Instructions to use sanaka87/BAGEL-RecA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use sanaka87/BAGEL-RecA with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("sanaka87/BAGEL-RecA", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -8,4 +8,48 @@ language:
|
|
| 8 |
base_model:
|
| 9 |
- ByteDance-Seed/BAGEL-7B-MoT
|
| 10 |
pipeline_tag: any-to-any
|
| 11 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
base_model:
|
| 9 |
- ByteDance-Seed/BAGEL-7B-MoT
|
| 10 |
pipeline_tag: any-to-any
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# 🔥 BAGEL-ReAlign • A Better BAGEL with better image generation and editing performance.
|
| 14 |
+
|
| 15 |
+
This repository hosts the model weights for **BAGEL-ReAlign**. We fine-tuned BAGEL on 6 80GB NVIDIA A800 for only 27 GPU hours. While the understanding capability remains unchanged, our ReAlign method brings +3.6 **zero-shot improvement** on GenEval , +1.26 on DPGBench, +0.37 on ImgEdit and +0.33 on GEdit.
|
| 16 |
+
|
| 17 |
+
For installation, usage instructions, and further documentation, please visit BAGEL's original repository [GitHub repository](https://github.com/bytedance-seed/BAGEL).
|
| 18 |
+
|
| 19 |
+
## 🧠 Method
|
| 20 |
+
|
| 21 |
+
We present **ReAlign**, a self-supervised training framework that aligns understanding and generation in modest compute.
|
| 22 |
+
|
| 23 |
+
Coming soon! Stay tuned~
|
| 24 |
+
|
| 25 |
+
## 📊 Benchmarks
|
| 26 |
+
|
| 27 |
+
### 1. Visual Understanding
|
| 28 |
+
|
| 29 |
+
Remains Unchanged.
|
| 30 |
+
|
| 31 |
+
### 2. Text-to-Image Generation · GenEval
|
| 32 |
+
|
| 33 |
+
We test it on 1024x1024 resolution.
|
| 34 |
+
|
| 35 |
+
| Model | GenEval ↑ | DPGBench ↑ |
|
| 36 |
+
| ------------ | --------- | --------- |
|
| 37 |
+
| **BAGEL** | 0.787 | 84.03 |
|
| 38 |
+
| **BAGEL-ReAligned** | **0.824** | **85.29** |
|
| 39 |
+
|
| 40 |
+
### 3. Image Editing
|
| 41 |
+
|
| 42 |
+
| Model | GEdit-Bench-EN (SC) ↑ | GEdit-Bench-EN (PQ) ↑ | GEdit-Bench-EN (O) ↑ | ImgEdit ↑ |
|
| 43 |
+
| ------------- | --------------------- | --------------------- | ------------------- | ------------------ |
|
| 44 |
+
| **BAGEL** | 7.96 | 6.64 | 6.94 | 3.38 |
|
| 45 |
+
| **BAGEL-NHR** | 8.04 | 6.87 | 7.08 | 3.48 |
|
| 46 |
+
| **BAGEL-ReAlign** | **8.24** | 6.87 | **7.27** | **3.75** |
|
| 47 |
+
| **FLUX Kontext** | 6.95 | **7.30** | 6.27 | 3.59 |
|
| 48 |
+
|
| 49 |
+
## License
|
| 50 |
+
|
| 51 |
+
BAGEL-ReAlign is licensed under the Apache 2.0 license.
|
| 52 |
+
|
| 53 |
+
## ✍️ Citation
|
| 54 |
+
|
| 55 |
+
Coming soon!
|