I changed my definitions to this:

@asset
def train_dataset():
    train_data : str = "data/train.csv"
    return read_data(train_data)

@asset
def test_dataset():
    test_data : str = "data/test.csv"
    return read_data(test_data)

input_datasets = [train_dataset, test_dataset]

encoder_op = define_dagstermill_op(
    name="encoder_op",
    notebook_path=file_relative_path(__file__, "../notebooks/encoder.ipynb"),
    output_notebook_name="output_encoder",
    outs={"encoders": Out(dict)},
    ins={"df_train": In(pd.DataFrame), "df_test": In(pd.DataFrame)}
)

@graph(out = {'encoders': GraphOut()})
def encoder_graph(df_train, df_test):
    encoders = encoder_op(df_train, df_test)
    return encoders

encoder_asset = AssetsDefinition.from_graph(encoder_graph,
    keys_by_input_name={"df_train": AssetKey("train_dataset"), "df_test": AssetKey("test_dataset")},
    keys_by_output_name={"encoders": AssetKey("encoders_asset")}
)

which I think reflects the structure better but i still get the same error

DagstermillError: dagstermill assets do not currently support dagstermill.yield_result

which means that my

dagstermill_op

is getting converted to a

dagstermill_asset

which is not what I want. What am I doing wrong?