metadata
license: creativeml-openrail-m
language:
- en
library_name: diffusers
pipeline_tag: text-to-image
tags:
- art
- tensorRT
Model Card for lyraSD
lyraSD is currently the fastest Stable Diffusion model available, boasting an inference cost of only 0.435 seconds for a 512x512 image, accelerating the process up to 10 times faster than the original version.
Among its main features are:
- weights: original Stable Diffusion 1.4 weights
- input image: 512x512 (in img2img mode)
- output image: 512x512
- device requirements: Nvidia Ampere architecture (A100) or compatable
- super-resultion: 4x by default, optional.
Speed
test environment
- device: Nvidia A100 40G
- img size: 512x512
- percision:fp16
- steps: 30
- solver: LMSD
text2img
| model | time cost(ms) | memory(MB) |
|---|---|---|
| Pytorch SD | ~5000ms | ~10240 |
| lyraSD | ~435ms | ~4026 |
superResolution(SR)
| model | time cost(ms) | memory(MB) |
|---|---|---|
| Pytorch SR | ~720ms | ~6650 |
| lyraSD | ~26ms | ~1600 |
Model Sources
Uses
from lyraSD import LyraSD
t2imodel = LyraSD("text2img", "./sd1.4-engine")
t2imodel.inference(prompt="A fantasy landscape, trending on artstation", use_super=False)
from PIL import Image
i2imodel = LyraSD("img2img", "./sd1.4-engine")
demo_img = Image.open("output/text2img_demo.jpg")
i2imodel.inference(prompt="A fantasy landscape, trending on artstation", image=demo_img)
Demo output
text2img
img2img
Environment
- hardware: Nvidia Ampere architecture (A100) or compatable
- docker image avaible: https://hub.docker.com/r/bigmoyan/lyra_aigc/tags
docker pull bigmoyan/lyra_aigc:v0.1
Citation
@Misc{lyraSD2023,
author = {Kangjian Wu, Zhengtao Wang, Bin Wu},
title = {lyraSD: Accelerating Stable Diffusion by 10x+},
howpublished = {\url{https://huggingface.co/TMElyralab/lyraSD}},
year = {2023}
}
Report bug
- start a discussion to report any bugs!--> https://huggingface.co/TMElyralab/lyraSD/discussions
- report bug with a
[bug]mark in the title.