search

naver-ai / Visual-Style-Prompting

Official Pytorch implementation of "Visual Style Prompting with Swapping Self-Attention"
https://curryjung.github.io/VisualStylePrompt/
Apache License 2.0
408 stars 30 forks source link

inference is so slow #5

Closed akk-123 closed 5 months ago

akk-123 commented 5 months ago

inference a img cost 4 min?

taki0112 commented 5 months ago

Based on the V100 GPU, it takes about 30sec. (preset image) If you want to reduce the time, try lowering the diffusion step from 50 to about 30.

if you want a user_image, takes 2min