I am trying to install and configure Stable Diffusion AI locally on my PC (Windows 11 Pro x64), following a How-To-Geek article (How to Run Stable Diffusion Locally With a GUI on Windows).
Log story short, I ran into problems, primarily with the installation of Torch. So, after several attempts to fix that, I tried installing Torch separately
I already have python 6.1 (64bit) installed on my C: drive and used the following command line to install Torch:
I got all the way through the download and install, then got the error message:
ERROR: Package 'torch' requires a different Python: 3.6.1 not in '>=3.6.2'
I don’t really know what that means. It appears to be suggesting a different version of python, but a casual search suggests people are getting much the same issue with later versions.
What version of python do I need (if that’s even the issue)?
Python 3.6’s EOL was in Dec. 2021 and the support was dropped from newer PyTorch releases.
Update your Python version and you should be able to install PyTorch.
[quote=“ptrblck, post:2, topic:162055, full:true”]Python 3.6’s EOL was in Dec. 2021 and the support was dropped from newer PyTorch releases.
Update your Python version and you should be able to install PyTorch.[/quote]
OK. So I uninstalled Python 3.6x and installed Python 3.10.7 … I checked using “python -V” that the system thinks it is is using that version.
ERROR: Could not find a version that satisfies the requirement torch==1.8.1+cu111 (from versions: 1.11.0, 1.11.0+cpu, 1.11.0+cu113, 1.11.0+cu115, 1.12.0, 1.12.0+cpu, 1.12.0+cu113, 1.12.0+cu116, 1.12.1, 1.12.1+cpu, 1.12.1+cu113, 1.12.1+cu116)
ERROR: No matching distribution found for torch==1.8.1+cu111
Sorry, I should have warned you about too new Python versions, since you want to use an older PyTorch release. 1.8.1 wasn’t built for Python 3.10, so if you really need to install this older PyTorch release, use Python 3.9. However, since 1.8.1 is also quite old you should have been able to install it for Python 3.6 in the first place via this wheel: 1.8.1
You could alternatively manually download the desired wheels and pip install them.