search

rapidsai / cuml

cuML - RAPIDS Machine Learning Library
https://docs.rapids.ai/api/cuml/stable/
Apache License 2.0
4.2k stars 525 forks source link

[QST] Can cuml use multiple cores if GPU fails? #5998

Open MikeB2019x opened 2 months ago

MikeB2019x commented 2 months ago

I have an installation with 32 CPU's and 2 GPU's. If the GPU's are not available and I'm using cuml, will the latter take advantage of the multiple cores or will all processing fallback to a single core?

viclafargue commented 2 months ago

It is possible to easily switch to CPU (backed by Scikit-Learn) for some of the estimators. Most calculations in Scikit-Learn are handled by Numpy and should thus be able to use CPU optimally. To do this please refer to this documentation page.