Update README.md
Browse files
README.md
CHANGED
|
@@ -19,17 +19,6 @@ This repository contains a FLUX.1-dev text-to-image diffusion model stored in Or
|
|
| 19 |
|
| 20 |
FLUX.1-dev is a powerful text-to-image generation model that uses a transformer-based architecture with dual text encoders (CLIP and T5) for enhanced text understanding and image generation capabilities.
|
| 21 |
|
| 22 |
-
### Model Specifications
|
| 23 |
-
|
| 24 |
-
- **Architecture**: Transformer-based diffusion model
|
| 25 |
-
- **Hidden Dimension**: 3072
|
| 26 |
-
- **Attention Heads**: 24
|
| 27 |
-
- **Double Layers**: 19
|
| 28 |
-
- **Single Layers**: 38
|
| 29 |
-
- **Precision**: BFloat16
|
| 30 |
-
- **Joint Attention Dimension**: 4096
|
| 31 |
-
- **Pooled Projection Dimension**: 768
|
| 32 |
-
- **In Channels**: 64
|
| 33 |
|
| 34 |
### Components
|
| 35 |
|
|
@@ -45,6 +34,21 @@ This model includes the following components:
|
|
| 45 |
|
| 46 |
To use this model, you should use the **Diffuse** library, which provides an easy-to-use interface for FLUX models.
|
| 47 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
### Installation
|
| 49 |
|
| 50 |
Install the Diffuse library:
|
|
@@ -146,27 +150,18 @@ img = Image.fromarray((image * 255).astype('uint8'))
|
|
| 146 |
img.save("output.png")
|
| 147 |
```
|
| 148 |
|
| 149 |
-
### Tutorial
|
| 150 |
-
|
| 151 |
-
For a comprehensive tutorial on using FLUX models with Diffuse, please refer to:
|
| 152 |
-
**[FLUX Tutorial Documentation](https://diffuse.readthedocs.io/en/latest/flux_tutorial.html)**
|
| 153 |
-
|
| 154 |
-
## Resources
|
| 155 |
-
|
| 156 |
-
- **Diffuse Library**: [https://github.com/jcopo/diffuse](https://github.com/jcopo/diffuse)
|
| 157 |
-
- **Documentation**: [https://diffuse.readthedocs.io/](https://diffuse.readthedocs.io/)
|
| 158 |
-
- **FLUX Tutorial**: [https://diffuse.readthedocs.io/en/latest/flux_tutorial.html](https://diffuse.readthedocs.io/en/latest/flux_tutorial.html)
|
| 159 |
-
|
| 160 |
-
## Model Format
|
| 161 |
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
## Requirements
|
| 165 |
|
| 166 |
-
-
|
| 167 |
-
-
|
| 168 |
-
-
|
| 169 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 170 |
|
| 171 |
## License
|
| 172 |
|
|
|
|
| 19 |
|
| 20 |
FLUX.1-dev is a powerful text-to-image generation model that uses a transformer-based architecture with dual text encoders (CLIP and T5) for enhanced text understanding and image generation capabilities.
|
| 21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
### Components
|
| 24 |
|
|
|
|
| 34 |
|
| 35 |
To use this model, you should use the **Diffuse** library, which provides an easy-to-use interface for FLUX models.
|
| 36 |
|
| 37 |
+
### Tutorial
|
| 38 |
+
|
| 39 |
+
For a comprehensive tutorial on using FLUX models with Diffuse, please refer to:
|
| 40 |
+
**[FLUX Tutorial Documentation](https://diffuse.readthedocs.io/en/latest/flux_tutorial.html)**
|
| 41 |
+
|
| 42 |
+
## Resources
|
| 43 |
+
|
| 44 |
+
- **Diffuse Library**: [https://github.com/jcopo/diffuse](https://github.com/jcopo/diffuse)
|
| 45 |
+
- **Documentation**: [https://diffuse.readthedocs.io/](https://diffuse.readthedocs.io/)
|
| 46 |
+
- **FLUX Tutorial**: [https://diffuse.readthedocs.io/en/latest/flux_tutorial.html](https://diffuse.readthedocs.io/en/latest/flux_tutorial.html)
|
| 47 |
+
|
| 48 |
+
## Model Format
|
| 49 |
+
|
| 50 |
+
This model is stored in **Orbax checkpoint format**, optimized for JAX/Flax frameworks. The Diffuse library handles loading and inference automatically.
|
| 51 |
+
|
| 52 |
### Installation
|
| 53 |
|
| 54 |
Install the Diffuse library:
|
|
|
|
| 150 |
img.save("output.png")
|
| 151 |
```
|
| 152 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 153 |
|
| 154 |
+
### Model Specifications
|
|
|
|
|
|
|
| 155 |
|
| 156 |
+
- **Architecture**: Transformer-based diffusion model
|
| 157 |
+
- **Hidden Dimension**: 3072
|
| 158 |
+
- **Attention Heads**: 24
|
| 159 |
+
- **Double Layers**: 19
|
| 160 |
+
- **Single Layers**: 38
|
| 161 |
+
- **Precision**: BFloat16
|
| 162 |
+
- **Joint Attention Dimension**: 4096
|
| 163 |
+
- **Pooled Projection Dimension**: 768
|
| 164 |
+
- **In Channels**: 64
|
| 165 |
|
| 166 |
## License
|
| 167 |
|