Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,107 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: tencent-hunyuan-community
|
4 |
+
license_link: https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/blob/main/LICENSE.txt
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
---
|
8 |
+
<!-- ## **HunyuanDiT** -->
|
9 |
+
|
10 |
+
<p align="center">
|
11 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/logo.png" height=100>
|
12 |
+
</p>
|
13 |
+
|
14 |
+
# Hunyuan-DiT : A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding
|
15 |
+
# 混元-DiT: 具有细粒度中文理解的多分辨率Diffusion Transformer
|
16 |
+
|
17 |
+
[[Arxiv]](https://arxiv.org/abs/2405.08748) [[project page]](https://dit.hunyuan.tencent.com/) [[github]](https://github.com/Tencent/HunyuanDiT)
|
18 |
+
|
19 |
+
This repo contains the distilled Hunyuan-DiT in 🤗 [Diffusers](https://github.com/huggingface/diffusers) format.
|
20 |
+
|
21 |
+
It supports 25-step text-to-image generation.
|
22 |
+
|
23 |
+
## Dependency
|
24 |
+
Please install PyTorch first, following the instruction in [https://pytorch.org](https://pytorch.org)
|
25 |
+
|
26 |
+
Install the latest version of transformers with `pip`:
|
27 |
+
```
|
28 |
+
pip install --upgrade transformers
|
29 |
+
```
|
30 |
+
|
31 |
+
Then install the latest github version of 🤗 Diffusers with `pip`:
|
32 |
+
```
|
33 |
+
pip install git+https://github.com/huggingface/diffusers.git
|
34 |
+
```
|
35 |
+
|
36 |
+
## Example Usage with 🤗 Diffusers
|
37 |
+
```py
|
38 |
+
import torch
|
39 |
+
from diffusers import HunyuanDiTPipeline
|
40 |
+
|
41 |
+
pipe = HunyuanDiTPipeline.from_pretrained("Tencent-Hunyuan/HunyuanDiT-v1.2-Diffusers-Distilled", torch_dtype=torch.float16)
|
42 |
+
pipe.to("cuda")
|
43 |
+
|
44 |
+
# You may also use English prompt as HunyuanDiT supports both English and Chinese
|
45 |
+
# prompt = "An astronaut riding a horse"
|
46 |
+
prompt = "一个宇航员在骑马"
|
47 |
+
image = pipe(prompt).images[0]
|
48 |
+
```
|
49 |
+
|
50 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/646b0bbdec9a61e871799339/xoO_-5N7eZ-aCt4KpBYY6.png)
|
51 |
+
|
52 |
+
## 📈 Comparisons
|
53 |
+
In order to comprehensively compare the generation capabilities of HunyuanDiT and other models, we constructed a 4-dimensional test set, including Text-Image Consistency, Excluding AI Artifacts, Subject Clarity, Aesthetic. More than 50 professional evaluators performs the evaluation.
|
54 |
+
|
55 |
+
<p align="center">
|
56 |
+
<table>
|
57 |
+
<thead>
|
58 |
+
<tr>
|
59 |
+
<th rowspan="2">Model</th> <th rowspan="2">Open Source</th> <th>Text-Image Consistency (%)</th> <th>Excluding AI Artifacts (%)</th> <th>Subject Clarity (%)</th> <th rowspan="2">Aesthetics (%)</th> <th rowspan="2">Overall (%)</th>
|
60 |
+
</tr>
|
61 |
+
</thead>
|
62 |
+
<tbody>
|
63 |
+
<tr>
|
64 |
+
<td>SDXL</td> <td> ✔ </td> <td>64.3</td> <td>60.6</td> <td>91.1</td> <td>76.3</td> <td>42.7</td>
|
65 |
+
</tr>
|
66 |
+
<tr>
|
67 |
+
<td>PixArt-α</td> <td> ✔</td> <td>68.3</td> <td>60.9</td> <td>93.2</td> <td>77.5</td> <td>45.5</td>
|
68 |
+
</tr>
|
69 |
+
<tr>
|
70 |
+
<td>Playground 2.5</td> <td>✔</td> <td>71.9</td> <td>70.8</td> <td>94.9</td> <td>83.3</td> <td>54.3</td>
|
71 |
+
</tr>
|
72 |
+
|
73 |
+
<tr>
|
74 |
+
<td>SD 3</td> <td>✘</td> <td>77.1</td> <td>69.3</td> <td>94.6</td> <td>82.5</td> <td>56.7</td>
|
75 |
+
|
76 |
+
</tr>
|
77 |
+
<tr>
|
78 |
+
<td>MidJourney v6</td><td>✘</td> <td>73.5</td> <td>80.2</td> <td>93.5</td> <td>87.2</td> <td>63.3</td>
|
79 |
+
</tr>
|
80 |
+
<tr>
|
81 |
+
<td>DALL-E 3</td><td>✘</td> <td>83.9</td> <td>80.3</td> <td>96.5</td> <td>89.4</td> <td>71.0</td>
|
82 |
+
</tr>
|
83 |
+
<tr style="font-weight: bold; background-color: #f2f2f2;">
|
84 |
+
<td>Hunyuan-DiT</td><td>✔</td> <td>74.2</td> <td>74.3</td> <td>95.4</td> <td>86.6</td> <td>59.0</td>
|
85 |
+
</tr>
|
86 |
+
</tbody>
|
87 |
+
</table>
|
88 |
+
</p>
|
89 |
+
|
90 |
+
## 🎥 Visualization
|
91 |
+
|
92 |
+
* **Chinese Elements**
|
93 |
+
<p align="center">
|
94 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/chinese elements understanding.png" height=220>
|
95 |
+
</p>
|
96 |
+
|
97 |
+
* **Long Text Input**
|
98 |
+
|
99 |
+
|
100 |
+
<p align="center">
|
101 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/long text understanding.png" height=310>
|
102 |
+
</p>
|
103 |
+
|
104 |
+
|
105 |
+
## 🔥🔥🔥 Tencent Hunyuan Bot
|
106 |
+
|
107 |
+
Welcome to [Tencent Hunyuan Bot](https://hunyuan.tencent.com/bot/chat), where you can explore our innovative products in multi-round conversation!
|