分布式缓存系统

Cachelib

git clone -b v2023.04.24.00 https://github.com/facebook/CacheLib.git  
cd CacheLib  
./contrib/update-submodules.sh  
# 有一定的gcc版本要求,参考软件安装中的gcc版本和glibc版本  
export CC=[/home/username/installed/gcc-12]/bin/gcc  
export CXX=[/home/username/installed/gcc-12]/bin/g++  
export PATH=[/home/username/installed/gcc-12]/bin:$PATH  
export LD_LIBRARY_PATH=[/home/username/installed/gcc-12]/lib64  
./contrib/build.sh -j -v  

后续修改cachelib代码,再编译:
cd build-cachelib
make -j[n]  
make install  

MLPerf

安装

git clone -b v1.0 --recurse-submodules https://github.com/mlcommons/storage.git  
conda create -n mlperf python=3.10.12  
conda activate mlperf  
pip3 install -r dlio_benchmark/requirements.txt  
# pip安装mpi4py时报错,conda单独安装一下,然后再pip3 install  
conda install mpi4py  

分布式测试

# 要求:多机的conda环境路径要一致,mlperf代码位置要一致

results matching ""

    No results matching ""