Make ComyUI 50 Percent Faster With Sage Attention - Easiest Guide!

Wouldn't it be nice if your comi workflows would run 50% faster? If you have an Nvidia graphics card, then this dream can become true. To achieve this, we are installing Triton and Sage Attention. Both are technologies that optimize GPU calculations in neural networks. To do this, first make sure you have the latest version of your GPU driver from Nvidia. Next, we need to download Python 3.2. 2. So go to this URL here and scroll down until you find the right installer. If you're on Windows, then use the 64bit installer. I can't show you the installation of Python 3.12 because I already have it. So it says modify up here instead. The last thing we need is Git. So go to this URL and click on the installer you need. In my case, Windows and here x64. So, when installing Git, all you need to do is click next and next again and next again and next and again. Do this a couple of times until you actually can install it. You can leave everything as the default settings here. Instead of downloading the portable package, we are installing it manually. So, to do this, click on code and then click this copy URL to clipboard icon. Then type git clone and paste the URL we copied from GitHub and hit enter. Now we enter the folder with the ID confi. Hit enter. And here we are going to create a new virtual environment by typing pi minus 3.12 minus m vn vn and we hit enter. And now we type call vn back back slashcripts back slashactivate and we hit enter. For the next command we go to the confi GitHub page and we scroll down until we see Nvidia install. So it's here and we copy this future Chris here. Some things changed in comui and you need to install pietorrch 2.8 8. Now, with this command, there will be another change later in the video. Back in our console window, we paste this and hit enter. If this is all too complicated for you, I have a one-click installer for my Patreon supporters. Just saying. So, this took about 3 minutes on my machine. We grab this command here, pip install requirements, and we paste it again here, and we hit enter.