My team has a number of models that train, organized like

myproject.models.alabama

,

myproject.models.california

,

myproject.models.texas

. Each one has the same API. How would you all create tasks/workflows for each of these models, without having to copy/paste the same code for each model? I've tried things like making a function/task factory, but Flyte keeps telling me

ValueError: TaskFunction cannot be a nested/inner or local function.

. Using importlib to import-by-string works okay, but then

pyflyte run --remote

doesn't package the model's code and its dependencies when submitting the workflow to the remote.

which version of flytekit you are using? we add an auto copy feature, and it should copy your code dependencies.

why the need for distinct tasks? can you add a string as an input/output?

1.13.12

but why can’t the single

fetch_training_data

task take a string input?

It certainly can, I'm just struggling to find a way to use that string that works well with both our project's organization and what flyte's expecting.