A county assessor hears that XGBoost can run faster on a GPU. What…

Question

A county assessor hears that XGBoost can run faster on a GPU. What does that mean at associate depth?

Accepted Answer

Use a GPU-accelerated implementation of the same tree algorithm. Do not author CUDA kernels or design a new network. — GPU-aware XGBoost is the same tree algorithm on an accelerated implementation. Do not author CUDA kernels or design a new network. A transformer and a Dask cluster are the wrong turn.

Answer

Write a custom CUDA kernel that replaces XGBoost, because GPU trees require device code you author.

Answer

Replace the tree booster with a transformer, because a GPU cannot run boosted trees.

Answer

Design a multi-node Dask cluster first, because GPU XGBoost is a Domain 4 ingest architecture exam.

A county assessor hears that XGBoost can run faster on a GPU. What does that mean at associate depth?

Short Explanation

Full Explanation