Update README.md
Browse files
README.md
CHANGED
|
@@ -1,7 +1,8 @@
|
|
| 1 |
# Datasets for the Direct Preference for Denoising Diffusion Policy Optimization (D3PO)
|
| 2 |
|
| 3 |
**Description**: The dataset for the image distortion experiment of the [`anything-v5`](https://huggingface.co/stablediffusionapi/anything-v5) model in the paper [Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model](https://arxiv.org/abs/2311.13231).
|
| 4 |
-
|
|
|
|
| 5 |
**Source Code**: The code used to generate this data can be found [here](https://github.com/yk7333/D3PO/tree/main).
|
| 6 |
|
| 7 |
**Directory**
|
|
@@ -15,10 +16,17 @@
|
|
| 15 |
- data.json (required for training)
|
| 16 |
- prompt.json
|
| 17 |
- sample.pkl(required for training)
|
| 18 |
-
- epoch2
|
| 19 |
- ...
|
| 20 |
- epoch5
|
| 21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
**Citation**
|
| 23 |
```
|
| 24 |
@article{yang2023using,
|
|
|
|
| 1 |
# Datasets for the Direct Preference for Denoising Diffusion Policy Optimization (D3PO)
|
| 2 |
|
| 3 |
**Description**: The dataset for the image distortion experiment of the [`anything-v5`](https://huggingface.co/stablediffusionapi/anything-v5) model in the paper [Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model](https://arxiv.org/abs/2311.13231).
|
| 4 |
+
(2024.1.22 Update: Add the dataset for evaluating text and image alignment before and after fine-tuning.)
|
| 5 |
+
|
| 6 |
**Source Code**: The code used to generate this data can be found [here](https://github.com/yk7333/D3PO/tree/main).
|
| 7 |
|
| 8 |
**Directory**
|
|
|
|
| 16 |
- data.json (required for training)
|
| 17 |
- prompt.json
|
| 18 |
- sample.pkl(required for training)
|
| 19 |
+
- epoch2`
|
| 20 |
- ...
|
| 21 |
- epoch5
|
| 22 |
|
| 23 |
+
|
| 24 |
+
- text2img_dataset:
|
| 25 |
+
- img
|
| 26 |
+
- data_*.json
|
| 27 |
+
- plot.ipynb
|
| 28 |
+
- prompt.txt
|
| 29 |
+
|
| 30 |
**Citation**
|
| 31 |
```
|
| 32 |
@article{yang2023using,
|