Locally-runnable DAG breaks down for complex examples

[elliotgunton]

Pre-bug-report checklist

1. This bug can be reproduced using pure Argo YAML

• Yes 👉 Please report a bug to the Argo Workflows GitHub 👈
• No

If yes, it is more likely to be an Argo bug unrelated to Hera. Please double check before submitting an issue to Hera.

2. This bug occurs in Hera when...

• exporting to YAML
• submitting to Argo
• running on Argo with the Hera runner
• other: running new dag decorator function locally

Bug report

Describe the bug
A clear and concise description of what the bug is:

When running locally, parameters/artifacts won't be loaded via Pydantic deserialization, meaning I cannot run locally for non-POD parameter types.

Error log if applicable:

pydantic_core._pydantic_core.ValidationError: 2 validation errors for FeatureScalingInput
X_train
  Input should be a valid dictionary [type=dict_type, input_value='{...}', input_type=str]

To Reproduce
Full Hera code to reproduce the bug:
Previous task output:

class DatasetsOutput(Output):
    X_train: Annotated[str, Artifact(name="X_train", archive=NoneArchiveStrategy())]
    X_test: Annotated[str, Artifact(name="X_test", archive=NoneArchiveStrategy())]
    y_train: Annotated[str, Artifact(name="y_train", archive=NoneArchiveStrategy())]
    y_test: Annotated[str, Artifact(name="y_test", archive=NoneArchiveStrategy())]

Task input:

class FeatureScalingInput(Input):
    X_train: Annotated[dict, Artifact(name="X_train", loader=ArtifactLoader.json)]
    X_test: Annotated[dict, Artifact(name="X_test", loader=ArtifactLoader.json)]

Dag:

@w.dag()
def run_training(_: Input):
    datasets = load_and_split_dataset(LoadDatasetInput())
    scaling = feature_scaling(
        FeatureScalingInput(X_train=datasets.X_train, X_test=datasets.X_test)
    )

Expected behavior
A clear and concise description of what you expected to happen:

To be able to run locally. Example works on-cluster

Environment

• Hera Version: 5.16.2
• Python Version: 3.12.4
• Argo Version: 3.5.10

Additional context
Add any other context about the problem here.

[elliotgunton]

Related/higher priority - #1223