Thanks for this @Xteve but sadly still no luck even if I reproduce your steps verbatim.
Looking in indexes: https://download.pytorch.org/whl/nightly/cpu
ERROR: Could not find a version that satisfies the requirement torch (from versions: none)
ERROR: No matching distribution found for torch
From what I can tell, I need this wheel cp313-none-macosx_11_0_arm64 but that tag isn’t available when I run pip debug --verbose. So at this point I’m going to throw in the towel. There’s clearly something bizarre with my setup as it ought to be very straight forward but no matter how deep I get into the dependencies, I can’t get the required versions to install.
Thanks all for the help. At least I’ve learned a bit about virtual environments so that’s a small win, haha.
I mean, yeah…that’s how I would expect mine to work as well…alas it does not. Again, I appreciate the help but something is clearly a little off on my end.
if you show me a picture(cropped screenshot) of what happen on your side, I can help you in an instant. if conda is installed (and you opened a NEW!!! Terminal window), then you build your environment:
conda create -n sammie python=3.13
after the python environment is ready, you just activate it:
conda activate sammie
I am sure your conda isn’t recognized yet after installation. therefore you just open a NEW terminal window.
Are you sure that minconda is running and activated when you try to install pythorch?
I have to execute this:
source ~/miniconda/bin/activate.
And then
conda activate sammieroto
This is for linux. ~/miniconda/ is my installation path. So, for mac and your install path it will be different. “sammieroto” is the name of env created following the guide of cnoellert
I put it all together into a script sh. In mac maybe an automator file could be a good option.
Try this one= pip install torch torchvision torchaudio
If you already have any torch installed, remove it by running= pip uninstall torch torchvision torchaudio
By default torch is cpu mode, if you want to use a cuda one you need to use an extra flag to specify which cuda you want torch to use. So just by installing torch without any extra flag will install the one that uses cpu instead of cuda.
To all interested.
This tool now has blockswapping (by default) which means you can now run longer and higher resolution clips without getting Out Of Memory. Just update to 1.7.0.
