search

glucauze / sd-webui-faceswaplab

Extended faceswap extension for StableDiffusion web-ui with multiple faceswaps, inpainting, checkpoints, ....
https://glucauze.github.io/sd-webui-faceswaplab/
GNU Affero General Public License v3.0
707 stars 92 forks source link

Excluding the model loading time, how long does inference take on T4 GPU? #128

Closed DaeseungLee closed 9 months ago

DaeseungLee commented 9 months ago

Hello!

I want to optimize inference time. How long does inference take on T4 GPU for single image?

I takes 14 seconds excluding loading model. There are any technique to reduce the time?

Thanks!

glucauze commented 9 months ago

Did you check GPU is really used and CUDA provider is selected ? It should be enabled on settings too. That's surprisingly slow. Without upscaling it should be less than 2s. It is the case on my RTX3080ti. I don't have experience with T4. I will check on a M40 if i have time.