Hi all, quick question: what is the correct way to...
# ask-the-community
t
Hi all, quick question: what is the correct way to pass a raw file or a pandas.DataFrame to a task as input using
FlyteRemote.execute
?
I was trying
Copy code
remote.execute(mytask, {"df": df})
but I get
ValueError: Failed to find a handler for , protocol gcs, fmt ||
(my FlyteRemote is configured with data_upload_location pointing to gcs)
k
It should work once you configure gcs correctly
Ohh what version of flytekit is this
It seems there is not gcs persistence plugin available
t
I have flytekit 1.5.0 (tried with 1.4.2 too, same result). How do I verify the gcs plugin? I see that gcsfs is installed and I can
fsspec.open()
files with the same prefix as I set as data_upload_location.
I can also download the results of executions (
.outputs.get('o0').open(pd.DataFrame)
), so the gcs libs and permissions seem to be ok.
oh, I figured out what was missing:
.execute(task, {"df": StructuredDataset(dataframe=df)})
works
k
Hmm, that should not be needed, maybe file a bug
102 Views