Getting started with XGBoost
WML CE includes XGBoost 0.90 and Dask-XGBoost 0.1.9.
Overview
- XGBoost
- XGBoost is an optimized distributed gradient boosting library designed to be highly efficient,
flexible, and portable. It implements machine learning algorithms under the Gradient Boosting
framework. XGBoost is available on both x86 and
IBM®
Power®.
Tutorials and examples are available from the community on the XGBoost Documentation page.
- Dask-XGBoost
- Dask-XGBoost enables you to perform distributed training with XGBoost on Dask.array and
Dask.dataframe collections.
Dask-XGBoost
is available only on IBM Power.Tutorials and examples are available from the community on the dask-xgboost github page.
Getting started with XGBoost
To install XGBoost, run the appropriate command:
- GPU variant and dependencies:
conda install py-xgboost-gpu
- CPU variant and dependencies:
conda install py-xgboost-cpu
The following XGBoost related conda packages are included with WML CE:
Name | Description | Installed with powerai | Installed with py-xgboost-gpu | Installed with powerai-cpu | Installed with py-xgboost-cpu |
---|---|---|---|---|---|
py-xgboost-gpu | XGboost GPU Meta package | ✓ | ✓ | ||
py-xgboost-cpu | XGboost CPU Meta package | ✓ | ✓ | ||
libxgboost-base | XGBoost C++ library | ✓ | ✓ | ✓ | ✓ |
py-xgboost-base | XGBoost Python library | ✓ | ✓ | ✓ | ✓ |
_py-xgboost-mutex | CPU/GPU selector metapackage | ✓ | ✓ | ✓ | ✓ |
Getting started with Dask-XGBoost
Dask-XGBoost is provided in the dask-xgboost
conda package, which is compatible
with both the CPU and GPU variants of XGBoost.
To install dask-xgboost, run the appropriate command:
- dask-xgboost and the GPU variant of xgboost:
conda install py-xgboost-gpu dask-xgboost
- dask-xgboost and the CPU variant of xgboost:
conda install py-xgboost-cpu dask-xgboost
Known issue: In a conda environment where the CPU or GPU XGBoost variant is already installed, there is a known conda issue with installing the other XGBoost variant. This issue can be avoided by uninstalling the existing variant before installing the desired XGBoost variant.
Note: Dask-XGBoost is a technology preview.