4 ms·
SDXL is already very very slow when compared to SD 1.5. They are claiming 200ms for 512x512 image in SDXL on A100. We need SD 1.5 turbo for even faster generati
by smusamashah 3y ago
SDXL is already very very slow when compared to SD 1.5. They are claiming 200ms for 512x512 image in SDXL on A100. We need SD 1.5 turbo for even faster generation.
- ShamelessC 3y agoIt _would_ be nice if they offered 512/768/1024 px variants of the models. I frequently don’t actually need the full 1024 px as it just needs to look good enough for a chat thumbnail. Then I could upscale it manually later. There’s other models like Kandinsky but it’s not super convenient to use multiple models with different code and what not.
- Sharlin 3y agoI haven't found SDXL to be inherently much slower than 1.5, besides the obvious 4x slowdown from having twice the linear resolution.
- thot_experiment 3y agoYeah, in my experience it's actually FASTER because I was already doing high res gens, and that required 2 or 3 passes previously.
- minimaxir 3y agoSDXL Turbo also uses a distilled version of SDXL so it gets a speed bonus from that too.
- swyx 3y agois it known how much larger SDXL is compared to SD1 or SD2?
- dragonwriter 3y agoSD1.x is around 1.1B parameters including VAE, SD2.x is slightly more (uses the same UNet and VAE, but a bigger text encoder; not finding stats as quickly as I'd like), and SDXL is 3.5B parameters (single model only, but StabilityAI's preferred base + refiner model setup is effectively 6.6B parameters -- some bits are shared.)
- dragonwriter 3y agoWell, its 2.1 instead of 1.5, but: https://huggingface.co/stabilityai/sd-turbo https://huggingface.co/stabilityai/sd-turbo I assume that this was ready for release because Stable Video Diffusion (which is also an SD2.1-based model) is essentially this plus a motion model.) I wouldn't be surprised if their hosted-only SD1.6 beta is, or has, a turbo version, and if that gets released publicly, that's where we'll see an SD1.x turbo.