Boosted trees are a fast, robust way to model all types of data. XGBoost is an incredibly popular library for building ML models, especially gradient boosted decision trees.
W&B users have reached out to ask if we could make our visualizations work easily with XGBoost. Our tool is framework agnostic, so you can build your own custom logging using wandb.log(dict), but I wanted to make it as easy to visualize XGBoost as it is for Keras, TensorFlow or PyTorch.
I built a callback called wandb_callback that you can pass into your XGBoost train function. For example:
# save the param_list as hyperparameter inputs
# log metrics as the model trains
bst = xgb.train(param_list, d_train, callbacks=[wandb_callback()])
The W&B integration made it quick and easy to monitor the success in a dashboard.