context_shift怎么使用

[hhzzlll]

在ContextDiff_finetune.py文件中虽然传入了context_shift为True

但在video_diffusion.pipelines.ddim_spatial_temporal.DDIMSpatioTemporalStableDiffusionPipeline的prepare_latents_ddim_inverted函数中却将context_shift注释了，这是什么意思，有使用到context_shift吗

以及ddim_clean2noisy_loop函数也将context_shift注释了

[ZhaochenYu0201]

In our paper, during the process of one-shot tuning for video editing, we utilize an adapter that has been trained on the LAION dataset during the text-to-image stage. We then fine-tune this pretrained adapter with the newly introduced spatio-temporal information from the original video, leading to further enhancement of the desired effects. However, for the sake of convenience in open-sourcing, the current version available is trained on a new adapter based on the original video. Due to the fact that one-shot tuning only involves the original video and its corresponding prompt data, the retrained adapter has a minimal impact on the semantic aspects. Therefore, in the previously open-sourced version, we did not enable it. We have updated our repo, you can directly use it now.

[hhzzlll]

很感谢您能回答我的问题，我还有一些问题想请教一下您，论文里的公式13在代码中哪里体现出来，您有空的时候麻烦帮我指出来一下，谢谢！