Unable to plug-in default values to Pipeline

[swiftdiaries]

I tried creating a custom pipeline but I'm ran into this error while compiling it.
ValueError: Default values of argument has to be type dsl.PipelineParam or its child.

@dsl.pipeline(
    name='Simple Pipeline for test',
    description='A simple pipeline for testing'
)
def run_step(project='some value'):

A workaround was to create a PipelineParam and then pass it.
project = dsl.PipelineParam(name='run-step-parameter')

But I'm unable to set default values to the parameter.
The default examples from here don't compile as well. I'm getting the same error.

[gaoning777]

Which DSL Compiler version did you install?
It might be that your compiler is outdated.

[qimingj]

The change to support direct values were checked in at 0b7120c#diff-618bb1fbca23caf0b3edc26f1b38caa8. Latest release of kfp SDK should have it. https://github.com/kubeflow/pipelines/releases

Make sure you install it with "--upgrade" flag.

[swiftdiaries]

Thank you for the reply.
I installed with the --upgrade flag, I tried it with a clean environment and a fresh install.
It still gives out the same error.

[Ark-kun]

Thank you for the reply.
I installed with the --upgrade flag, I tried it with a clean environment and a fresh install.
It still gives out the same error.

Please try pip3 uninstall kfp and use pip3 to install it again. (You can also use python3 -m pip install to be extra sure that the env is the same) Also make sure your python and pip use the same virtual environments. I've experienced similar issues when I installed the package using pip (instead of pip3) which installed it into the python2 package location.

ValueError: Default values of argument has to be type dsl.PipelineParam or its child.

This error was removed from the code base several months ago.

[swiftdiaries]

Thanks for the reply :) That worked. I mixed up my environments.
Closing this issue now.

[Ark-kun]

/assign @Ark-kun