pushpendre

a workaround is to configure autogluon as follows ```python import autogluon.utils.tabular.features.utils autogluon.utils.tabular.features.utils.logger.disabled = True predictor = task.fit(..., excluded_model_types=['NN']) ```

I was actually going into the internals of auto-gluon yesterday to try and make this thing work, since I'll really prefer to use auto-gluon as a top-level orchestrator for lightgbm...

@Innixma here's a minimal patch that I wrote to appease the `abstract_feature_generator` and the `auto_ml_feature_generator.py`. https://gist.github.com/pushpendre/a17ce02e8bc2ee293f93360495070954 With these changes my MWE passes ```bash Feature Generator processed 10000 data points with...

I went ahead and did some work to use Pandas SparseDataframe with lightgbm. It turns out that lightgbm does not support the extension dtype `Sparse[]` that is defined in pandas....

Focusing on a case where the input is an array, can also help in adding a flag where the preprocessing can be skipped. If the user is passing in an...

i have a couple of deadlines coming up, I'll try to put something together probably after Sep 15.

Weirdly enough when I run this same code I get the error that "MTLLibrary is not formatted as a MetalLib file" ```bash (tg) [20:12:22]Mac:~/w/(02-22_06:16) $ python -c 'import objc; print(objc.__version__)'...

Can confirm, I had to make some more changes to get this thing to work but the bug disappeared with the system python3. Specifically here were the steps 1. Install...