Konduktor Launch Task Yamls

​

Konduktor Launch Task Yaml

​

Schema

name: <string>                    # required

envs:                             # optional
  key: value
workdir: <string>                 # optional
num_nodes: <int>                  # optional; defaults to 1

resources:
  cpus: <int/float/string>        # required; unit is vCPUs
  memory: <int/string>            # required; unit is GiB
  image_id: <string>              # required
  accelerators: <string>          # optional
  labels:                           
    kueue.x-k8s.io/queue-name: <string>  # required
    maxRunDurationSeconds: <string>      # optional; defaults to 604800 (7 days)
  job_config:
    max_restarts: <int>           # optional; defaults to 0
    completions: <int>            # optional; defaults to num_nodes

file_mounts:                      # optional
  /remote/path: ./local/path

run: |                            # optional
  <shell command(s)>

​

Details

• local directory to sync into the remote workdir before running commands

• default and only unit is vCPU
• accepts int, float, or numeric string:
  • 2, 3.1, 0.2, “4.2”, “0.5”, etc (0.5 vCPU → 500m (millicores))
• default and only unit is vCPU
• unsupported formats:
  • “500m”
  • “4 vCPU”
  • anything containing letters

• default and only unit is GiB (Gi)
• accepts int or numeric string:
  • 2, 16, etc (16 → 16Gi)
• unsupported formats:
  • 8.5
  • “8.5”
  • floats
  • “16G”
  • “0.5 T”
  • anything containing letters

• required Kueue queue label so the scheduler knows which queue to place the JobSet in

• expected wall-clock runtime DWS uses for fairness/preemption; omitted values default to 604800s (7 days) which effectively behaves like unlimited on on-prem clusters

• mapping of local files/dirs to remote paths; Konduktor syncs them before launching your job
• Deprecated: syncing via cloud storage buckets/credentials is deprecated and will be removed in a future version of Konduktor