Issue with Donwloading the model and running generate? about ai-friends

[Skquark]

So trying to figure out the best way to resolve this issue.. As I've mentioned before, I haven't tested enough on the local desktop due to my personal GPU being trash, so I appreciate the bug reports,, Now to figure out how to fix..
The error coming from the line running torch.Generator("cuda") is strange since it seems pretty basic. When I looked up that RuntimeError, looks like the cause is from it having an incompatible version of torch or CUDA installed on the venv. I have it setup to install torch if it's not found or version below 2.0 using "pip install -U --force-reinstall torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118" which possibly wasn't working on your computer. Wondering if in your console you saw the message "Installing PyTorch 2.0.1 with CUDA 11.8"? When looking into troubleshooting the error, one cause is an older version of the NVidia drivers which isn't compatible with cuda 1.18, so if you run "nvidia-smi" from your terminal, tell me what CUDA Version number you see on the top right. Maybe updating your vid drivers could fix the issue. Also curious what torch version is installed, which you can see running "pip freeze", but you may have to check it after activating the virtual environment in command prompt. Ideally we want to see torch==2.0.1 so I'm thinking on my install line I should explicitly state that version number.
This isn't the kind of problem I enjoy troubleshooting, but hopefully my suggestions make sense and we can get it working. Was much easier debugging in Colab since it's consistent and has cuda/torch preinstalled. Lemme know how it goes and hopefully there's something I can correct on my end to prevent others from getting this issue.. Thanks.

[Skquark]

Checking if you got this working on your end... You may want to download latest setup and reinstall it, updated some of the bat files, and even included a debug-venv.bat that lets you open command prompt easily inside the virtual environment to manually run installs and see what the error messages are saying. I hide the console outputs when I install things to make it cleaner, but sometimes if things don't work it's harder to know what's up. Installing Torch & CUDA can be tricky on some systems, but trying to make things universal as long as GPU is supported. Hope it's good now, please let me know.