基本安装
港大LMFlow地址
https://github.com/OptimalScale/LMFlow
个人gitlab地址
http://yaoisss.ddns.net:9003/yaoisss/LMFlow
相关下载个人NAS地址
个人NAS地址
https://yaoisss.ddns.net:52311/portal/apis/fileExplorer/share_link.cgi?link=Hm34QCg1DeZTn2m6N2NQdw
- ubuntu server
- 安装好后查看 驱动是否占用 安装依赖
sudo vim /etc/modprobe.d/blacklist-nouveau.conf blacklist nouveau options nouveau modeset=0 sudo update-initramfs -u reboot lspci | grep nouvea apt install gcc g++ make - 安装NVIDIA驱动
nvidia驱动查找
https://www.nvidia.cn/download/find.aspx?lang=cn
- 安装CUDA Toolkit
CUDA11.7 https://developer.nvidia.cn/cuda-11-7-0-download-archive
报错后面加入 --override 安装完后 vim ~/.bashrc export PATH="/usr/local/cuda-11.7/bin:$PATH" export LD_LIBRARY_PATH="/usr/local/cuda-11.7/lib64:$LD_LIBRARY_PATH" - 安装Anaconda
- 运行
git clone https://github.com/OptimalScale/LMFlow.git conda create -n lmflow python=3.9 -y conda activate lmflow conda install mpi4py pip install -e . -i https://pypi.tuna.tsinghua.edu.cn/simple - 出错修改 requirements.txt
numpy==1.24.2 datasets==2.10.1 peft @ git+http://yaoisss.ddns.net:9003/yaoisss/peft torch==2.0.0 wandb==0.14.0 deepspeed==0.8.2 trl @ git+http://yaoisss.ddns.net:9003/yaoisss/trl.git#egg=trl-0.4.1 sentencepiece transformers @ git+http://yaoisss.ddns.net:9003/yaoisss/transformers flask flask_cors - 手动安装 deepspeed 0.8.3，试了一下不一定0.8.2出问题，需要的可以手动安装
pip install deepspeed-0.8.3.tar.gz - 基本就好了
- 使用都在 ./script 目录下
原文链接：https://blog.csdn.net/u011288483/article/details/129986194?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522168476299816800217251136%2522%252C%2522scm%2522%253A%252220140713.130102334.pc%255Fblog.%2522%257D&request_id=168476299816800217251136&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~blog~first_rank_ecpm_v1~times_rank-2-129986194-null-null.article_score_rank_blog&utm_term=NAS%E3%80%81%E7%BE%A4%E6%99%96%E3%80%81%E9%98%BF%E9%87%8C%E4%BA%91%E3%80%81%E5%9F%9F%E5%90%8D%E8%A7%A3%E6%9E%90%E3%80%81%E5%86%85%E7%BD%91%E7%A9%BF%E9%80%8F%E3%80%81ipv6%E3%80%81ddns%E3%80%81%E8%BD%BB%E9%87%8F%E7%BA%A7%E4%BA%91%E6%9C%8D%E5%8A%A1%E5%99%A8%E3%80%81%E9%93%81%E5%A8%81%E9%A9%AC%E3%80%81%E5%A8%81%E8%81%94%E9%80%9A%E3%80%81DSM%E3%80%81DSM6.0%E3%80%81%E7%BE%A4%E6%99%96nas%E3%80%81%E4%BA%91%E6%9C%8D%E5%8A%A1%E5%99%A8%E3%80%81%E8%9C%97%E7%89%9B%E6%98%9F%E9%99%85%E3%80%81%E9%BB%91%E7%BE%A4%E6%99%96%E3%80%81docker%E3%80%81%E5%AE%B9%E5%99%A8%E9%95%9C%E5%83%8F%E3%80%81%E5%9F%9F%E5%90%8D%E6%B3%A8%E5%86%8C%E3%80%81%E5%AE%9D%E5%A1%94%E3%80%81%E5%8F%8D%E5%90%91%E4%BB%A3%E7%90%86%E3%80%81nginx%E3%80%81frp%E3%80%81%E5%8A%A8%E6%80%81%E5%9F%9F%E5%90%8D%E8%A7%A3%E6%9E%90
