To get started, we'll need the old version of Comfyui.

unpacking to a place convenient for you

moving the custom_nodes and models folders from our Comfyui to the new one

we launch a new Comfyui and see this picture

updating comfyui using manager

after the update, we see that the red is no longer there and close comfyui.

next, we put this file in the ComfyUI_windows_portable\update

далее распаковываем этот архив в папке ComfyUI_windows_portable

next, go to the folder ComfyUI_windows_portable\update

and enter cmd into the address bar and press enter

next, copy the lines and install them one at a time

..\python_embeded\python.exe -s -m pip install "accelerate >= 1.1.1"

..\python_embeded\python.exe -s -m pip install "diffusers >= 0.31.0"

..\python_embeded\python.exe -s -m pip install "transformers >= 4.39.3"

..\python_embeded\python.exe -s -m pip install ninja 

..\python_embeded\python.exe -s -m pip install --upgrade torch  torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

..\python_embeded\python.exe -s -m pip install triton-3.1.0-cp311-cp311-win_amd64.whl

downloading the archive

and unpack it into a folder ComfyUI_windows_portable\python_embeded

Если у вас не установлен Microsoft Visual Studio then install https://visualstudio.microsoft.com/ru/downloads/

checkboxes that need to be checked

next, go to the ComfyUI_windows_portable\SageAttention folder

run cmd and enter

..\python_embeded\python.exe -m pip install .

The installation is long, I have more than 10 minutes.

go to the ComfyUI_windows_portable folder and click edit on the run_nvidia_gpu.bat file and add the line

--use-sage-attention

that's all, we launch it and see the difference.

On my 4060 ti 16vram at 720*1024 resolution, 53 frames 10 steps are generated for

4min 40sec and it was 7min 50sec.