New RunLLM response generated: >>>*# Question* Hello! I am trying to use a remote workflow and execute it like this:

data = {
        'column1': np.random.randint(0, 100, size=10),
        'column2': np.random.rand(10),
        'column3': np.random.choice(['A', 'B', 'C', 'D'], size=10)
    }

    df = pd.DataFrame(data)

    inputs = {
        "data_source_name": "dashboards-test",
        "data_source_description": "test",
        "df_input": df
    }
    workflow_name = (
        "workflowname"
    )
    remote = FlyteRemote(config=Config.auto())
    workflow = remote.fetch_workflow(
        project="project",
        domain="production",
        name=workflow_name,
    )
    execution = remote.execute(workflow, inputs=inputs)

    execution = remote.wait(execution)

But it fails on this line

remote.execute(workflow, inputs=inputs)

with the error:

ValueError: Error encountered while executing 'load_data':
  Failed to find a handler for <class 
'flytekit.types.structured.structured_dataset.StructuredDataset'>, protocol 
[flyte], fmt ['']

I am assuming that they problem is related to the dataframe but in the workflow's code I saw that it is also being passed a dataframe without any problem. Any suggestion? # Answer The error you're encountering is due to the

pd.DataFrame

not being correctly wrapped as a

StructuredDataset

when using

FlyteRemote.execute

. According to this discussion, you need to wrap the dataframe with

StructuredDataset

before passing it as an input. Here's how you can modify your code:

from flytekit import FlyteRemote, Config
from flytekit.types.structured import StructuredDataset
import pandas as pd
import numpy as np

data = {
    'column1': np.random.randint(0, 100, size=10),
    'column2': np.random.rand(10),
    'column3': np.random.choice(['A', 'B', 'C', 'D'], size=10)
}

df = pd.DataFrame(data)

inputs = {
    "data_source_name": "dashboards-test",
    "data_source_description": "test",
    "df_input": StructuredDataset(dataframe=df)
}

workflow_name = "workflowname"
remote = FlyteRemote(config=Config.auto())
workflow = remote.fetch_workflow(
    project="project",
    domain="production",
    name=workflow_name,
)
execution = remote.execute(workflow, inputs=inputs)
execution = remote.wait(execution)

This should resolve the issue by ensuring the dataframe is correctly wrapped as a

StructuredDataset

. Would you like more detail on this? Sources: - Hi all quick question what is the correct way to pass a raw Slack Link https://.slack.com/archives/CP2HDHKE1/p1727366866.522569 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.