【PyTorch】2024保姆级安装教程-Python-（CPU+GPU详细完整版）-

PyTorch 安装指南（CPU/GPU 版本）

环境准备

确保已安装 Python（推荐 3.8-3.10 版本）和 pip 包管理工具。可通过以下命令验证：

python --version pip --version

对于 GPU 版本，需提前安装 CUDA 工具包（推荐 CUDA 11.8）和 cuDNN。通过以下命令检查 CUDA 是否可用：

nvidia-smi

安装 PyTorch

访问 PyTorch 官网获取最新安装命令（https://pytorch.org/get-started/locally/），或直接使用以下通用命令：

CPU 版本

pip install torch torchvision torchaudio

https://www.zhihu.com/zvideo/1994565209331885213/
https://www.zhihu.com/zvideo/1994565208023258822/
https://www.zhihu.com/zvideo/1994565207540925794/
https://www.zhihu.com/zvideo/1994565206941127434/
https://www.zhihu.com/zvideo/1994565206744003950/
https://www.zhihu.com/zvideo/1994565206433625958/
https://www.zhihu.com/zvideo/1994565206098077303/
https://www.zhihu.com/zvideo/1994565204680397752/
https://www.zhihu.com/zvideo/1994565203636003581/
https://www.zhihu.com/zvideo/1994565201907955097/
https://www.zhihu.com/zvideo/1994565201689875180/
https://www.zhihu.com/zvideo/1994565200905524719/
https://www.zhihu.com/zvideo/1994565200305737977/
https://www.zhihu.com/zvideo/1994565198812561760/
https://www.zhihu.com/zvideo/1994565198850327780/
https://www.zhihu.com/zvideo/1994565197306815368/
https://www.zhihu.com/zvideo/1994565196514103586/
https://www.zhihu.com/zvideo/1994565196014966417/
https://www.zhihu.com/zvideo/1994565190486890060/
https://www.zhihu.com/zvideo/1994565189287318935/
https://www.zhihu.com/zvideo/1994565187051726155/
https://www.zhihu.com/zvideo/1994565188213569352/
https://www.zhihu.com/zvideo/1994565185751512843/
https://www.zhihu.com/zvideo/1994565186321921600/
https://www.zhihu.com/zvideo/1994565184279298762/
https://www.zhihu.com/zvideo/1994565184161858093/
https://www.zhihu.com/zvideo/1994565183075526026/
https://www.zhihu.com/zvideo/1994565181632692694/
https://www.zhihu.com/zvideo/1994565181045486231/
https://www.zhihu.com/zvideo/1994565181108413045/
https://www.zhihu.com/zvideo/1994565180231812002/
https://www.zhihu.com/zvideo/1994565179363571467/
https://www.zhihu.com/zvideo/1994565177887199780/
https://www.zhihu.com/zvideo/1994565177253855407/
https://www.zhihu.com/zvideo/1994565176838595115/
https://www.zhihu.com/zvideo/1994565175676777824/
https://www.zhihu.com/zvideo/1994565174481404230/
https://www.zhihu.com/zvideo/1994565174229750335/
https://www.zhihu.com/zvideo/1994565171113375318/
https://www.zhihu.com/zvideo/1994565168638742902/
https://www.zhihu.com/zvideo/1994565169397912751/
https://www.zhihu.com/zvideo/1994565168420638729/
https://www.zhihu.com/zvideo/1994565167074268258/
https://www.zhihu.com/zvideo/1994565168265442742/
https://www.zhihu.com/zvideo/1994565167678261233/
https://www.zhihu.com/zvideo/1994565166956823859/
https://www.zhihu.com/zvideo/1994565166822618564/
https://www.zhihu.com/zvideo/1994565166868763135/
https://www.zhihu.com/zvideo/1994565165560140329/

GPU（CUDA 11.8）版本

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

验证安装

运行 Python 解释器执行以下代码：

import torch print(torch.__version__) # 查看版本 print(torch.cuda.is_available()) # GPU 是否可用

常见问题解决

• CUDA 不匹配：卸载现有版本后重新安装指定 CUDA 版本的 PyTorch。
• 权限问题：添加--user参数或使用虚拟环境。
• 下载慢：更换 pip 源（如-i https://pypi.tuna.tsinghua.edu.cn/simple）。

虚拟环境推荐

使用 conda 或 venv 创建独立环境：

conda create -n pytorch_env python=3.9 conda activate pytorch_env