renyuxi Yanzuo commited on
Commit
5024c1d
β€’
1 Parent(s): 64b477d

Update README.md (#52)

Browse files

- Update README.md (f959d506c7a1b80d5c0750abeb703b98e57026bf)


Co-authored-by: Yanzuo Lu <[email protected]>

Files changed (1) hide show
  1. README.md +22 -3
README.md CHANGED
@@ -18,7 +18,8 @@ Project Page: https://hyper-sd.github.io/
18
 
19
  ## NewsπŸ”₯πŸ”₯πŸ”₯
20
 
21
- * Aug.19, 2024. πŸ’₯πŸ’₯πŸ’₯ **SD3-related CFG LoRAs** are available now! We recommend setting guidance scale to 3.0/5.0/7.0 at 4/8/16-steps. Don't forget to fuse lora with a relatively small scale (e.g. 0.125 that is adaptive with training) before inference with diffusers. Note that 8-steps and 16-steps LoRA can also inference on a little bit smaller steps like 6-steps and 12-steps, respectively. Hope to hear your feedback, FLUX-related models will be coming next week. πŸ’₯πŸ’₯πŸ’₯
 
22
  * May.13, 2024. The 12-Steps CFG-Preserved [Hyper-SDXL-12steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SDXL-12steps-CFG-lora.safetensors) and [Hyper-SD15-12steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SD15-12steps-CFG-lora.safetensors) is also available now(support 5~8 guidance scales), this could be more practical with better trade-off between performance and speed. Enjoy!
23
  * Apr.30, 2024. Our 8-Steps CFG-Preserved [Hyper-SDXL-8steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SDXL-8steps-CFG-lora.safetensors) and [Hyper-SD15-8steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SD15-8steps-CFG-lora.safetensors) is available now(support 5~8 guidance scales), we strongly recommend making the 8-step CFGLora a standard configuration for all SDXL and SD15 models!!!
24
  * Apr.28, 2024. ComfyUI workflows on 1-Step Unified LoRA πŸ₯° with TCDScheduler to inference on different steps are [released](https://huggingface.co/ByteDance/Hyper-SD/tree/main/comfyui)! Remember to install ⭕️ [ComfyUI-TCD](https://github.com/JettHu/ComfyUI-TCD) in your `ComfyUI/custom_nodes` folder!!! You're encouraged to adjust the eta parameter to get better results 🌟!
@@ -37,10 +38,11 @@ Hyper-SDXL One-step Text-to-Image demo host on [πŸ€— T2I](https://huggingface.co
37
  ## Introduction
38
 
39
  Hyper-SD is one of the new State-of-the-Art diffusion model acceleration techniques.
40
- In this repository, we release the models distilled from [SD3-Medium](https://huggingface.co/stabilityai/stable-diffusion-3-medium-diffusers), [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)。
41
 
42
  ## Checkpoints
43
 
 
44
  * `Hyper-SD3-Nsteps-CFG-lora.safetensors`: Lora checkpoint, for SD3-related models.
45
  * `Hyper-SDXL-Nstep-lora.safetensors`: Lora checkpoint, for SDXL-related models.
46
  * `Hyper-SD15-Nstep-lora.safetensors`: Lora checkpoint, for SD1.5-related models.
@@ -48,6 +50,24 @@ In this repository, we release the models distilled from [SD3-Medium](https://hu
48
 
49
  ## Text-to-Image Usage
50
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
51
  ### SD3-related models
52
  ```python
53
  import torch
@@ -64,7 +84,6 @@ pipe.fuse_lora(lora_scale=0.125)
64
  pipe.to("cuda", dtype=torch.float16)
65
  image=pipe(prompt="a photo of a cat", num_inference_steps=8, guidance_scale=5.0).images[0]
66
  image.save("output.png")
67
-
68
  ```
69
 
70
  ### SDXL-related models
 
18
 
19
  ## NewsπŸ”₯πŸ”₯πŸ”₯
20
 
21
+ * Aug.26, 2024. πŸ’₯πŸ’₯πŸ’₯ Our 8-steps and 16-steps **FLUX.1-dev-related LoRAs** are available now! We recommend LoRA scales around 0.125 that is adaptive with training and guidance scale could be kept on 3.5. Lower step LoRAs would be coming soon. πŸ’₯πŸ’₯πŸ’₯
22
+ * Aug.19, 2024. SD3-related CFG LoRAs are available now! We recommend setting guidance scale to 3.0/5.0/7.0 at 4/8/16-steps. Don't forget to fuse lora with a relatively small scale (e.g. 0.125 that is adaptive with training) before inference with diffusers. Note that 8-steps and 16-steps LoRA can also inference on a little bit smaller steps like 6-steps and 12-steps, respectively. Hope to hear your feedback, FLUX-related models will be coming next week.
23
  * May.13, 2024. The 12-Steps CFG-Preserved [Hyper-SDXL-12steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SDXL-12steps-CFG-lora.safetensors) and [Hyper-SD15-12steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SD15-12steps-CFG-lora.safetensors) is also available now(support 5~8 guidance scales), this could be more practical with better trade-off between performance and speed. Enjoy!
24
  * Apr.30, 2024. Our 8-Steps CFG-Preserved [Hyper-SDXL-8steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SDXL-8steps-CFG-lora.safetensors) and [Hyper-SD15-8steps-CFG-LoRA](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-SD15-8steps-CFG-lora.safetensors) is available now(support 5~8 guidance scales), we strongly recommend making the 8-step CFGLora a standard configuration for all SDXL and SD15 models!!!
25
  * Apr.28, 2024. ComfyUI workflows on 1-Step Unified LoRA πŸ₯° with TCDScheduler to inference on different steps are [released](https://huggingface.co/ByteDance/Hyper-SD/tree/main/comfyui)! Remember to install ⭕️ [ComfyUI-TCD](https://github.com/JettHu/ComfyUI-TCD) in your `ComfyUI/custom_nodes` folder!!! You're encouraged to adjust the eta parameter to get better results 🌟!
 
38
  ## Introduction
39
 
40
  Hyper-SD is one of the new State-of-the-Art diffusion model acceleration techniques.
41
+ In this repository, we release the models distilled from [FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev), [SD3-Medium](https://huggingface.co/stabilityai/stable-diffusion-3-medium-diffusers), [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)。
42
 
43
  ## Checkpoints
44
 
45
+ * `Hyper-FLUX.1-dev-Nsteps-lora.safetensors`: Lora checkpoint, for FLUX.1-dev-related models.
46
  * `Hyper-SD3-Nsteps-CFG-lora.safetensors`: Lora checkpoint, for SD3-related models.
47
  * `Hyper-SDXL-Nstep-lora.safetensors`: Lora checkpoint, for SDXL-related models.
48
  * `Hyper-SD15-Nstep-lora.safetensors`: Lora checkpoint, for SD1.5-related models.
 
50
 
51
  ## Text-to-Image Usage
52
 
53
+ ### FLUX.1-dev-related models
54
+ ```python
55
+ import torch
56
+ from diffusers import FluxPipeline
57
+ from huggingface_hub import hf_hub_download
58
+ base_model_id = "black-forest-labs/FLUX.1-dev"
59
+ repo_name = "ByteDance/Hyper-SD"
60
+ # Take 8-steps lora as an example
61
+ ckpt_name = "Hyper-FLUX.1-dev-8steps-lora.safetensors"
62
+ # Load model, please fill in your access tokens since FLUX.1-dev repo is a gated model.
63
+ pipe = FluxPipeline.from_pretrained(base_model_id, token="xxx")
64
+ pipe.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
65
+ pipe.fuse_lora(lora_scale=0.125)
66
+ pipe.to("cuda", dtype=torch.float16)
67
+ image=pipe(prompt="a photo of a cat", num_inference_steps=8, guidance_scale=3.5).images[0]
68
+ image.save("output.png")
69
+ ```
70
+
71
  ### SD3-related models
72
  ```python
73
  import torch
 
84
  pipe.to("cuda", dtype=torch.float16)
85
  image=pipe(prompt="a photo of a cat", num_inference_steps=8, guidance_scale=5.0).images[0]
86
  image.save("output.png")
 
87
  ```
88
 
89
  ### SDXL-related models