with1015 / horovod-benchmark

Pytorch horovod benchmark

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

PyTorch Benchmarks

Setup

Software packages

- CUDA 10.1
- Nvidia driver 418.56
- Ubuntu 16.04
- PyTorch >= 1.1

Build Conda Environment

$ conda create -n pt_1.8 python=3.6
$ conda activate pt_1.8
$ conda install pytorch==1.8.1 torchvision cudatoolkit=10.1 -c pytorch
$ pip install -r requirements.txt

Thanks to

All code is modified from https://github.com/NVIDIA/DeepLearningExamples/tree/master

About

Pytorch horovod benchmark


Languages

Language:Python 83.6%Language:HTML 7.0%Language:Shell 6.4%Language:Cuda 2.1%Language:Dockerfile 0.5%Language:C++ 0.4%Language:Batchfile 0.1%Language:Makefile 0.0%