当搭建deepforest模型时需考虑各基学习器的gpu加速问题。以下
LightGBM
读文档没看到明确说多卡的设置
好像要用到clinfo?哪位用过可否告诉我一下==
CatBoost
最简单,原生支持多gpu,直接设置参数:
params['task_type'] = 'GPU'
params['devices'] = '1,2,3'
XGBoost
可支持多gpu,需使用dask,可参考:
https://github.com/dmlc/xgboost/blob/master/demo/dask/gpu_training.py
Reference:
https://lightgbm.readthedocs.io/en/latest/GPU-Tutorial.html