| PyTorch Build | Compute Platform | Python Version | Your OS | Arch | |
|---|---|---|---|---|---|
| 2.3.0 | CUDA 12.1 | Python 3.12 | Windows | x86_64 | |
| 2.3.0 | CUDA 12.1 | Python 3.12 | Linux | x86_64 | |
| 2.2.2 | CUDA 12.1 | Python 3.12 | Windows | x86_64 | |
| 2.2.2 | CUDA 12.1 | Python 3.12 | Linux | x86_64 | |
| 2.2.1 | CUDA 12.1 | Python 3.12 | Windows | x86_64 | |
| 2.2.1 | CUDA 12.1 | Python 3.12 | Linux | x86_64 | |
| 2.2.0 | CUDA 12.1 | Python 3.12 | Windows | x86_64 | |
| 2.2.0 | CUDA 12.1 | Python 3.12 | Linux | x86_64 |
Install PyTorch
Select your CUDA, ROCm, Python and OS to find PyTorch version you need.
• Why did you build this website?
I found it hard to find the right PyTorch version for my project. I have to go to the official repository and search by browser, it is not a good experience. So I build this website to help me and others like me find the right PyTorch version easily.
• CPU vs Cuda vs ROCm?
If you have no GPU or only want to use PyTorch on CPU, select CPU. If you have Nvidia GPU, select Cuda. If you have AMD GPU, select ROCm.
• Which Cuda version should I choose?
If you have Nvidia GPU, you should choose the Cuda version that is compatible with your GPU. You can check the compatibility at Nvidia official website. And a easy way is to use nvidia-smi command to check. Of course, you should make sure that you have installed the right driver.
• Which ROCm version should I choose?
If you have AMD GPU, you should choose the ROCm version that is compatible with your GPU. You can check the compatibility at AMD official website. And you should make sure that you have installed the right driver and ROCm.
• I have install ROCm 6.0, but I can not find ROCm 6.0 in the list?
You could install ROCm 5.7 instead. However, when you use PyTorch, you need to set the environment variable HSA_OVERRIDE_GFX_VERSION=10.3.0.
• Which PyTorch version should I choose for my project?
Firstly, You should choose the PyTorch version that is compatible with your model. You also need to consider the compatibility of the PyTorch with your device, Python and OS.
• What should I do next after I have downloaded the PyTorch wheel file?
You can install the PyTorch wheel file by using the command pip install your_wheel_file.whl. You can also copy the command from the website and paste it into shell and run.