Set up RunAI#

The RunAI integration supports NCCL. For direct Kubernetes execution of Dynamo or NCCL, use Run on Kubernetes.

Configure your RunAI account and project in the system TOML:

name = "runai-cluster"
scheduler = "runai"

install_path = "./install"
output_path = "./results"

base_url = "http://runai.example.com"       # The URL of your RunAI system, typically the same as used for the web interface.
verify_certs = true                        # Applies to HTTPS requests and training log connections.
user_email = "your_email"              # The email address used to log into the RunAI system.
app_id = "your_app_id"                      # Obtained by creating an application in the RunAI web interface.
app_secret = ""                             # Fill in the secret obtained with the app_id.
project_id = "your_project_id"              # Project ID assigned or created in the RunAI system (usually an integer).
cluster_id = "your_cluster_id"              # Cluster ID in UUID format (e.g., a69928cc-ccaa-48be-bda9-482440f4d855).

  • After logging into the RunAI web interface, navigate to Access → Applications and create a new application to obtain app_id and app_secret

  • Use your assigned project and cluster IDs. Contact your administrator if they are not available

  • All other fields follow the same semantics as in the Slurm system schema (e.g., install_path, output_path)

RunAI HTTP requests use a 30-second timeout. The optional verify_certs setting defaults to true and controls server TLS certificate verification for both HTTPS requests and training log connections. Set it to false to disable verification. For training log connections with a private certificate authority, install its CA certificate in the system trust store.