DistStat.jl: Towards unified programming for high-performance statistical computing environments in Julia
Dependencies:
- Julia >= 1.4
- An MPI installation (tested on OpenMPI, MPICH, and Intel MPI)
- MPI.jl >= 0.15.0 (and its dependencies)
- Select proper MPI backend when building MPI.jl, as described in this page)
- CustomUnitRanges.jl
- See
Project.toml
For CUDA support:
- CUDA >= 9.0
- CUDA.jl, GPUArrays.jl (and their dependencies)
- CUDA-aware MPI installation (of OpenMPI, MPICH, and Intel MPI, only OpenMPI supports CUDA)
- MPI.jl should be built with the environment variable
JULIA_MPI_BINARY=system
; see this page).
To install the package, run the following code in Julia.
using Pkg
pkg"add https://github.com/kose-y/DistStat.jl"
Examples of nonnegative matrix factorization, multidimensional scaling, and l1-regularized Cox regression is provided in the directory examples/
. Settings for multi-gpu experiments and multi-instance cloud experiments are also provided.
This work was supported by AWS Cloud Credits for Research. This research has been conducted using the UK Biobank Resource under application number 48152.
Ko S, Zhou H, Zhou J, and Won J-H (2020+). DistStat.jl: Towards Unified Programming for High-Performance Statistical Computing Environments in Julia. arXiv:2010.16114.