Update README.md
Browse files
README.md
CHANGED
|
@@ -19,19 +19,60 @@ widget:
|
|
| 19 |
url: images/4444.png
|
| 20 |
base_model: black-forest-labs/FLUX.1-dev
|
| 21 |
instance_prompt: Belle, the female character
|
| 22 |
-
|
| 23 |
---
|
|
|
|
|
|
|
|
|
|
| 24 |
# Belle
|
| 25 |
|
| 26 |
<Gallery />
|
|
|
|
|
|
|
| 27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
|
| 31 |
-
|
| 32 |
|
| 33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
|
|
|
| 35 |
|
| 36 |
## Download model
|
| 37 |
|
|
|
|
| 19 |
url: images/4444.png
|
| 20 |
base_model: black-forest-labs/FLUX.1-dev
|
| 21 |
instance_prompt: Belle, the female character
|
|
|
|
| 22 |
---
|
| 23 |
+
|
| 24 |
+

|
| 25 |
+
|
| 26 |
# Belle
|
| 27 |
|
| 28 |
<Gallery />
|
| 29 |
+
# Model description for Belle
|
| 30 |
+
Image Processing Parameters
|
| 31 |
|
| 32 |
+
| Parameter | Value | Parameter | Value |
|
| 33 |
+
|---------------------------|--------|---------------------------|--------|
|
| 34 |
+
| LR Scheduler | constant | Noise Offset | 0.03 |
|
| 35 |
+
| Optimizer | AdamW | Multires Noise Discount | 0.1 |
|
| 36 |
+
| Network Dim | 64 | Multires Noise Iterations | 10 |
|
| 37 |
+
| Network Alpha | 32 | Repeat & Steps | 18 & 2100 |
|
| 38 |
+
| Epoch | 21 | Save Every N Epochs | 1 |
|
| 39 |
|
| 40 |
+
Labeling: florence2-en(natural language & English)
|
| 41 |
+
|
| 42 |
+
Total Images Used for Training : 14
|
| 43 |
+
|
| 44 |
+
Source : Pinterest
|
| 45 |
+
|
| 46 |
+
## Best Dimensions & Inference
|
| 47 |
+
|
| 48 |
+
| **Dimensions** | **Aspect Ratio** | **Recommendation** |
|
| 49 |
+
|-----------------|------------------|---------------------------|
|
| 50 |
+
| 1280 x 832 | 3:2 | Best |
|
| 51 |
+
| 1024 x 1024 | 1:1 | Default |
|
| 52 |
|
| 53 |
+
### Inference Range
|
| 54 |
|
| 55 |
+
- **Recommended Inference Steps:** 30–35
|
| 56 |
+
|
| 57 |
+
## Setting Up
|
| 58 |
+
```python
|
| 59 |
+
import torch
|
| 60 |
+
from pipelines import DiffusionPipeline
|
| 61 |
+
|
| 62 |
+
base_model = "black-forest-labs/FLUX.1-dev"
|
| 63 |
+
pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.bfloat16)
|
| 64 |
+
|
| 65 |
+
lora_repo = "strangerzonehf/Belle"
|
| 66 |
+
trigger_word = "Belle the female character"
|
| 67 |
+
pipe.load_lora_weights(lora_repo)
|
| 68 |
+
|
| 69 |
+
device = torch.device("cuda")
|
| 70 |
+
pipe.to(device)
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
## Trigger words
|
| 74 |
|
| 75 |
+
You should use `Belle the female character` to trigger the image generation.
|
| 76 |
|
| 77 |
## Download model
|
| 78 |
|