Command Line Interface#
kfp-toolbox
command can be used.
Global options#
Help:
kfp-toolbox --help
Show the installed version:
kfp-toolbox --version
kfp-toolbox submit
#
submit
subcommand submits pipeline jobs to the appropriate environment.
Vertex AI Pipelines#
To submit a job to Vertex AI Pipelines, a pipeline package must be created using the V2 compiler.
dsl-compile-v2 --py /path/to/pipeline.py --function pipeline \
--output ./pipeline.json
If an --endpoint
option is not specified, the job is submitted to Vertex AI Pipelines.
kfp-toolbox submit -f ./pipeline.json
Kubeflow Pipelines#
To submit a job to Kubeflow Pipelines, a pipeline package must be created using the V1 compiler with V2_COMPATIBLE
mode.
dsl-compile --py /path/to/pipeline.py --function pipeline \
--output ./pipeline.yaml \
--mode V2_COMPATIBLE
If an --endpoint
option is specified, the job is submitted to Kubeflow Pipelines pointed to by that endpoint.
kfp-toolbox submit -f ./pipeline.yaml --endpoint http://localhost:8080
Pipeline parameters#
Pipeline parameters can be specified after a double dash (--
).
kfp-toolbox submit -f ./pipeline.json -- \
--parameter-name value \
--another-parameter-name another_value
Pipeline parameters defined in the pipeline package can be viewed by specifying --help
option.
kfp-toolbox submit -f ./pipeline.json --help