Initial update for Pick-Style cartoon split
Browse filesApplied [this template](https://huggingface.co/mapo-t2i/mapo-beta/discussions/1) as suggested. Please feel free to give some feedback!
README.md
CHANGED
@@ -1,4 +1,6 @@
|
|
1 |
---
|
|
|
|
|
2 |
dataset_info:
|
3 |
features:
|
4 |
- name: caption
|
@@ -23,3 +25,38 @@ configs:
|
|
23 |
- split: train
|
24 |
path: data/train-*
|
25 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
license: openrail++
|
3 |
+
library_name: diffusers
|
4 |
dataset_info:
|
5 |
features:
|
6 |
- name: caption
|
|
|
25 |
- split: train
|
26 |
path: data/train-*
|
27 |
---
|
28 |
+
|
29 |
+
# Margin-aware Preference Optimization for Aligning Diffusion Models without Reference
|
30 |
+
|
31 |
+
<div align="center">
|
32 |
+
<img src="assets/mapo_overview.png" width=750/>
|
33 |
+
</div><br>
|
34 |
+
|
35 |
+
We propose **MaPO**, a reference-free, sample-efficient, memory-friendly alignment technique for text-to-image diffusion models. For more details on the technique, please refer to our paper [here] (TODO).
|
36 |
+
|
37 |
+
## Developed by
|
38 |
+
|
39 |
+
* Jiwoo Hong<sup>*</sup> (KAIST AI)
|
40 |
+
* Sayak Paul<sup>*</sup> (Hugging Face)
|
41 |
+
* Noah Lee (KAIST AI)
|
42 |
+
* Kashif Rasul (Hugging Face)
|
43 |
+
* James Thorne (KAIST AI)
|
44 |
+
* Jongheon Jeong (Korea University)
|
45 |
+
|
46 |
+
## Dataset
|
47 |
+
|
48 |
+
This dataset is *cartoon* split of **Pick-Style**, self-curated with [Stable Diffusion XL](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0). Using the context prompts (i.e., without stylistic specifications), we generate (1) cartoon style generation with stylistic prefix prompt and (2) normal generation with context prompt. Then, (1) is used as the chosen image, and (2) as the rejected image.
|
49 |
+
|
50 |
+
## Citation
|
51 |
+
|
52 |
+
```bibtex
|
53 |
+
@misc{todo,
|
54 |
+
title={Margin-aware Preference Optimization for Aligning Diffusion Models without Reference},
|
55 |
+
author={Jiwoo Hong and Sayak Paul and Noah Lee and Kashif Rasuland James Thorne and Jongheon Jeong},
|
56 |
+
year={2024},
|
57 |
+
eprint={todo},
|
58 |
+
archivePrefix={arXiv},
|
59 |
+
primaryClass={cs.CV,cs.LG}
|
60 |
+
}
|
61 |
+
```
|
62 |
+
|