NimbusML
NimbusML copied to clipboard
Documentation of Handler's "replace_with" parameter is misguiding
The documentation for the Handler class (link) mentions the following for the "replace_with" parameter:
replace_with
The method to use to replace NaN values. The following choices are available. Def: Replace with default value of that type, usually 0. If no replace method is specified, this is the default strategy. Mean: Replace NaN values with the mean of the values in that column. Min: Replace with minimum value in the column. Max: Replace with maximum value in the column.
This is misguiding, since the actual options that it accepts are the following:
https://github.com/microsoft/NimbusML/blob/d08b702ade3e4d8bf487fe583f9632c40a7a774b/src/python/nimbusml/internal/entrypoints/transforms_missingvaluehandler.py#L58-L62
If I try to use replace_with='Max'
then I get the following cryptic error message:
parameter passed not in values.
Since it's actually expected to use replace_with='Maximum'
, although the documentation doesn't say it explicitly, and the error message doesn't say which parameter of what component is not accepted.