Update README.md
Browse files
README.md
CHANGED
|
@@ -5,10 +5,6 @@ tags:
|
|
| 5 |
- lora
|
| 6 |
- diffusers
|
| 7 |
- template:sd-lora
|
| 8 |
-
widget:
|
| 9 |
-
- text: '-'
|
| 10 |
-
output:
|
| 11 |
-
url: images/result_juggernaut_tscd_8step.png
|
| 12 |
base_model: digiplay/Juggernaut_final
|
| 13 |
instance_prompt: null
|
| 14 |
license: mit
|
|
@@ -27,7 +23,61 @@ Efficient Image Synthesis](https://arxiv.org/abs&
|
|
| 27 |
This LoRA is fine-tuned on [laion/conceptual-captions-12m-webdataset](https://huggingface.co/datasets/laion/conceptual-captions-12m-webdataset) for 16.000 iterations using unofficial training implementation at [https://github.com/rootonchair/consistency_models_distill](https://github.com/rootonchair/consistency_models_distill)
|
| 28 |
|
| 29 |
# Usage
|
|
|
|
| 30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
## Download model
|
| 33 |
|
|
|
|
| 5 |
- lora
|
| 6 |
- diffusers
|
| 7 |
- template:sd-lora
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
base_model: digiplay/Juggernaut_final
|
| 9 |
instance_prompt: null
|
| 10 |
license: mit
|
|
|
|
| 23 |
This LoRA is fine-tuned on [laion/conceptual-captions-12m-webdataset](https://huggingface.co/datasets/laion/conceptual-captions-12m-webdataset) for 16.000 iterations using unofficial training implementation at [https://github.com/rootonchair/consistency_models_distill](https://github.com/rootonchair/consistency_models_distill)
|
| 24 |
|
| 25 |
# Usage
|
| 26 |
+
```python
|
| 27 |
|
| 28 |
+
import os
|
| 29 |
+
import torch
|
| 30 |
+
from diffusers import StableDiffusionPipeline, TCDScheduler
|
| 31 |
+
|
| 32 |
+
pipeline = StableDiffusionPipeline.from_pretrained("digiplay/Juggernaut_final", torch_dtype=torch.float16, safety_checker = None).to("cuda")
|
| 33 |
+
pipeline.load_lora_weights("rootonchair/tscd_juggernaut_final")
|
| 34 |
+
pipeline.scheduler = TCDScheduler.from_config(pipeline.scheduler.config)
|
| 35 |
+
test_prompts = [
|
| 36 |
+
"cgmech, white mecha robot, cape, science fiction, torn clothes, glowing, standing, robot joints, mecha, armor, cowboy shot, intense sunlight, silver dragonborn, outdoors, landscape, nature, volumetrics dtx",
|
| 37 |
+
"Portrait photo of muscular bearded guy in a worn mech suit, elegant, sharp focus, photo by greg rutkowski, soft lighting, vibrant colors",
|
| 38 |
+
"photo of a supercar, 8k uhd, high quality, road, sunset, motion blur, depth blur, cinematic, filmic image 4k",
|
| 39 |
+
"a portrait of a white cat wearing glasses, highly detailed",
|
| 40 |
+
"a beautiful runrise at the beach, cinematic, masterpiece",
|
| 41 |
+
"A photo of beautiful mountain with realistic sunset and blue lake, highly detailed, masterpiece",
|
| 42 |
+
"Self-portrait oil painting, a beautiful cyborg with golden hair, 8k",
|
| 43 |
+
"Astronaut in a jungle, cold color palette, muted colors, detailed, 8k";
|
| 44 |
+
]
|
| 45 |
+
seed = 1234
|
| 46 |
+
images = pipeline(test_prompts, width=512, height=512, num_inference_steps=4, guidance_scale=0, generator=torch.Generator(device="cuda").manual_seed(seed)).images
|
| 47 |
+
```
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
## Generation Result
|
| 51 |
+
|
| 52 |
+
### Original model
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
### 2-steps inference
|
| 58 |
+
#### TSCD
|
| 59 |
+
|
| 60 |
+
#### LCM-Lora
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
### 4-steps inference
|
| 65 |
+
#### TSCD
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
#### LCM-Lora
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
|
| 73 |
+
### 8-steps inference
|
| 74 |
+
#### TSCD
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
|
| 78 |
+
#### LCM-Lora
|
| 79 |
+
|
| 80 |
+
|
| 81 |
|
| 82 |
## Download model
|
| 83 |
|