Wenliang Zhao

Thanks for your help! I think you are right and fixing the bug perhaps will make the training better. I will conduct experiments later and update the results.

Hi, Could you please provide more details about how you tried to reproduce (e.g., dataset, method)? About the visualization, we simply visualize the [cross-attention map](https://github.com/wl-zhao/DIML/blob/c15dbea696f68ddf889dcacfcaacd315d16a34ac/criteria/margin_diml.py#L78) and some representative values of...

The test result of each stream is saved as python dict, where the order of items will be ignored. To ensemble the streams correctly, you can do either of the...

Hi, thanks for your efforts! We are glad to see UniPC can also achieve good results in stable-diffusion-webui project. By the way, UniPC has already been integrated into diffusers, you...