用Conda配置Pytorch环境 (pytorch==2.2.1)

本文主要讲解: 如何用Conda搭建Pytorch环境，用Conda的方式安装，不需要单独去安装Cuda了。

1. 安装miniconda

https://docs.anaconda.com/free/miniconda/index.html

2. 搭建虚拟环境

激活python虚拟环境

conda create -n env310 python==3.10.14 -y
conda activate env310

安裝Pytorch

conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=11.8 -c pytorch -c nvidia

安裝常用包

pip install scikit-learn einops ipywidgets pandas tqdm jupyterlab matplotlib seaborn

3. Conda设置清华源

conda config --show channels

conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --set show_channel_urls yes

4. pip设置清华源

pip config list
pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple

5. 一个分类网络的例子

测试Pytorch环境是否Okay

conda activate env310
python mnist.py

文件mnist.py内容：

# Task
# Our task is simple, recognize handwritten digits. We will use MNIST dataset for this tutorial.
# 

# # Import necessary library
# In this tutorial, we are going to use pytorch, the cutting-edge deep learning framework to complete our task.

# In[2]:


import torch
import torchvision


# In[3]:


## Create dataloader, in PyTorch, we feed the trainer data with use of dataloader
## We create dataloader with dataset from torchvision, 
## and we dont have to download it seperately, all automatically done

# Define batch size, batch size is how much data you feed for training in one iteration
batch_size_train = 64 # We use a small batch size here for training
batch_size_test = 1024 #

# define how image transformed
image_transform = torchvision.transforms.Compose([
                               torchvision.transforms.ToTensor(),
                               torchvision.transforms.Normalize(
                                 (0.1307,), (0.3081,)