Open ghostsquad opened 6 days ago
Have you run %reset -f out
to free the RAM, as discussed in #48?
And, if you are using ComfyUI loaded in the same process (the standalone runtime), connecting to a ComfyUI server running separately (e.g. load('http://127.0.0.1:8188/')
) may provide better stability.
I'll try not running the server from VSCode and see if that helps. Right now I don't have another machine. That may come later.
I'm curious if there's some setting or something I need to change. It doesn't seem to take long before image generations take multiple minutes to complete. As if maybe I've run out of VRAM and processing is now happening on the CPU or something, which correlates strongly with the fact that applications of all sort slow down significantly too. I'm wondering what is going on that's causing GPU VRAM not be reclaimed. Anyways, this only seems to happen when using ComfyScript via VSCode. I haven't had this issue at all in ComfyUI.