When we install torch through the Pytorch official website, we will find that the installation is often unsuccessful due to network speed problems. The following provides a simple method to successfully install Cuda12.1, and the personal test is effective.
Table of contents
2.1 Create and activate a virtual environment
1. Conventional method
Install torch through the Pytorch official website, the URL link is as follows: PyTorch
You will find that the installation speed is very slow, and it is easy to fail.
2. Effective methods
2.1 Create and activate a virtual environment
Create a virtual environment
conda create -n pytorch39 python==3.9
Activate the virtual environment
conda activate pytorch39
2.2 Add Tsinghua source
Adding Tsinghua source can achieve network acceleration and greatly improve the success rate of installation
pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
2.3 install torch
pip install torch==2.0.0+cu118 torchvision==0.15.1+cu118 torchaudio==2.0.1+cu118 -f https://download.pytorch.org/whl/torch_stable.html
Verify that the installation was successful
Anaconda prompt 里输入 python
>>> import torch
>>> torch.__version__
'2.0.0+cu118'
>>> torch.cuda.is_available()
True
>>> torch.cuda.get_device_name(0)
'NVIDIA GeForce RTX 3090'
As shown in the figure above, pytorch (Cuda12.1) is installed successfully!