Author: Dalvan Griebler (Contact: [email protected])
This project aims to provide a simpler parallel programming abstraction for Python programs over/using MPI. It is designed based on the structured parallel programming approach and stream processing applications. It intends to provide parallel patterns suitable to structure and modeling stream-based computations.
ssp4py uses the mpi4py as the backend.
Hopefully you can speed-up your machine-learning tasks for pre-processing data, training algorithms and testing models in Python.
- Support dynamic/ondemand schedule for the Farm pattern
- Support ordering for the Farm pattern
- Support Pipeline pattern
- Support arbitrary nesting of Pipeline and Farm patterns
- Support parallel Map and MapReduce pattern
- Support arbitrary nesting of Pipeline, Farm, Map, and MapReduce patterns