File size: 3,234 Bytes
14a581c 3bd8537 14a581c 130b747 93c6fb2 130b747 f8178b3 80511fe 130b747 8765927 9347c75 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 |
---
license: mit
pipeline_tag: text-to-video
tags:
- Text-to-Video
---
<div align="center">
<!-- TITLE -->
# **Video Diffusion Alignment via Reward Gradient**
[](http://vader-vid.github.io)
</div>
For more information on how to use it, please check [GitHub](https://github.com/mihirp1998/VADER).
This is a checkpoint trained via PickScore Reward Model with a LoRA rank of 16 on prompts in `assets/chatgpt_custom_animal.txt` for our paper [Video Diffusion Alignment via Reward Gradient](https://vader-vid.github.io/) by
Mihir Prabhudesai*, Russell Mendonca*, Zheyang Qin*, Katerina Fragkiadaki, Deepak Pathak.
<!-- DESCRIPTION -->
## Abstract
We have made significant progress towards building foundational video diffusion models. As these models are trained using large-scale unsupervised data, it has become crucial to adapt these models to specific downstream tasks, such as video-text alignment or ethical video generation. Adapting these models via supervised fine-tuning requires collecting target datasets of videos, which is challenging and tedious. In this work, we instead utilize pre-trained reward models that are learned via preferences on top of powerful discriminative models. These models contain dense gradient information with respect to generated RGB pixels, which is critical to be able to learn efficiently in complex search spaces, such as videos. We show that our approach can enable alignment of video diffusion for aesthetic generations, similarity between text context and video, as well long horizon video generations that are 3X longer than the training sequence length. We show our approach can learn much more efficiently in terms of reward queries and compute than previous gradient-free approaches for video generation.
## Demo
| | | |
| ----------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- |
| <img src="assets/videos/8.gif" width=""> | <img src="assets/videos/5.gif" width=""> | <img src="assets/videos/7.gif" width=""> |
| <img src="assets/videos/10.gif" width=""> | <img src="assets/videos/3.gif" width=""> | <img src="assets/videos/4.gif" width=""> |
| <img src="assets/videos/9.gif" width=""> | <img src="assets/videos/1.gif" width=""> | <img src="assets/videos/11.gif" width=""> |
## Citation
If you find this work useful in your research, please cite:
```bibtex
@misc{prabhudesai2024videodiffusionalignmentreward,
title={Video Diffusion Alignment via Reward Gradients},
author={Mihir Prabhudesai and Russell Mendonca and Zheyang Qin and Katerina Fragkiadaki and Deepak Pathak},
year={2024},
eprint={2407.08737},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2407.08737},
}
``` |