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

Flyte

When using Flyte remote, is there a way to convert inputs/outputs back to the original types if I have access to the code?
I.e. if I define a dataclass `TrainingArgs` and use that as an input to a task and then fetch an execution of that task using remote, I'll get a `TrainingargsSchema` object which is generated from the protobuf I guess. Now I'd like to convert that back to a `TrainingArgs` instance. Is this possible?

`FlyteWorkflowExecution` used to have `raw_outputs` but I don’t find it in the recent flytekit versions. <@U0265RTUJ5B>, any idea why this has been removed?

For the time being, you can convert the schema to dict `<http://dataclass_schema.to|dataclass_schema.to>_dict()` or extract the fields directly from the schema `dataclass_schema.epochs`, and create a dataclass.

You can simply type - execution.inputs and this returns a resolver

The resolver can return raw, guessed or you can pass it a hint

Thanks <@U01J90KBSU9> and <@UNZB4NW3S>, using a type hint works and will return an instance of `TrainingArgs`
```execution.inputs.get("training_args", as_type=TrainingArgs)```