File size: 15,833 Bytes
22e6485
 
ed14162
758b1b3
ed14162
 
 
 
22e6485
ed14162
 
e8e7a23
ed14162
 
 
 
 
1506f97
 
 
c4334b4
 
46fbfc8
 
1506f97
46fbfc8
 
1506f97
ed14162
98b008f
ed14162
98b008f
ed14162
 
 
98b008f
ed14162
 
 
 
 
 
 
 
831b863
 
 
 
ed14162
 
 
 
 
 
 
 
 
 
 
 
831b863
ed14162
 
 
 
 
 
831b863
 
ed14162
 
 
 
 
 
 
 
 
 
 
 
 
831b863
ed14162
 
 
 
 
831b863
 
ed14162
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
831b863
ed14162
831b863
 
ed14162
 
 
 
e8e7a23
ed14162
 
 
 
 
 
 
831b863
ed14162
 
 
 
 
 
831b863
 
ed14162
 
 
 
e8e7a23
ed14162
 
 
 
 
 
 
 
831b863
ed14162
 
 
 
 
831b863
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6856747
1a091a0
 
 
46fbfc8
 
 
831b863
ed14162
 
e8e7a23
 
 
 
 
 
 
 
ed14162
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
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
---
license: openrail++
library_name: diffusers
inference: false
tags:
  - lora
  - text-to-image
  - stable-diffusion
---

# Hyper-SD
Official Repository of the paper: *[Hyper-SD](https://arxiv.org/abs/2404.13686)*.

Project Page: https://hyper-sd.github.io/

![](./hypersd_tearser.jpg)


## News🔥🔥🔥

* Apr.26, 2024. 💥💥💥 Our CFG-Preserved Hyper-SD15/SDXL that facilitate negative prompts and larger guidance scales (e.g. 5~10) will be coming soon!!! 💥💥💥
* Apr.26, 2024. Thanks to @[Pete](https://huggingface.co/pngwn) for contributing to our [scribble demo](https://huggingface.co/spaces/ByteDance/Hyper-SD15-Scribble) with larger canvas right now 👏.
* Apr.24, 2024. The ComfyUI [workflow](https://huggingface.co/ByteDance/Hyper-SD/blob/main/comfyui/Hyper-SDXL-1step-Unet-workflow.json) and [checkpoint](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SDXL-1step-Unet-Comfyui.fp16.safetensors) on 1-Step SDXL UNet ✨ is also available! Don't forget ⭕️ to install the custom [scheduler](https://huggingface.co/ByteDance/Hyper-SD/tree/main/comfyui/ComfyUI-HyperSDXL1StepUnetScheduler) in your `ComfyUI/custom_nodes` folder!!!
* Apr.23, 2024. ComfyUI workflows on N-Steps LoRAs are [released](https://huggingface.co/ByteDance/Hyper-SD/tree/main/comfyui)! Worth a try for creators 💥!
* Apr.23, 2024. Our technical report 📚 is uploaded to [arXiv](https://arxiv.org/abs/2404.13686)! Many implementation details are provided and we welcome more discussions👏.
* Apr.21, 2024. Hyper-SD ⚡️ is highly compatible and work well with different base models and controlnets. To clarify, we also append the usage example of controlnet [here](https://huggingface.co/ByteDance/Hyper-SD#controlnet-usage).
* Apr.20, 2024. Our checkpoints and two demos 🤗 (i.e. [SD15-Scribble](https://huggingface.co/spaces/ByteDance/Hyper-SD15-Scribble) and [SDXL-T2I](https://huggingface.co/spaces/ByteDance/Hyper-SDXL-1Step-T2I)) are publicly available on [HuggingFace Repo](https://huggingface.co/ByteDance/Hyper-SD).

## Try our Hugging Face demos: 
Hyper-SD Scribble demo host on [🤗 scribble](https://huggingface.co/spaces/ByteDance/Hyper-SD15-Scribble) 

Hyper-SDXL One-step Text-to-Image demo host on [🤗 T2I](https://huggingface.co/spaces/ByteDance/Hyper-SDXL-1Step-T2I)

## Introduction

Hyper-SD is one of the new State-of-the-Art diffusion model acceleration techniques.
In this repository, we release the models distilled from [SDXL Base 1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0) and [Stable-Diffusion v1-5](https://huggingface.co/runwayml/stable-diffusion-v1-5)。

## Checkpoints

* `Hyper-SDXL-Nstep-lora.safetensors`: Lora checkpoint, for SDXL-related models.
* `Hyper-SD15-Nstep-lora.safetensors`: Lora checkpoint, for SD1.5-related models.
* `Hyper-SDXL-1step-unet.safetensors`: Unet checkpoint distilled from SDXL-Base.

## Text-to-Image Usage
### SDXL-related models
#### 2-Steps, 4-Steps, 8-steps LoRA
Take the 2-steps LoRA as an example, you can also use other LoRAs for the corresponding inference steps setting.  
```python
import torch
from diffusers import DiffusionPipeline, DDIMScheduler
from huggingface_hub import hf_hub_download
base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
repo_name = "ByteDance/Hyper-SD"
# Take 2-steps lora as an example
ckpt_name = "Hyper-SDXL-2steps-lora.safetensors"
# Load model.
pipe = DiffusionPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
pipe.fuse_lora()
# Ensure ddim scheduler timestep spacing set as trailing !!!
pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
# lower eta results in more detail
prompt="a photo of a cat"
image=pipe(prompt=prompt, num_inference_steps=2, guidance_scale=0).images[0]
```

#### Unified LoRA (support 1 to 8 steps inference)
You can flexibly adjust the number of inference steps and eta value to achieve best performance. 
```python
import torch
from diffusers import DiffusionPipeline, TCDScheduler
from huggingface_hub import hf_hub_download
base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
repo_name = "ByteDance/Hyper-SD"
ckpt_name = "Hyper-SDXL-1step-lora.safetensors"
# Load model.
pipe = DiffusionPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
pipe.fuse_lora()
# Use TCD scheduler to achieve better image quality
pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
# Lower eta results in more detail for multi-steps inference
eta=1.0
prompt="a photo of a cat"
image=pipe(prompt=prompt, num_inference_steps=1, guidance_scale=0, eta=eta).images[0]
```

#### 1-step SDXL Unet
Only for the single step inference.
```python
import torch
from diffusers import DiffusionPipeline, UNet2DConditionModel, LCMScheduler
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file
base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
repo_name = "ByteDance/Hyper-SD"
ckpt_name = "Hyper-SDXL-1step-Unet.safetensors"
# Load model.
unet = UNet2DConditionModel.from_config(base_model_id, subfolder="unet").to("cuda", torch.float16)
unet.load_state_dict(load_file(hf_hub_download(repo_name, ckpt_name), device="cuda"))
pipe = DiffusionPipeline.from_pretrained(base_model_id, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")
# Use LCM scheduler instead of ddim scheduler to support specific timestep number inputs
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
# Set start timesteps to 800 in the one-step inference to get better results
prompt="a photo of a cat"
image=pipe(prompt=prompt, num_inference_steps=1, guidance_scale=0, timesteps=[800]).images[0]
```


### SD1.5-related models

#### 2-Steps, 4-Steps, 8-steps LoRA
Take the 2-steps LoRA as an example, you can also use other LoRAs for the corresponding inference steps setting.
```python
import torch
from diffusers import DiffusionPipeline, DDIMScheduler
from huggingface_hub import hf_hub_download
base_model_id = "runwayml/stable-diffusion-v1-5"
repo_name = "ByteDance/Hyper-SD"
# Take 2-steps lora as an example
ckpt_name = "Hyper-SD15-2steps-lora.safetensors"
# Load model.
pipe = DiffusionPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
pipe.fuse_lora()
# Ensure ddim scheduler timestep spacing set as trailing !!!
pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
prompt="a photo of a cat"
image=pipe(prompt=prompt, num_inference_steps=2, guidance_scale=0).images[0]
```


#### Unified LoRA (support 1 to 8 steps inference)
You can flexibly adjust the number of inference steps and eta value to achieve best performance.
```python
import torch
from diffusers import DiffusionPipeline, TCDScheduler
from huggingface_hub import hf_hub_download
base_model_id = "runwayml/stable-diffusion-v1-5"
repo_name = "ByteDance/Hyper-SD"
ckpt_name = "Hyper-SD15-1step-lora.safetensors"
# Load model.
pipe = DiffusionPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
pipe.fuse_lora()
# Use TCD scheduler to achieve better image quality
pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
# Lower eta results in more detail for multi-steps inference
eta=1.0
prompt="a photo of a cat"
image=pipe(prompt=prompt, num_inference_steps=1, guidance_scale=0, eta=eta).images[0]
```

## ControlNet Usage
### SDXL-related models

#### 2-Steps, 4-Steps, 8-steps LoRA
Take Canny Controlnet and 2-steps inference as an example:
```python
import torch
from diffusers.utils import load_image
import numpy as np
import cv2
from PIL import Image
from diffusers import ControlNetModel, StableDiffusionXLControlNetPipeline, AutoencoderKL, DDIMScheduler
from huggingface_hub import hf_hub_download

# Load original image
image = load_image("https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd_controlnet/hf-logo.png")
image = np.array(image)
# Prepare Canny Control Image
low_threshold = 100
high_threshold = 200
image = cv2.Canny(image, low_threshold, high_threshold)
image = image[:, :, None]
image = np.concatenate([image, image, image], axis=2)
control_image = Image.fromarray(image)
control_image.save("control.png")
control_weight = 0.5  # recommended for good generalization

# Initialize pipeline
controlnet = ControlNetModel.from_pretrained(
    "diffusers/controlnet-canny-sdxl-1.0",
    torch_dtype=torch.float16
)
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
pipe = StableDiffusionXLControlNetPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", controlnet=controlnet, vae=vae, torch_dtype=torch.float16).to("cuda")

pipe.load_lora_weights(hf_hub_download("ByteDance/Hyper-SD", "Hyper-SDXL-2steps-lora.safetensors"))
# Ensure ddim scheduler timestep spacing set as trailing !!!
pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
pipe.fuse_lora()
image = pipe("A chocolate cookie", num_inference_steps=2, image=control_image, guidance_scale=0, controlnet_conditioning_scale=control_weight).images[0]
image.save('image_out.png')
```

#### Unified LoRA (support 1 to 8 steps inference)
Take Canny Controlnet as an example:
```python
import torch
from diffusers.utils import load_image
import numpy as np
import cv2
from PIL import Image
from diffusers import ControlNetModel, StableDiffusionXLControlNetPipeline, AutoencoderKL, TCDScheduler
from huggingface_hub import hf_hub_download

# Load original image
image = load_image("https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd_controlnet/hf-logo.png")
image = np.array(image)
# Prepare Canny Control Image
low_threshold = 100
high_threshold = 200
image = cv2.Canny(image, low_threshold, high_threshold)
image = image[:, :, None]
image = np.concatenate([image, image, image], axis=2)
control_image = Image.fromarray(image)
control_image.save("control.png")
control_weight = 0.5  # recommended for good generalization

# Initialize pipeline
controlnet = ControlNetModel.from_pretrained(
    "diffusers/controlnet-canny-sdxl-1.0",
    torch_dtype=torch.float16
)
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    controlnet=controlnet, vae=vae, torch_dtype=torch.float16).to("cuda")

# Load Hyper-SD15-1step lora
pipe.load_lora_weights(hf_hub_download("ByteDance/Hyper-SD", "Hyper-SDXL-1step-lora.safetensors"))
pipe.fuse_lora()
# Use TCD scheduler to achieve better image quality
pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
# Lower eta results in more detail for multi-steps inference
eta=1.0
image = pipe("A chocolate cookie", num_inference_steps=4, image=control_image, guidance_scale=0, controlnet_conditioning_scale=control_weight, eta=eta).images[0]
image.save('image_out.png')
```

### SD1.5-related models

#### 2-Steps, 4-Steps, 8-steps LoRA
Take Canny Controlnet and 2-steps inference as an example:
```python
import torch
from diffusers.utils import load_image
import numpy as np
import cv2
from PIL import Image
from diffusers import ControlNetModel, StableDiffusionControlNetPipeline, DDIMScheduler

from huggingface_hub import hf_hub_download

controlnet_checkpoint = "lllyasviel/control_v11p_sd15_canny"

# Load original image
image = load_image("https://huggingface.co/lllyasviel/control_v11p_sd15_canny/resolve/main/images/input.png")
image = np.array(image)
# Prepare Canny Control Image
low_threshold = 100
high_threshold = 200
image = cv2.Canny(image, low_threshold, high_threshold)
image = image[:, :, None]
image = np.concatenate([image, image, image], axis=2)
control_image = Image.fromarray(image)
control_image.save("control.png")

# Initialize pipeline
controlnet = ControlNetModel.from_pretrained(controlnet_checkpoint, torch_dtype=torch.float16)
pipe = StableDiffusionControlNetPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", controlnet=controlnet, torch_dtype=torch.float16).to("cuda")
pipe.load_lora_weights(hf_hub_download("ByteDance/Hyper-SD", "Hyper-SD15-2steps-lora.safetensors"))
pipe.fuse_lora()
# Ensure ddim scheduler timestep spacing set as trailing !!!
pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
image = pipe("a blue paradise bird in the jungle", num_inference_steps=2, image=control_image, guidance_scale=0).images[0]
image.save('image_out.png')
```


#### Unified LoRA (support 1 to 8 steps inference)
Take Canny Controlnet as an example:
```python
import torch
from diffusers.utils import load_image
import numpy as np
import cv2
from PIL import Image
from diffusers import ControlNetModel, StableDiffusionControlNetPipeline, TCDScheduler
from huggingface_hub import hf_hub_download

controlnet_checkpoint = "lllyasviel/control_v11p_sd15_canny"

# Load original image
image = load_image("https://huggingface.co/lllyasviel/control_v11p_sd15_canny/resolve/main/images/input.png")
image = np.array(image)
# Prepare Canny Control Image
low_threshold = 100
high_threshold = 200
image = cv2.Canny(image, low_threshold, high_threshold)
image = image[:, :, None]
image = np.concatenate([image, image, image], axis=2)
control_image = Image.fromarray(image)
control_image.save("control.png")

# Initialize pipeline
controlnet = ControlNetModel.from_pretrained(controlnet_checkpoint, torch_dtype=torch.float16)
pipe = StableDiffusionControlNetPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", controlnet=controlnet, torch_dtype=torch.float16).to("cuda")
# Load Hyper-SD15-1step lora
pipe.load_lora_weights(hf_hub_download("ByteDance/Hyper-SD", "Hyper-SD15-1step-lora.safetensors"))
pipe.fuse_lora()
# Use TCD scheduler to achieve better image quality
pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
# Lower eta results in more detail for multi-steps inference
eta=1.0
image = pipe("a blue paradise bird in the jungle", num_inference_steps=1, image=control_image, guidance_scale=0, eta=eta).images[0]
image.save('image_out.png')
```
## Comfyui Usage
* `Hyper-SDXL-Nsteps-lora.safetensors`: [text-to-image workflow](https://huggingface.co/ByteDance/Hyper-SD/blob/main/comfyui/Hyper-SDXL-Nsteps-lora-workflow.json)
* `Hyper-SD15-Nsteps-lora.safetensors`: [text-to-image workflow](https://huggingface.co/ByteDance/Hyper-SD/blob/main/comfyui/Hyper-SD15-Nsteps-lora-workflow.json)
* `Hyper-SDXL-1step-Unet-Comfyui.fp16.safetensors`: [text-to-image workflow](https://huggingface.co/ByteDance/Hyper-SD/blob/main/comfyui/Hyper-SDXL-1step-Unet-workflow.json)
  * **REQUIREMENT / INSTALL** for 1-Step SDXL UNet: Please install our [scheduler folder](https://huggingface.co/ByteDance/Hyper-SD/tree/main/comfyui/ComfyUI-HyperSDXL1StepUnetScheduler) into your `ComfyUI/custom_nodes` to enable sampling from 800 timestep instead of 999. 
  * i.e. making sure the `ComfyUI/custom_nodes/ComfyUI-HyperSDXL1StepUnetScheduler` folder exist.
  * For more details, please refer to our [technical report](https://arxiv.org/abs/2404.13686).

## Citation
```bibtex
@misc{ren2024hypersd,
      title={Hyper-SD: Trajectory Segmented Consistency Model for Efficient Image Synthesis}, 
      author={Yuxi Ren and Xin Xia and Yanzuo Lu and Jiacheng Zhang and Jie Wu and Pan Xie and Xing Wang and Xuefeng Xiao},
      year={2024},
      eprint={2404.13686},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
```