Skip to main content

Cirun configuration

File: .cirun.yml

API documentation for Cirun.yml file.

Runners: runners​

It containers the spec for the runners that will be spun on each commit.

runners:
  - name: "gpu-runner"
    cloud: "gcp"
    gpu: "nvidia-tesla-t4"
    instance_type: "n1-standard-1"
    machine_image: "ubuntu-minimal-2004-lts"
    preemptible: true
    labels:
      - "cirun-gpu-runner"

Name: name​

Name of the runner.

Cloud: cloud​

Name of the cloud provider, here are the possible options:

  • Google Cloud Platform
cloud: "gcp"
  • Amazon Web Services
cloud: "aws"
  • DigitalOcean
cloud: "do"
  • Azure
cloud: "azure"
  • Openstack
cloud: "openstack"
  • Oracle
cloud: "oracle"

GPU: gpu​

Name of the GPU if runner is required with a GPU. See Cloud provider's documentation for gpu names.

gpu: "nvidia-tesla-t4"
note

This flag also installs nvidia-drivers in the given VM

note

For using GPUs on AWS, we recommend using NVIDIA Deep Learning AMI from AWS Marketplace. See the Custom Images section for the documentation on how to use that. Also note this flag is not required for NVIDIA Deep Learning AMI image on AWS.

Instance Type: instance_type​

Name of the instance depending on the cloud provider.

instance_type: "n1-standard-1"  # For GCP
instance_type: "t3.nano"  # For AWS
instance_type: "s-1vcpu-1gb"  # For DigitalOcean
instance_type: "Standard_DS1_v2"  # For Azure
instance_type: "m1.small"  # For Openstack

Machine Image: machine_image​

The OS image for the runner.

GCP​

It can be seen from "gcloud compute images list"

machine_image: "ubuntu-minimal-2004-lts"

AWS​

Amazon Machine image ID

machine_image: "ami-06fd8a495a537da8b"

DigitalOcean​

Slug of the machine image, can be obtained from DigitalOcean CLI via: doctl compute image list --public

machine_image: "ubuntu-20-04-x64"

Azure​

Machine image takes 4 parameters(publisher, offer, sku, version) for available images on Azure. For custom images, it takes only one parameter(id). Images can be obtained from Azure CLI via: az vm image list --output table

machine_image:
  publisher: "Canonical"
  offer: "UbuntuServer"
  sku: "18.04-LTS"
  version: "latest"

Openstack​

In OpenStack cloud by default you are provided with cirros image, you can also build your custom image and use them. You can see all your images in the dashboard or can also use the OpenStack CLI command openstack image list.

machine_image: "cirros-0.5.2-x86_64-disk"

Preemptible: preemptible​

Option to chose low cost instances, also knows as preemptible in GCP and Spot in AWS, default is false. This is not applicable for DigitalOcean and Openstack as they don't support preemptible instances.

preemptible: true

Region: region​

Region to spin up runners in your cloud account.

region: "us-central1-b"  # Example for GCP

Labels: labels​

Add label(s) to your runner definition, so that you can add the same or subset of these labels in the "runs-on" param in .github/workflows/.yml So that this runner is created for running the workflow.

labels:
  - "cirun-runner"
  - "cirun-gpu-runner"
Labels

Labels must start with prefix cirun-