microsoft
Support y as a single-column pd.DataFrame
When using FLAML(task="classification") as a component model in EconML/DoWhy, sometimes it gets called with y being a pd.DataFrame with a single column, and that throws an error as it only expects a pd.Series or np.ndarray.
Can we please also support dataframes with a single column, in which case we'd take that column as y?
Currently I monkey-patch around it, but it would be nicer if it just handled it.
Happy to do the PR myself if you tell me where in the code the changes should live :)
Great suggestion. These are the relevant places to be changed:
https://github.com/microsoft/FLAML/blob/a64956a7c8f98db6a7de7e8cc5d0ce53e9c86a03/flaml/automl.py#L1052 https://github.com/microsoft/FLAML/blob/a64956a7c8f98db6a7de7e8cc5d0ce53e9c86a03/flaml/automl.py#L1164 allow one-column dataframe and convert to series.
https://github.com/microsoft/FLAML/blob/a64956a7c8f98db6a7de7e8cc5d0ce53e9c86a03/flaml/automl.py#L1560 https://github.com/microsoft/FLAML/blob/a64956a7c8f98db6a7de7e8cc5d0ce53e9c86a03/flaml/automl.py#L2086 https://github.com/microsoft/FLAML/blob/a64956a7c8f98db6a7de7e8cc5d0ce53e9c86a03/flaml/automl.py#L2178 update docstr.
@qingyun-wu @ekzhu Please comment if I missed anything.
Thanks! Would I not want to convert the one-column dataframe to a column before using it, or will that happen automatically if it passes the assertions?
Thanks! Would I not want to convert the one-column dataframe to a column before using it, or will that happen automatically if it passes the assertions?
Please convert it. It won't happen automatically. Thanks.