1.点击下载并安装：Anaconda（个人习惯一般安装放在D盘）
- 打开系统属性，点击环境变量;（温馨提示：在
cmd+r打开终端terminal，输入sysdm.cpl，在高级中打开环境变量） - 在系统变量中找到
Path变量，选择后点击编辑; - 新建添加三个变量（根目录、Scripts目录、Library下bin目录）；
🙌示例：
D:\Anaconda
D:\Anaconda\Scripts
D:\Anaconda\Library\bin
win+r输入cmd，然后在命令窗口输入conda -V命令，显示conda版本就代表配置成功；- 打开
Anaconda Navigator选择打开python IDE--spyder、pycharm、vscode任君选择；推荐在安装CUDA之前安装好vscode
- 查看CUDA版本：
WIN+X打卡终端Windows PowerShell，然后输入nvidia-smi查看CUDA版本
PS C:\Users\maihuanzhuo> nvidia-smi
Tue Mar 19 18:22:02 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 546.29 Driver Version: 546.29 CUDA Version: 12.3 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 4060 ... WDDM | 00000000:01:00.0 On | N/A |
| N/A 50C P5 8W / 122W | 1985MiB / 8188MiB | 22% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+-
这里可以看显卡对应的算力，下载CUDA对应电脑的版本，默认安装即可
-
安装完CUDA后，
WIN+X打卡终端Windows PowerShell，然后输入nvcc --version，出现对应的CUDA版本信息
PS C:\Users\maihuanzhuo> nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Fri_Nov__3_17:51:05_Pacific_Daylight_Time_2023
Cuda compilation tools, release 12.3, V12.3.103
Build cuda_12.3.r12.3/compiler.33492891_0-
注册英伟达账号，然后申请验证一下，通过后即可下载对应CUDA版本的CUDNN（CUDA12.2对应CUDNN8.9.3，亲测4090装的是CUDNN8.2可兼容）
-
然后解压之后，除了==LICENSE==文件之外，其他三个文件夹复制到==C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.2==目录下
-
接着添加CUDNN环境变量
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.2\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.2\libnvvp
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.2\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.2\include
- 检查是否安装成功，
win+r输入cmd打开终端
cd C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\extras\demo_suite- 分别运行 ==deviceQuery.exe==和==bandwidthTest.exe==，出现PASS则为成功
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\extras\demo_suite>deviceQuery.exe
deviceQuery.exe Starting...
................................
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 12.3, CUDA Runtime Version = 12.3, NumDevs = 1, Device0 = NVIDIA GeForce RTX 4060 Laptop GPU
Result = PASS
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\extras\demo_suite>bandwidthTest.exe
[CUDA Bandwidth Test] - Starting...
Running on...
................................
Result = PASS
NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.- 查看CUDNN版本，在==C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.2\include==目录下，打开==cudnn_version.h==文件
C:\Users\maihuanzhuo>cd C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include>cudnn_version.h
#define CUDNN_MAJOR 8
#define CUDNN_MINOR 9
#define CUDNN_PATCHLEVEL 6
#因此我的cudnn版本是8.9.6-
首先根据TensorFlow官网安装教程 ==tensorflow-2.6==要求==Python3.6-3.9==，CUDA跟CUDNN应该是往下兼容的
-
因为spyder默认安装的是python 3.10或者是3.11，因此我们要创建一个虚拟环境env（跟linux一样），在
终端terminal中输入：
conda create -n tensorflow_2.6 python=3.9 -y- 创建环境后激活环境
(base) C:\Users\maihuanzhuo>conda activate tensorflow_2.6
(tensorflow_2.6) C:\Users\maihuanzhuo>conda install tensorflow-gpu==2.6.0 -y #GPU版本
conda install tensorflow==2.6.0 -y #CPU版本
pip install tensorflow-gpu==2.6.0 -i https://pypi.tuna.tsinghua.edu.cn/simple #GPU版本，网不行的可以加个镜像
pip install tensorflow==2.6.0 -i https://pypi.tuna.tsinghua.edu.cn/simple #CPU版本- 在
terminal中输入python打开python交互界面，导库检查是否报错import tensorflow as tf
import tensorflow as tf
print(tf.__version__)(tensorflow_2.6) C:\Users\maihuanzhuo>python
Python 3.9.18 (main, Sep 11 2023, 13:30:38) [MSC v.1916 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
>>> print(tf.__version__)
2.6.0- 没有报错就代表安装成功，输入quit()来退出
- 正常来说应该是在tensorflow_2.6环境下conda install spyder，不知道为什么一直装不上，在Anaconda Navigator页面里，spyder是显示没有安装的，所以比较麻烦
- 那么我们就通过pip源来安装spyder
pip install spyder -i https://pypi.tuna.tsinghua.edu.cn/simple- 通过pip源安装后，有个问题，现在进入spyder需要先激活tensorflow_2.6环境，然后在cmd prompt中输入spyder启动
(base) C:\Users\maihuanzhuo>conda activate tensorflow_2.6
(tensorflow_2.6) C:\Users\maihuanzhuo>spyder
这个算是tensorflow的老毛病吧，迟迟没有更新，毕竟从1.0推出至今快十年了，当然现在更多是推荐pytorch。pytorch安装方法也一样，安装好了CUDA和CUDNN之后，为pytorch创建一个新环境，官方要求python版本为3.8-3.11，创建好之后激活环境，然后在pytorch官网；根据对应配置复制代码，在新激活的环境中输入自动下载pytorch，同样也通过pip源安装spyder
请注意：根据python版本和CUDA版本对应安装，一般来说CUDA都会往下兼容，自身电脑CUDA版本太高也没关系
NOTE: Latest PyTorch requires Python 3.8 or later. For more details, see Python section below.
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 #CUDA 11.8
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121 #CUDA 12.1
pip3 install torch torchvision torchaudio #CPU当我们配置好env进入spyder，导库的时候会发现numpy报错，这是因为numpy版本太高，已经不支持 np.object所导致
AttributeError: module 'numpy' has no attribute 'object'.
`np.object` was a deprecated alias for the builtin `object`. To avoid this error in existing code, use `object` by itself. Doing this will not modify any behavior and is safe.
The aliases was originally deprecated in NumPy 1.20; for more details and guidance see the original release note at:
https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations解决：将numpy版本修复到仍然支持使用 np.object 的最后一个版本1.23.4
pip uninstall numpy
pip install numpy==1.23.4当然也会出现如下一系列库的版本不兼容，以及缺乏相应的库
ERROR: ERROR: pip’s dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
tensorflow 2.6.0 requires clang=5.0, which is not installed.
tensorflow 2.6.0 requires keras>=2.4.0, which is not installed.
pandas 2.2.1 requires numpy<2,>=1.22.4;python version<“3.11”, but you have numpy 1.20.0 which is incompatible.
scipy 1.11.4 requires numpy<1.28.0, >=1.21.6, but you have numpy 1.20.0 which is incompatible.
tensorflow 2.6.0 requires google-auth<2, >=1.6.3, but you have google-auth 2.22.0 which is incompatible.
tensorflow 2.6.0 requires absl-py =0.10, but you have absl-py 1.4.0 which is incompatible.
tensorflow 2.6.0 requires flatbuffers =1.12, but you have flatbuffers 20210226132247 which is incompatible.那就需要一一对应安装所提示的库，譬如需要requires keras>=2.4.0
pip install keras==2.4.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/