search

open-mmlab / PIA

[CVPR 2024] PIA, your Personalized Image Animator. Animate your images by text prompt, combing with Dreambooth, achieving stunning videos. PIA，你的个性化图像动画生成器，利用文本提示将图像变为奇妙的动画
https://pi-animator.github.io/
Apache License 2.0
888 stars 70 forks source link

demo give error #12

Closed loboere closed 9 months ago

loboere commented 9 months ago

i try in 1024 image but give error

LeoXing1996 commented 9 months ago

Hey @loboere, thanks for your interest in our project. Which demo are you referring to? Openxlab's online demo, or the Gradio demo running on your local machine?

If you mean the Openxlab demo, could you please provide the input image you used? If you are referring to the local Gradio demo, could you provide the error traceback?

LeoXing1996 commented 9 months ago

Hey @loboere, we now support scaled_dot_product_attention for memory-efficient attention. Now you can animate 1024x1024 image with 16GB gpu memory. Please pull our newest code and try this feature. If you still have any problem, feel free to re-open this issue.