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

Flyte

Hi!

I've been having some issues with flyte not respecting types of data between tasks.

<https://github.com/flyteorg/flyte/issues/4740>
<https://github.com/flyteorg/flyte/issues/4505>

Its making it so I need to have hard casting between task and introducing weird bugs. Is this something I'm missing or i a known issue?

Cc <@UNR3C6Y4T> / <@UNW4VP36V> the problem of protobuf struct

But this seems odd to be up converting to string

dict[int, float] gets represented not as a protobuf struct in the backend but as a literal map.  The flyte <https://github.com/flyteorg/flyte/blob/5e2224af6c2a8c395751abcf765663ce7e0703de/flyteidl/protos/flyteidl/core/literals.proto#L123|literal map> key types are always str

you can try leaving off the types - a pure `dict` will trigger the proto struct transformer.

good, and what about the int being casted to flat? Need to make that one a dataclass too?

When using `dict` instead of `Dict[int, str]` it still gets casted to str <@UNR3C6Y4T>

```from flytekit import task, workflow

@task
def create_dict() -> dict:
    sample_dict = {1: 1.1, 2: 2.2, 3: 3.3}
    print(sample_dict.keys())
    return sample_dict

@task
def process_dict(input_dict: dict):
    print(input_dict.keys())

@workflow
def workflow() -> None:
    dict_created = create_dict()
    process_dict(input_dict=dict_created)```
The second print will still have str keys :disappointed: