Ma Xinyin

Hi @bigmover , if the SD models fine-tuned/retrained by diffusers still have the same structure and module name, then you can follow the [general instructions](https://github.com/horseee/DeepCache?tab=readme-ov-file#usage) to employ DeepCache. However, if...

Hi, thanks for your interest in our paper! In the updated camera-ready version, we included a new section showing the text-to-video results. We believe there might be unexplored things for...

Hi, @Barry0121 The code in Colab is outdated and I have updated it. Feel free to give it a try and see the results. Thanks for the gentle reminder! 😊

The calculation is on the latent space, and we resize it back to the pixel space.

Hi @jyoung105, DeepCache is a dynamic model inference algorithm, which makes it incompatible with torch.compile(). One possible solution (though I'm not certain) is to split the entire pipeline into two...

Hi @ledrose , We tested the issue you mentioned and found that the possible cause of this issue is a conflict between the cpu offloa and our implementation of deepcache...

Hi @weirdo0913 , Yes. You can check the instructions [here](https://github.com/horseee/DeepCache/tree/master/experiments#experiment-code-for-ddpm)

Hi @Barry0121, Q1: The variable responsible for storing cached features is named as `prv_features` at Line 723 and undergoes updates at Line 753. To understand the mechanism of feature caching,...

Hi @Barry0121, That sounds like an exciting project! Thanks for your great work on this, and I added the link to the section: [Other Implementations and Plugins](https://github.com/horseee/DeepCache?tab=readme-ov-file#other-implementations-and-plugins)

Hi @l-dawei, Could you check which UNet2DConditionModel the code would call? The one in diffusers or the one in DeepCache/sdxl/unet_2d_condition.py