Zeqiang Lai

Your torchvision is too old, try to upgrade the torchvison via ``` pip install torchvision --upgrade ```

https://github.com/Zeqiang-Lai/DragGAN/issues/60 需要转换格式

欧欧不好意思没注意看，应该是draggan里的模型某些网络层的默认channel数和ckpt对不上，要对应改下。 训练的时候我建议是follow stylegan2 ada与放出ckpt的模型相同的网络配置，这样应该就可以避免不一致的问题。 我这几天有空的时候会增加一个无需转换ckpt的版本，到时候就会很简单了。

stylegan ada 官方没有256的ckpt，所以我也不确定能不能找到合适的config。 256分辨率通过转换ckpt不修改代码就能够正确load的模型这里有 https://github.com/self-distilled-stylegan/self-distilled-internet-photos 和他们的256模型对齐配置应该就可以了

根据其他issuse的反馈是代理问题，关闭代理可能就好了。 https://github.com/Zeqiang-Lai/DragGAN/issues/50

Fixed with https://github.com/Zeqiang-Lai/DragGAN/pull/78

pip install gradio==3.28.1

The download url is fixed at version 1.1.5. Colab should work now as well.

Thanks for the contribution. As we are working on the next version, this pull request might not be able to be merged due to the conflict. Feel free to resolve...

> it takes time to load the plugin for the first time. if ubuntu complains about app not responding, click wait. How long does it take in your cases ?...