Looking for suggestion to achieve Faster SDXL Outputs
Hi, I am currently trying to generate large amount of images (200+) every session via Automatic1111/Forge UI SDXL Model and was wondering how can I generate them fast? I tried using RTX 3090 for generation and it's about 1.5-2 it/s which is pretty slow in the long run.
Perhaps there is a faster alternative to this or workflow? Please let me know.
Provide me a workflow and GPU suggestion that can generate large amount of images swiftly.
7 Replies
ComfyUI + rtx4090 i think that setup can be faster
+ xformers
I would think you could increase your speed for generate large numbers of images by running more workers. Say you are going to need 300 images you could either have 1 worker generate all 300 images, sequentially. Or you could have 30 workers generate just 10 images each in parallel.
Ooops when responding to this, I assume you were using serverless workers. I suggest you look into serverless workers.
yes thats a verygood one ig, at s cale with many workers you can do much more
you should consider the serverless
@Encyrption @nerdylive Thanks for the suggestions but is there any proper guide on how I can setup serverless properly with my preferred model and extensions?
there are serverless worker code's there that you can copy
like from https://github.com/runpod-workers
GitHub
RunPod | Endpoints | Workers
Offical collection of worker repositories for serverless endpoints on RunPod. - RunPod | Endpoints | Workers
thank you very much