Skip to navigation

runai inference standard describe

View as Markdown

describe an inference workload

runai inference standard describe [WORKLOAD_NAME] [flags]

Examples

# Describe a workload with a default project
runai inference describe <workload-name>
# Describe a workload in a specific project
runai inference describe <workload-name> -p <project-name>
# Describe a workload by UUID
runai inference describe --uuid=<workload-id>
# Describe a workload with specific output format
runai inference describe <workload-name> -o json
# Describe a workload with specific sections
runai inference describe <workload-name> --general --compute --pods --events --networks
# Describe a workload with container details and custom limits
runai inference describe <workload-name> --containers --pod-limit 20 --event-limit 100

Options

--compute Show compute resources information (CPU, GPU, memory, etc.). (default true)
--containers Include container-level details in pod information.
--event-limit int32 Limit the number of events displayed. Use -1 for no limit. (default 50)
--events Show recent events for the resource. (default true)
--general Show general resource information. (default true)
-h, --help help for describe
--networks Show network-related information. (default true)
-o, --output string Output format: table, json, or yaml. Default is table. (default "table")
--pod-limit int32 Limit the number of pods displayed. Use -1 for no limit. (default 10)
--pods Show information about pods associated with the resource. (default true)
-p, --project string Specify the project for the command to use. Defaults to the project set in the context, if any. Use 'runai project set <project>' to set the default.
--sortEvents string Sort events in ascending or descending order. Valid values: asc, desc. (default "asc")
--storage Show storage information (PVCs, ConfigMaps, Secrets, etc.). (default true)

Options inherited from parent commands

--config-file string config file name; can be set by environment variable RUNAI_CLI_CONFIG_FILE (default "config.json")
--config-path string config path; can be set by environment variable RUNAI_CLI_CONFIG_PATH
-d, --debug enable debug mode
-q, --quiet enable quiet mode, suppress all output except error messages
--verbose enable verbose mode

SEE ALSO

  • runai inference standard - Runs a single inference process on one node. Suitable for smaller models or simpler inference tasks.