Overview

This page includes information about the self-hosted runners that are available from NVIDIA and how to use them.

CPU Runners

The CPU labeled runners are backed by various EC2 instances and do not have any GPUs installed.

Label	EC2 Machine Type
linux-amd64-cpu4	m7i.xlarge (specs)
linux-amd64-cpu8	m7i.2xlarge (specs)
linux-amd64-cpu16	m7i.4xlarge (specs)
linux-arm64-cpu4	m7g.xlarge (specs)
linux-arm64-cpu8	m7g.2xlarge (specs)
linux-arm64-cpu16	m7g.4xlarge (specs)

Info

The CPU label names consist of the following components:

linux-amd64-cpu4
^     ^     ^  ^
|     |     |  |
|     |     |  CPU Core Count
|     |     CPU Designator
|     Architecture
Operating System

GPU Runners

The GPU labeled runners have the GPUs specified in the table below installed.

Important!

GPU jobs have two requirements:

1. They must run in a container (i.e. nvidia/cuda:12.0.0-base-ubuntu22.04)
2. They must set the NVIDIA_VISIBLE_DEVICES: ${{ env.NVIDIA_VISIBLE_DEVICES }} container environment variable

If these requirements aren't met, the GitHub Actions job will fail. See the Example section below for an example.

These requirements do not apply to non-NVIDIA GitHub organizations.

Label	GPU	Driver Version	# of GPUs
linux-amd64-gpu-v100-earliest-1	V100	470	1
linux-amd64-gpu-v100-latest-1	V100	535	1
linux-arm64-gpu-a100-latest-1	A100	535	1
linux-amd64-gpu-t4-latest-1	T4	535	1

Info

The GPU label names consist of the following components:

linux-amd64-gpu-t4-latest-1
^     ^     ^   ^  ^      ^
|     |     |   |  |      |
|     |     |   |  |      Number of GPUs Available
|     |     |   |  GPU Driver Version
|     |     |   GPU Type
|     |     GPU Designator
|     Architecture
Operating System

Due to limited GPU capacity and the overhead associated with manually rotating self-hosted runner labels when GPU drivers are updated, there are no driver-specific self-hosted runner labels (e.g. linux-amd64-gpu-t4-535-1).

Instead, the driver-version designators earliest and latest are used. The values of these designators represent the oldest and newest NVIDIA supported drivers at any given time.

The chart above will be kept up-to-date with the corresponding driver versions at any given time.

Supported organizations will be notified whenever these versions are scheduled to be updated.

Example

The code snippet below shows how the labels above may be utilized in a GitHub Action workflow.

name: Test Self Hosted Runners
on: push
jobs:
  job1_cpu:
    runs-on: linux-amd64-cpu8
    steps:
      - name: hello
        run: echo "hello"
  job2_gpu:
    runs-on: linux-amd64-gpu-v100-latest-1
    container: # GPU jobs must run in a container
      image: nvidia/cuda:12.0.0-base-ubuntu22.04
      env:
        NVIDIA_VISIBLE_DEVICES: ${{ env.NVIDIA_VISIBLE_DEVICES }} # GPU jobs must set this container env variable
    steps:
      - name: hello
        run: |
          echo "hello"
          nvidia-smi

For additional details on self-hosted runner usage, see the official GitHub Action documentation page here.