Speed Slow Comparative to SDXL Normal???

[verifiedsyckozis]

This is taking minutes on my setup, which isn't much, but I get less than a minute with SDXL and about 3 minutes full trip 1st stage and 2nd stage both. Sometimes 7. I'm wondering if with 8GB V 4070 would result in about 3-4 it/s at best or if something is wrong and if something is wrong how would I fix it? Any help appreciated. No help understood.

Have a great day!

[SoftologyPro]

For a comparison, on a 24GB 4090 I can generate an image in 30 seconds. That includes loading the models and generating both stages. This is on Windows using the same code/models from the text_to_image.ipynb notebook.

[FurkanGozukara]

with my gradio app RTX 4090 able to generate images in 4-5 seconds. faster than SDXL. I can say like 80% faster for 1024x1024. for 2048x2048 many many times faster.

the example code provided in notebook is like 6 times slower than diffusers pipeline with optimization which my gradio app uses

[laolongboy]

with my gradio app RTX 4090 able to generate images in 4-5 seconds. faster than SDXL. I can say like 80% faster for 1024x1024. for 2048x2048 many many times faster.我的 Gradio 应用程序 RTX 4090 能够在 4-5 秒内生成图像。比 SDXL 更快。我可以说 1024x1024 的速度快了 80%。 2048x2048 的速度要快很多倍。关闭

the example code provided in notebook is like 6 times slower than diffusers pipeline with optimization which my gradio app uses笔记本中提供的示例代码比我的渐变应用程序使用的经过优化的扩散器管道慢 6 倍关闭

cool, will you share your gradio code? the example code is really slow.

[FurkanGozukara]

with my gradio app RTX 4090 able to generate images in 4-5 seconds. faster than SDXL. I can say like 80% faster for 1024x1024. for 2048x2048 many many times faster.我的 Gradio 应用程序 RTX 4090 能够在 4-5 秒内生成图像。比 SDXL 更快。我可以说 1024x1024 的速度快了 80%。 2048x2048 的速度要快很多倍。关闭 the example code provided in notebook is like 6 times slower than diffusers pipeline with optimization which my gradio app uses笔记本中提供的示例代码比我的渐变应用程序使用的经过优化的扩散器管道慢 6 倍关闭

cool, will you share your gradio code? the example code is really slow.

Hello.

Currently shared here with including RunPod installer and also Kaggle notebook - working with gradio as well on a free kaggle account

https://www.patreon.com/posts/stable-cascade-1-98410661

You can also see my sharings here : https://www.reddit.com/r/StableDiffusion/comments/1aqbydi/stable_cascade_prompt_following_is_amazing_this/