Hi all, What is the intended way to execute a remote launch plan using an input that exists locally? I am running into issues doing something like

pyflyte run remote-workflow workflow.simple_input_flow --input=input.txt

The issue is that flytekit doesn't upload

input.txt

to s3 the way it would if I were creating the workflow via

pyflyte run --remote workflow.py simple_input_flow --input=input.txt

In the former example flytekit will use a path like

/tmp/flyte-3x5f57__/raw/6b1dee242a380ced7fe1e44af4f3a36b/input.txt

to refer to the file, which will cause the task to fail. In contrast, the latter case correctly uploads the file to s3 and the file will be referenced by a path like

<s3://noahfly37041-flyte-data/flytesnacks/development/WTIDKJAQCPBKDYDRFRUHFJPKBM======/input.txt>

I'm not sure if I'm going about this in the wrong way or if this is a bug that impacts my specific use case

interesting. This is a bug. Can you please file it. The problem is type guessing may be little awkward. Also if you can offer a fix that would be awesome. Can you also for me check what the type of the param is.

pyflyte run remote-workflow workflow.simple_input_flow --input=input.txt --help

Can take a look later, working on other issues

@freezing-airport-6809 https://github.com/flyteorg/flytekit/pull/2222 should fix the issue. could you review the changes and confirm if they look correct to you?

Looks good will take a deeper look, did you try regular run and also remote-task etc just to ensure

Thanks for the quick turnaround, I just tried this using the fix branch and my use case works now

@freezing-airport-6809 yes, remote-task, remote-workflow and regular run work.

awesome merge it