Using Graphical Processing Units (GPUs)

Tier: Free, Premium, Ultimate Offering: GitLab.com, Self-managed
History
  • Introduced in GitLab Runner 13.9.

GitLab Runner supports the use of Graphical Processing Units (GPUs). The following section describes the required configuration to enable GPUs for various executors.

Shell executor

No runner configuration is needed.

Docker executor

Use the gpus configuration option in the runners.docker section. For example: 

[runners.docker]
    gpus = "all"

Docker Machine executor

See the documentation for the GitLab fork of Docker Machine.

Kubernetes executor

No runner configuration should be needed. Be sure to check that the node selector chooses a node with GPU support.

GitLab Runner has been tested on Amazon Elastic Kubernetes Service with GPU-enabled instances.

Validate that GPUs are enabled

You can use runners with NVIDIA GPUs. For NVIDIA GPUs, one way to ensure that a GPU is enabled for a CI job is to run nvidia-smi at the beginning of the script. For example: 

train:
  script:
    - nvidia-smi

If GPUs are enabled, the output of nvidia-smi displays the available devices. In the following example, a single NVIDIA Tesla P4 is enabled: 

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.51.06    Driver Version: 450.51.06    CUDA Version: 11.0     |
|-------------------------------+----------------------+----------------------+
| 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  Tesla P4            Off  | 00000000:00:04.0 Off |                    0 |
| N/A   43C    P0    22W /  75W |      0MiB /  7611MiB |      3%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

If the hardware does not support a GPU, nvidia-smi should fail either because it’s missing or because it can’t communicate with the driver: 

modprobe: ERROR: could not insert 'nvidia': No such device
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.