DaskML with XGBoost and using eval_set requires pre-computed data

Question

I am trying to run dask_ml.xgboost using eval_set to allow for early stopping in an attempt to avoid overfitting.

Currently, I have a sample