donutn
donutn
may i ask if you have solved this problem i often report mistake "out of memory" could you tell what is the result of running this script
Thanks for your reply. It really seemed not well when i just scaling the resolution of the input image.
> 你的环境是多卡的吗?运行`python -c "import torch; print(torch.cuda.device_count())"`看看? 对的,是8张,但我运行之后的run.py是没问题的,可以在24G显卡上跑完全部的,可能是app.py,部署gradio需要显卡一些额外的资源(oom主要是在重建时
> 我这边没有这样的观察,可以发几张测试图吗?  比如代码里这张图,DAPE提取的提示也是“baseball glove, currency, nebula, woodpecker, ,clean, high-resolution, 8k”,您测试时的随机数种子是固定了嘛,我去看了代码部分发现这几个tag对应的概率低于平均,是这个问题吗