: 11.6 |
|-------------------------------+----------------------+----------------------+
| 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 ... Off | 00000000:03:00.0 On | N/A |
| 35% 35C P5 25W / 260W | 666MiB / 11264MiB | 33% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 1640 G 342MiB |
| 0 N/A N/A 3095 G 54MiB |
| 0 N/A N/A 3107 G 36MiB |
| 0 N/A N/A 3311 G 48MiB |
| 0 N/A N/A 3528 G 139MiB |
| 0 N/A N/A 7828 G 40MiB |
+-----------------------------------------------------------------------------+
``