Hello! Could someone advise me on the correct way to run a workflow locally for which I've created a launch plan? This is with Python 3.10 in my own Conda env, using the CLI. Currently I'm trying

pyflyte --verbose run workflows/computer_vision/object_detector_training.py object_detector_launchplan

where object_detector_training.py contains my workflow and the launch plan, and

object_detector_launchplan

is a variable used like

object_detector_launchplan = LaunchPlan.create(...

. When I run the above command I get

Error: Missing option '--model_parameters'

, but I'm specifying

model_parameters

in the launch plan with

default_inputs={

"model_parameters": FasterRCNNParameters(...

If iam not mistaken, it is not meant to run a launchplan like this. Launchplan should be rather registered with pyflyte register. When using pyflyte run you specify the name of the workflow (inside the py file, @workflow decorator) after the python file that includes the workflow. pyflyte run file.py workflow

I previously had a Bash script set up that specified all the parameters for the workflow at the CLI using

pyflyte run

, is there a way to utilize launch plans for something like that? Or is scripting it out either with a Python wrapper or Bash the way to go?

pyflyte run won’t register launch plan by default unless you use

--copy-all

, which will register all the flyte entities (task, wf, LP) in the file.

I am not aware of an option to utilize launch plan parameters with pyflyte run. I think you have to pass them as cli parameters

Thanks!