Update README.md
Browse files
README.md
CHANGED
|
@@ -3,6 +3,10 @@ license: apache-2.0
|
|
| 3 |
---
|
| 4 |
|
| 5 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
|
| 7 |
# 🥯 BAGEL • Unified Model for Multimodal Understanding and Generation
|
| 8 |
|
|
@@ -38,7 +42,6 @@ license: apache-2.0
|
|
| 38 |
|
| 39 |
> We present **BAGEL**, an open‑source multimodal foundation model with 7B active parameters (14B total) trained on large‑scale interleaved multimodal data. BAGEL outperforms the current top‑tier open‑source VLMs like Qwen2.5-VL and InternVL-2.5 on standard multimodal understanding leaderboards, and delivers text‑to‑image quality that is competitive with strong specialist generators such as SD3.
|
| 40 |
Moreover, BAGEL demonstrates superior qualitative results in classical image‑editing scenarios than the leading open-source models. More importantly, it extends to free-form visual manipulation, multiview synthesis, and world navigation, capabilities that constitute "world-modeling" tasks beyond the scope of previous image-editing models.
|
| 41 |
-
Below is a showcase of BAGEL's qualitative performance.
|
| 42 |
|
| 43 |
|
| 44 |
This repository hosts the model weights for **BAGEL**. For installation, usage instructions, and further documentation, please visit our [GitHub repository](https://github.com/bytedance-seed/BAGEL).
|
|
@@ -60,12 +63,12 @@ This repository hosts the model weights for **BAGEL**. For installation, usage i
|
|
| 60 |
| Janus-Pro-7B | 0.80 |
|
| 61 |
| **BAGEL** | **0.88** |
|
| 62 |
### 3. Image Editing
|
| 63 |
-
|
|
| 64 |
-
| ------------------------
|
| 65 |
-
|
|
| 66 |
-
|
|
| 67 |
-
| **
|
| 68 |
-
| **
|
| 69 |
|
| 70 |
## License
|
| 71 |
BAGEL is licensed under the Apache 2.0 license. It is finetuned from [Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) and [siglip-so400m-14-980-flash-attn2-navit](https://huggingface.co/HuggingFaceM4/siglip-so400m-14-980-flash-attn2-navit) model, and uses the [FLUX.1-schnell VAE model](https://huggingface.co/black-forest-labs/FLUX.1-schnell), all under Apache 2.0.
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
|
| 6 |
+
<p align="left">
|
| 7 |
+
<img src="https://lf3-static.bytednsdoc.com/obj/eden-cn/nuhojubrps/banner.png" alt="BAGEL" width="480"/>
|
| 8 |
+
</p>
|
| 9 |
+
|
| 10 |
|
| 11 |
# 🥯 BAGEL • Unified Model for Multimodal Understanding and Generation
|
| 12 |
|
|
|
|
| 42 |
|
| 43 |
> We present **BAGEL**, an open‑source multimodal foundation model with 7B active parameters (14B total) trained on large‑scale interleaved multimodal data. BAGEL outperforms the current top‑tier open‑source VLMs like Qwen2.5-VL and InternVL-2.5 on standard multimodal understanding leaderboards, and delivers text‑to‑image quality that is competitive with strong specialist generators such as SD3.
|
| 44 |
Moreover, BAGEL demonstrates superior qualitative results in classical image‑editing scenarios than the leading open-source models. More importantly, it extends to free-form visual manipulation, multiview synthesis, and world navigation, capabilities that constitute "world-modeling" tasks beyond the scope of previous image-editing models.
|
|
|
|
| 45 |
|
| 46 |
|
| 47 |
This repository hosts the model weights for **BAGEL**. For installation, usage instructions, and further documentation, please visit our [GitHub repository](https://github.com/bytedance-seed/BAGEL).
|
|
|
|
| 63 |
| Janus-Pro-7B | 0.80 |
|
| 64 |
| **BAGEL** | **0.88** |
|
| 65 |
### 3. Image Editing
|
| 66 |
+
| Model | GEdit-Bench-EN (SC) ↑ | GEdit-Bench-EN (PQ) ↑ | GEdit-Bench-EN (O) ↑ | IntelligentBench ↑ |
|
| 67 |
+
| ------------- | --------------------- | --------------------- | ------------------- | ------------------ |
|
| 68 |
+
| Step1X-Edit | 7.09 | 6.76 | **6.70** | 14.9 |
|
| 69 |
+
| Gemini-2-exp. | 6.73 | 6.61 | 6.32 | **57.6** |
|
| 70 |
+
| **BAGEL** | **7.36** | **6.83** | 6.52 | 44.0 |
|
| 71 |
+
| **BAGEL+CoT** | – | – | – | 55.3 |
|
| 72 |
|
| 73 |
## License
|
| 74 |
BAGEL is licensed under the Apache 2.0 license. It is finetuned from [Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) and [siglip-so400m-14-980-flash-attn2-navit](https://huggingface.co/HuggingFaceM4/siglip-so400m-14-980-flash-attn2-navit) model, and uses the [FLUX.1-schnell VAE model](https://huggingface.co/black-forest-labs/FLUX.1-schnell), all under Apache 2.0.
|