Will Smith

Related to #1456

As a partial workaround, try this ``` from autogluon.tabular.configs.hyperparameter_configs import get_hyperparameter_config custom_hyperparameters = get_hyperparameter_config('default') ``` Now observe that every model type has a key in the dict. You can remove...

Note for developers : users seem to want an included_model_types parameter to make this functionality more intuitive.

Would this apply to tabular only, or all prediction types?

A few questions / thoughts to aid grooming. 1) what's the motivation? A) datasets large enough to fit into autogluon for subsampling, but too big for the models to train...

Assign to me!

Am working on this now. Some very interesting results so far. I'm actually looking at model performance vs CPU cores across all models.

Could column type inference hints be done through an API call, or maybe more simply, by hints in the name of the column? E.g. if a column is called expiry.ag_date...

I see a problem with ints never getting recognised as categoricals, even if the cardinality is low. The workaround is to use `df['x'] = df['x'].astype('category')` in manual feature engineering, as...

I've done some preliminary work and found that taking a categorical and duplicating it as both int and categorical can improve model quality using all AutoGluon presets. So far I...