File size: 9,922 Bytes
553cbc3
29c674a
632fa50
 
 
553cbc3
632fa50
 
553cbc3
 
29c674a
632fa50
 
 
 
 
 
 
29c674a
632fa50
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
---
title: DALL路E mini
emoji: 馃
colorFrom: yellow
colorTo: green
sdk: streamlit
app_file: app/streamlit/app.py
pinned: True
---

# DALL路E Mini

[![Join us on Discord](https://img.shields.io/discord/823813159592001537?color=5865F2&logo=discord&logoColor=white)](https://discord.gg/xBPBXfcFHd)

_Generate images from a text prompt_

<img src="https://github.com/borisdayma/dalle-mini/raw/main/img/logo.png" width="200">

Our logo was generated with DALL路E mini using the prompt "logo of an armchair in the shape of an avocado".

You can create your own pictures with [the demo](https://huggingface.co/spaces/flax-community/dalle-mini).

## How does it work?

Refer to [our report](https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini--Vmlldzo4NjIxODA).

## Inference Pipeline

To generate sample predictions and understand the inference pipeline step by step, refer to [`tools/inference/inference_pipeline.ipynb`](tools/inference/inference_pipeline.ipynb).

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/borisdayma/dalle-mini/blob/main/tools/inference/inference_pipeline.ipynb)

## Contributing

Join the community on the [DALLE-Pytorch Discord](https://discord.gg/xBPBXfcFHd).
Any contribution is welcome, from reporting issues to proposing fixes/improvements or testing the model with cool prompts!

## Development

### Dependencies Installation

For inference only, use `pip install git+https://github.com/borisdayma/dalle-mini.git`.

For development, clone the repo and use `pip install -e ".[dev]"`.
Before making a PR, check style with `make style`.

### Image Encoder

We use a VQGAN from [taming-transformers](https://github.com/CompVis/taming-transformers), which can also be fine-tuned.

Use [patil-suraj/vqgan-jax](https://github.com/patil-suraj/vqgan-jax) if you want to convert a checkpoint to JAX (does not support Gumbel).

Any image encoder that turns an image into a fixed sequence of tokens can be used.

### Training of DALL路E mini

Use [`tools/train/train.py`](tools/train/train.py).

You can also adjust the [sweep configuration file](https://docs.wandb.ai/guides/sweeps) if you need to perform a hyperparameter search.

## FAQ

### Where to find the latest models?

Trained models are on 馃 Model Hub:

- [VQGAN-f16-16384](https://huggingface.co/dalle-mini/vqgan_imagenet_f16_16384) for encoding/decoding images
- [DALL路E mini](https://huggingface.co/flax-community/dalle-mini) for generating images from a text prompt

### Where does the logo come from?

The "armchair in the shape of an avocado" was used by OpenAI when releasing DALL路E to illustrate the model's capabilities. Having successful predictions on this prompt represents a big milestone to us.

## Acknowledgements

- 馃 Hugging Face for organizing [the FLAX/JAX community week](https://github.com/huggingface/transformers/tree/master/examples/research_projects/jax-projects)
- Google [TPU Research Cloud (TRC) program](https://sites.research.google/trc/) for providing computing resources
- [Weights & Biases](https://wandb.com/) for providing the infrastructure for experiment tracking and model management

## Authors & Contributors

DALL路E mini was initially developed by:

- [Boris Dayma](https://github.com/borisdayma)
- [Suraj Patil](https://github.com/patil-suraj)
- [Pedro Cuenca](https://github.com/pcuenca)
- [Khalid Saifullah](https://github.com/khalidsaifullaah)
- [Tanishq Abraham](https://github.com/tmabraham)
- [Ph煤c L锚 Kh岷痗](https://github.com/lkhphuc)
- [Luke Melas](https://github.com/lukemelas)
- [Ritobrata Ghosh](https://github.com/ghosh-r)

Many thanks to the people who helped make it better:

- the [DALLE-Pytorch](https://discord.gg/xBPBXfcFHd) and [EleutherAI](https://www.eleuther.ai/) communities for testing and exchanging cool ideas
- [Rohan Anil](https://github.com/rohan-anil) for adding Distributed Shampoo optimizer
- [Phil Wang](https://github.com/lucidrains) has provided a lot of cool implementations of transformer variants and gives interesting insights with [x-transformers](https://github.com/lucidrains/x-transformers)
- [Katherine Crowson](https://github.com/crowsonkb) for [super conditioning](https://twitter.com/RiversHaveWings/status/1478093658716966912)

## Citing DALL路E mini

If you find DALL路E mini useful in your research or wish to refer, please use the following BibTeX entry.

```text
@misc{Dayma_DALL路E_Mini_2021,
      author = {Dayma, Boris and Patil, Suraj and Cuenca, Pedro and Saifullah, Khalid and Abraham, Tanishq and L锚 Kh岷痗, Ph煤c and Melas, Luke and Ghosh, Ritobrata},
      doi = {10.5281/zenodo.5146400},
      month = {7},
      title = {DALL路E Mini},
      url = {https://github.com/borisdayma/dalle-mini},
      year = {2021}
}
```

## References

Original DALL路E from "[Zero-Shot Text-to-Image Generation](https://arxiv.org/abs/2102.12092)" with image quantization from "[Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020)".

Image encoder from "[Taming Transformers for High-Resolution Image Synthesis](https://arxiv.org/abs/2012.09841v2)".

Sequence to sequence model based on "[BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461v1)" with implementation of a few variants:

- "[GLU Variants Improve Transformer](https://arxiv.org/abs/2002.05202)"
- "[Deepnet: Scaling Transformers to 1,000 Layers](https://arxiv.org/abs/2203.00555)"
- "[NormFormer: Improved Transformer Pretraining with Extra Normalization](https://arxiv.org/abs/2110.09456)"
- "[Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030)"
- "[CogView: Mastering Text-to-Image Generation via Transformers](https://arxiv.org/abs/2105.13290v2)"
- "[Root Mean Square Layer Normalization](https://arxiv.org/abs/1910.07467)"
- "[Sinkformers: Transformers with Doubly Stochastic Attention](https://arxiv.org/abs/2110.11773)"

Main optimizer (Distributed Shampoo) from "[Scalable Second Order Optimization for Deep Learning](https://arxiv.org/abs/2002.09018)".

### Citations

```text
@misc{
  title={Zero-Shot Text-to-Image Generation}, 
  author={Aditya Ramesh and Mikhail Pavlov and Gabriel Goh and Scott Gray and Chelsea Voss and Alec Radford and Mark Chen and Ilya Sutskever},
  year={2021},
  eprint={2102.12092},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}
```

```text
@misc{
  title={Learning Transferable Visual Models From Natural Language Supervision}, 
  author={Alec Radford and Jong Wook Kim and Chris Hallacy and Aditya Ramesh and Gabriel Goh and Sandhini Agarwal and Girish Sastry and Amanda Askell and Pamela Mishkin and Jack Clark and Gretchen Krueger and Ilya Sutskever},
  year={2021},
  eprint={2103.00020},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}
```

```text
@misc{
  title={Taming Transformers for High-Resolution Image Synthesis}, 
  author={Patrick Esser and Robin Rombach and Bj枚rn Ommer},
  year={2021},
  eprint={2012.09841},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}
```

```text
@misc{
  title={BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension}, 
  author={Mike Lewis and Yinhan Liu and Naman Goyal and Marjan Ghazvininejad and Abdelrahman Mohamed and Omer Levy and Ves Stoyanov and Luke Zettlemoyer},
  year={2019},
  eprint={1910.13461},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}
```

```text
@misc{
  title={Scalable Second Order Optimization for Deep Learning},
  author={Rohan Anil and Vineet Gupta and Tomer Koren and Kevin Regan and Yoram Singer},
  year={2021},
  eprint={2002.09018},
  archivePrefix={arXiv},
  primaryClass={cs.LG}
}
```

```text
@misc{
  title={GLU Variants Improve Transformer},
  author={Noam Shazeer},
  year={2020},
  url={https://arxiv.org/abs/2002.05202}    
}
```

```text
 @misc{
  title={DeepNet: Scaling transformers to 1,000 layers},
  author={Wang, Hongyu and Ma, Shuming and Dong, Li and Huang, Shaohan and Zhang, Dongdong and Wei, Furu},
  year={2022},
  eprint={2203.00555}
  archivePrefix={arXiv},
  primaryClass={cs.LG}
} 
```

```text
@misc{
  title={NormFormer: Improved Transformer Pretraining with Extra Normalization},
  author={Sam Shleifer and Jason Weston and Myle Ott},
  year={2021},
  eprint={2110.09456},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}
```

```text
@inproceedings{
  title={Swin Transformer V2: Scaling Up Capacity and Resolution}, 
  author={Ze Liu and Han Hu and Yutong Lin and Zhuliang Yao and Zhenda Xie and Yixuan Wei and Jia Ning and Yue Cao and Zheng Zhang and Li Dong and Furu Wei and Baining Guo},
  booktitle={International Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2022}
}
```

```text
@misc{
  title = {CogView: Mastering Text-to-Image Generation via Transformers},
  author = {Ming Ding and Zhuoyi Yang and Wenyi Hong and Wendi Zheng and Chang Zhou and Da Yin and Junyang Lin and Xu Zou and Zhou Shao and Hongxia Yang and Jie Tang},
  year = {2021},
  eprint = {2105.13290},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV}
}
```

```text
@misc{
  title = {Root Mean Square Layer Normalization},
  author = {Biao Zhang and Rico Sennrich},
  year = {2019},
  eprint = {1910.07467},
  archivePrefix = {arXiv},
  primaryClass = {cs.LG}
}
```

```text
@misc{
  title = {Sinkformers: Transformers with Doubly Stochastic Attention},
  url = {https://arxiv.org/abs/2110.11773},
  author = {Sander, Michael E. and Ablin, Pierre and Blondel, Mathieu and Peyr茅, Gabriel},
  publisher = {arXiv},
  year = {2021},
}
```

```text
@misc{
  title = {Smooth activations and reproducibility in deep networks},
  url = {https://arxiv.org/abs/2010.09931},
  author = {Shamir, Gil I. and Lin, Dong and Coviello, Lorenzo},
  publisher = {arXiv},
  year = {2020},
}
```