DASF is a generic framework specialized in acceleration and scaling common techniques for Machine Learning. DASF uses most methods and function from the most common libraries to increase the speed up of most algorithms. Part of this is to use Dask data to scale computation and RAPIDS AI algorithms to extend the support to GPUs as well.
https://github.com/discovery-unicamp/dasf-core
For the main DASF project, please visit the link above.
This project aims at enabling the deployment of a ARGO and DASF environment in a kubernetes cluster. The project report is available at:
PFG 66 Ramon Galate.pdf