Common Commands#

Most of mcli is used with an <action> <object> pattern. Here are some common commands for using mcli:

Interacting with clusters#

mcli get clusters

View available clusters

mcli util

View cluster utilization

mcli run -f <your_yaml>

Submits a run with the provided YAML configuration.

mcli get runs

Lists the status of all your submitted runs.

mcli describe run <run_name>

Get detailed information about a run, including the config that was used to launch it.

mcli logs <run_name>

Retrieves the console log of the indicated run.

mcli logs <run_name> --failed

Retrieves the console log of a failed run from the node rank that failed first. This is extremely useful for large multinode runs.

mcli stop run <run_name>

Stops the provided run. The run will be stopped but not deleted from the cluster.

mcli delete run <run_name>

Deletes the provided run name. The run (and its associated logs) will be deleted from the cluster.

mcli create secret <secret-type>

Creates a secret

mcli get secrets

Lists secrets you’ve created

mcli delete secret <secret_name>

Deletes the secret