Stable Diffusion Performance Optimization Tutorial

Veröffentlicht am: 04 Juni 2023
auf dem Kanal: Archive
46,248
681

Stable Diffusion Performance Optimization
I'll show you how to generate Images faster in Automatic1111. It should also work for Vlad Automatic. You can turn off GPU Scheduling in Windows Settings. You may also turn off Hardware Acceleration in the Browsers Settings you use the Stable Diffusion WebUI with. Additionally, you can increase the power budget of your GPU to maximum performance. This can be done via the GPU Control Utility. For Nvidia that is the "Nvidia Control Panel". You can use command line arguments for Automatic1111 to generate images quicker. I personally use these: "--upcast-sampling --no-half-vae --opt-split-attention --xformers". Note that --medvram & --lowvram are recommended often, though they reduce generation speed to lower vram usage. If you're after speed don't use them. If you're on a low vram card (less than 4 GB) try --medvram first. If you want to upscale it may be best to batch upscale images. Just restart Automatic1111 with different command line arguments (this time using --medvram) and batch upscale your generated images while you're away. This way you can swiftly improve upon your pompts without waiting for the upscaling and the slowness of the --medvram setting. If you have any other tips please tell them.
If you have any question feel free to ask away.

Automatic1111 Commandline Arguments Docs: https://github.com/AUTOMATIC1111/stab...
SD Webui Benchmark Data:
https://vladmandic.github.io/sd-exten...
_________________________________________________________________
#xformers #automatic1111 #stablediffusion
_________________________________________________________________
Danke!


Auf dieser Seite können Sie das Online-Video Stable Diffusion Performance Optimization Tutorial mit der Dauer stunde minuten sekunde in guter Qualität ansehen, das der Benutzer Archive 04 Juni 2023 hochgeladen hat, den Link mit Freunden und Bekannten teilen, dieses Video wurde auf Youtube bereits 46,248 Mal angesehen und es wurde von 681 den Zuschauern gefallen. Viel Spaß beim Betrachtenden Zuschauern gefallen!