search

JuliaAI / MLJXGBoostInterface.jl

MIT License
11 stars 5 forks source link

Impossible to train a XGBoost model #11

Closed egolep closed 3 years ago

egolep commented 3 years ago

Anytime I try to fit a model, I keep getting the same error:

got unsupported keyword argument "label"

and this happens in the DMatrix call.

Looking directly to XGBoost.jl repo, they actually use DMatrix with the label argument, so is it a version problem between the MLJ interface and XGBoost.jl or am I missing something?

egolep commented 3 years ago

The problem was missing values in the dataset. Not very intuitive from the error message, but still a problem of mine.

Forgive me

ablaom commented 3 years ago

Curious as to how you could present missing values without a scitype warning being raised. I would have expected a complaint about data input_scitype being incompatible with the model input_sciype. Are you not constructing a machine, perhaps?

If you can provide reproducible example, I can try to improve the error reporting.