LLM 本地化部署1:安装 CUDA 环境

目录

1. 安装 CUDA Toolkit

1.1. 检查 CUDA 支持

首先,确保你的 GPU 支持 CUDA

https://developer.nvidia.com/cuda-gpus

1.2. 在线安装(新版本)

https://developer.nvidia.com/cuda-downloads

bash
# cu13.0
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda-toolkit-13-0
sudo apt-get install -y cuda-drivers

注:最新版 cu13 存在非常非常多的兼容问题,大量第三方框架没有适配,不推荐安装。

1.3. 离线安装(旧版本)

https://developer.nvidia.com/cuda-toolkit-archive

1.3.1. 下载安装包

bash
# cuda 12.9 (ubuntu24.04)
# pytorch 最新稳定版适配到 cu129
# https://pytorch.org/get-started/locally/
wget https://developer.download.nvidia.com/compute/cuda/12.9.1/local_installers/cuda_12.9.1_575.57.08_linux.run
sudo sh cuda_12.9.1_575.57.08_linux.run

# cuda 12.4
# cu124 兼容性最好,但只能在 ubuntu22.4 中安装
wget https://developer.download.nvidia.com/compute/cuda/12.4.1/local_installers/cuda_12.4.1_550.54.15_linux.run
sudo sh cuda_12.4.1_550.54.15_linux.run

注:若第一次安装失败,需重启后再安装。

1.3.2. 环境设置

bash
# sudo vim /etc/profile
PATH=$PATH:/usr/local/cuda-12.9/bin
export PATH

1.3.3. 检查卸载

bash
nvidia-smi
cuda-uninstaller
nvidia-uninstall

2. 安装 Container Toolkit

NVIDIA Container Toolkit 是一组库和实用程序,使用户能够构建和运行 GPU 加速的容器。它目前包括:

  • NVIDIA Container Runtime (nvidia-container-runtime)
  • NVIDIA Container Toolkit CLI (nvidia-ctk)
  • NVIDIA CDI Hooks (nvidia-cdi-hook)
  • NVIDIA Container Runtime Hook (nvidia-container-runtime-hook)
  • NVIDIA Container CLI (nvidia-container-cli)
  • NVIDIA Container Library (libnvidia-container1)

2.1. 安装 Docker

https://docs.docker.com/engine/install/ubuntu/

bash
curl https://code.wangyan.cloud/wangyan/docker-compose/raw/branch/main/shell-script/docker-setup-ubuntu-aliyun.sh | bash

2.2. 安装 container-toolkit

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

bash
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit

2.3. 配置 Docker 运行时

bash
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker
ESC