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

Flyte

Hi, I have one question regarding the for loop in workflow. I wanted to have a code like the following
```@task
def example_task(num:int) -&gt; None:
    .....

@workflow
def test_workflow() -&gt; None:
    for i in range(5):
        example_test(num=i)```
It seems one way is that I can use dynamic workflow. However, in dynamic workflow, the tasks in the for loop will be running in parallel. I wanted the tasks can be running in sequence. So do you have a method to realize the for loop in sequence. Thanks!

do you have many tasks in the loop? maybe you can add chaining to your task or add dummy variable that feeds into the next iteration

i dont think workflow will work since during registration time everything needs to be defined

see <https://docs.flyte.org/projects/cookbook/en/latest/auto_examples/control_flow/chain_entities.html#id1|here> for the chaining syntax

If using the chaining syntax, the code will be something like
```@task
def example_task(num:int) -&gt; None:
    .....

@workflow
def test_workflow() -&gt; None:
    example_test_wf1 = example_test(num=0)
    example_test_wf2 = example_test(num=1)
    example_test_wf3 = example_test(num=2)
    example_test_wf4 = example_test(num=3)
    example_test_wf5 = example_test(num=4)

    example_test_wf1 &gt;&gt; example_test_wf2
    example_test_wf2 &gt;&gt; example_test_wf3
    example_test_wf3 &gt;&gt; example_test_wf4
    example_test_wf4 &gt;&gt; example_test_wf5```
The code is kind of redundant and it will be more ugly if I have more tasks in the for loop. I am exploring if there is any more concise way.

maybe dummy output works :thinking_face:
```dep_out = 0
for i in range(5):
  output, dep_out = example_test_wf(..., dep_out=dep_out)```
and use dynamic task
i actually never tried this before so let me know if this works :sweat_smile:

Thanks for the suggestion! However I also wanted to run these tasks in sequence. One issue with dynamic task is that it seems these `example_test_wf` will be running in parallel task as far as I know (please let me know if there is some configuration that can make them running in sequence).

so by adding that dummy out should ensure the sequential operation right, dep_out from previous loop is fed back into the next task
well thats the idea but not sure it will work haha

ah, got it now. thanks! will try it and update the result later here.

it seems the above method works :+1:. Actually there is no need to use dynamic task. A normal `workflow` works well. It's kind of surprising as I thought for loop is not supported in workflow. It seems there are other people have the same impression as me (see the discussion here <https://flyte-org.slack.com/archives/CP2HDHKE1/p1685726254115599?thread_ts=1685464169.404789&amp;cid=CP2HDHKE1>). I guess maybe Flyte official did some improvements there. Btw, I am using the latest Flytekit 1.9.0.

To test the for loop works in workflow, I tried the following, it worked and it was running in parallel, just like dynamic task.
```@task
def example_task(num: int) -&gt; None:
    print(f"current num: {num}")


@workflow
def quick_test_workflow():
    for i in range(6):
        example_task(num=i)```

oh thats good to hear, yeah i think i observed for loop works only if the loop is deterministic at registration time, maybe flyte is just expanding the list directly