Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

:wave: does anyone know of any examples from docs or tutorials of launching a workflow from another workflow in an async way where i just want to launch the other workflow but not have it block or hold up my main workflow?

i have a feature importance workflow thats a little flakey and can oom and fail sometimes but i want it to be launched from my training workflow but thats all - just launch it and whatever happens happens - dont want it to impact my training workflow outcome at all.

chatgpt had some suggestions but i think is one or two hallucinations in there : <https://chatgpt.com/share/e/67813421-b258-800e-90d2-c146b389b092>

runllm suggested this actually:
```from flytekit import task, workflow
from flytekit.remote import FlyteRemote
from flytekit.configuration import Config

@task
def launch_subworkflow_async():
    remote = FlyteRemote(config=Config.auto())
    launch_plan = remote.fetch_launch_plan(name="subworkflow_launch_plan")
    execution = remote.execute(launch_plan, inputs={"num_samples": 10})
    # Do not wait for the execution to complete

@workflow
def parent_workflow() -&gt; str:
    launch_subworkflow_async()
    return "Subworkflow launched asynchronously"```

`@eager` would let you run an async task without blocking the main workflow
<https://docs.flyte.org/en/latest/user_guide/advanced_composition/eager_workflows.html>

This is beta but making strides to GA soon

oh interesting - will read up and see if we on a version that has it - thanks!

sure. I think it has been available at least since `1.10`

We have used flyte remote successfully but eager looks interesting. 

eager is getting a rewrite in 1.15, still in beta but will be backwards incompatible with the experimental syntax in &lt;=1.14

but actually the original question… i don’t think you need it.

why not just run the other workflow and not consume any outputs?  you can set failure policy on the parent workflow if you need

yeah am trying this approach where i just launch the other workflow. <https://flyte-org.slack.com/archives/CP2HDHKE1/p1736520927900819?thread_ts=1736520647.385129&amp;cid=CP2HDHKE1>

hopefully thats all i need for what im doing here.

sounds good.  you may need `@workflow(failure_policy=WorkflowFailurePolicy.FAIL_AFTER_EXECUTABLE_NODES_COMPLETE)`