Alex Li

It seems like it's not possible, since the VQ-GAN downsamples in the time dimension.

How are you fine-tuning your diffusion model?

What is the accuracy using default SD2.0 vs your LORA fine-tuned version? Are you sure you're changing the prompts correctly? For the DiT we used in the paper, we downloaded...

To get calibrated probabilities, you need to find the correct temperature scaling. If you have a small amount of training data, you can check out the approach in [this paper...

Hi Zhangzhi, Thanks for bringing this up! I am a big fan of the [score SDE ](https://arxiv.org/abs/2011.13456) and [FFJORD](https://arxiv.org/abs/1810.01367) papers. I did look into treating the diffusion model as a...