Hi everybody! I'm running an example pipeline from a web UI. I notice very strange behavior. After the first local run, I can create a NEW RUN and pass initialization parameters there, but after a successful run, I lose the ability to create new runs with passing new parameters there.

As far as I understand, when I first start the server, the clearml can read the configuration parameters, but after the remote launch, it apparently loses this ability. Is there a possibility that the feedback from the server is broken during the remote start, or something should be corrected in the configuration file.

from clearml import PipelineDecorator


@PipelineDecorator.component(cache=True, execution_queue="default")
def step(size: int):
    import numpy as np
    return np.random.random(size=size)


@PipelineDecorator.pipeline(
    name='ingest',
    project='data processing',
    version='0.1'
)
def pipeline_logic(do_stuff: bool):
    if do_stuff:
        return step(size=42)


if __name__ == '__main__':
    # run the pipeline on the current machine, for local debugging
    # for scale-out, comment-out the following line (Make sure a
    # 'services' queue is available and serviced by a ClearML agent
    # running either in services mode or through K8S/Autoscaler)
    PipelineDecorator.run_locally()

    pipeline_logic(do_stuff=True)