This section provides examples of popular use cases illustrating how to use the Command Line Interface (CLI).
Logging In
Logging In via NVIDIA Run:ai Sign In Page (Web)
You can log in from the UI, if you are using SSO or credentials
Logging In via Terminal (Credentials)
runai login user -u [email protected] -p "password"
Configuration
Setting a Default Project
runai project set "project-name"
Submitting a Workload
Naming a Workload
Use the commands below to provide a name for a workload.
Setting the Workload Name ( my_workload_name)
runai workspace submit my-workload-name -p test -i ubuntu
Setting a Random Name with Prefix (prefix=workload type)
runai workspace submit -p test -i ubuntu
Setting a Random Name with Specific Prefix (prefix determined by flag)
runai workspace submit --prefix-name my-prefix-workload-name -p test -i ubuntu
Labels and Annotations
Labels
runai workspace submit -p test -i ubuntu --label name=value --label name2=value2
Annotations
runai workspace submit -p test -i ubuntu --annotation name=value --annotation name2=value2
Container's Environment Variables
runai workspace submit -p test -i ubuntu -e name=value -e name2=value2
Requests and Limits
runai workspace submit -p "project-name" -i runai.jfrog.io/demo/quickstart-demo --cpu-core-request 0.3 --cpu-core-limit 1 --cpu-memory-request 50M --cpu-memory-limit 1G --gpu-devices-request 1 --gpu-memory-request 1G
Submitting and Attaching to Process
runai workspace submit -p "project-name" -i python --attach -- python3
Submitting a Jupyter Notebook
runai workspace submit --image jupyter/scipy-notebook -p "project-name" --gpu-devices-request 1 --external-url container=8888 --name-prefix jupyter --command -- start-notebook.sh --NotebookApp.base_url='/${RUNAI_PROJECT}/${RUNAI_JOB_NAME}' --NotebookApp.token=''
Submitting Distributed Training Workload with TensorFlow
runai tensorflow submit --workers=5 --no-master -g 1 -i kubeflow/tf-mnist-with-summaries:latest -p "project-name" --command -- python /var/tf_mnist/mnist_with_summaries.py --max_steps 1000000
Submitting a Multi-Pod Workload
runai training submit -i alpine -p test --parallelism 2 --completions 2 -- sleep 100000
Submit and Bash
Submitting a Workload with Bash Command
runai training pytorch submit -p "project-name" -i nvidia/cuda:11.8.0-cudnn8-runtime-ubuntu20.04 -g 1 --workers 3 --command -- bash -c 'trap : TERM INT; sleep infinity & wait'
Bashing into the Workload
runai training pytorch bash pytorch-06027b585626 -p "project-name"
Submitting Distributed Training Workload with MPI
runai mpi submit dist1 --workers=2 -g 1 \
-i runai.jfrog.io/demo/quickstart-distributed:v0.3.0 -e RUNAI_SLEEP_SECS=60 -p "project-name"
Submitting with PVC
New PVC Bounded to the Workspace
New PVCs will be deleted when the workload is deleted
runai workspace submit -i ubuntu --new-pvc claimname=yuval-3,size=10M,path=/tmp/test
New Ephemeral PVC
New ephemeral PVCs will be deleted when the workload is deleted or paused
runai workspace submit -i ubuntu --new-pvc claimname=yuval2,size=10M,path=/tmp/test,ephemeral
Existing PVC
Existing PVCs will not be deleted when the workload is deleted
runai workspace submit -i ubuntu --existing-pvc claimname=test-pvc-2-project-mn2xs,path=/home/test
Master/Worker Configuration
--command flag and -- are set both leader (master) and workers command/arguments
--master-args flag sets the master arguments
--master-command flag sets the master commands with arguments
--master-args and --master-command flags can be set together
Overriding Both the Leader (Master) and Worker Image's Arguments
runai pytorch submit -i ubuntu -- -a argument_a -b argument_b -c
Overriding Both the Leader (master) and Worker Image's Commands with Arguments
runai pytorch submit -i ubuntu --command -- python -m pip install
Overriding Arguments of the Leader (Master) and Worker Image's Arguments with Different Values
runai pytorch submit -i ubuntu --master-args "-a master_arg_a -b master-arg_b'" -- '-a worker_arg_a'
Overriding Command with Arguments of the Leader (Master) and Worker Image's Arguments
runai pytorch submit -i ubuntu --master-command "python_master -m pip install'" --command -- 'python_worker -m pip install'
Listing Objects
Submitting with Clean Pod Policy
Submitting a distributed workload with a policy to clean:
runai mpi submit --clean-pod-policy All -i nvidia/cuda:11.8.0-cudnn8-runtime-ubuntu20.04 -g 1
Listing All Workloads in the User's Scope
runai workload list -A
runai project list --yaml
runai node list --json
CLI Reference
For the full guide of the CLI syntax, see the CLI reference.
