Update README.md

README.md

@@ -8,6 +8,7 @@ tags:
 - multimodal-retrieval
 - embedding-model
 ---
+
 <h1 align="center">MegaPairs: Massive Data Synthesis For Universal Multimodal Retrieval</h1>
 
 <p align="center">
@@ -17,56 +18,56 @@ tags:
     <a href="https://github.com/VectorSpaceLab/MegaPairs">
         <img alt="Build" src="https://img.shields.io/badge/Github-Code-blue">
     </a>
-    <a href="https://huggingface.co/datasets/JUNJIE99/MegaPairs">
+    <a href="https://huggingface.co/datasets/BAAI/MegaPairs">
         <img alt="Build" src="https://img.shields.io/badge/🤗 Datasets-MegaPairs-yellow">
 </p>
 
 <p align="center">
 </a>
-    <a href="https://huggingface.co/JUNJIE99/MMRet-base">
-        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-MMRet_base-yellow">
+    <a href="https://huggingface.co/BAAI/BGE-VL-base">
+        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-BGE_VL_base-yellow">
     </a>
-    <a href="https://huggingface.co/JUNJIE99/MMRet-large">
-        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-MMRet_large-yellow">
+    <a href="https://huggingface.co/BAAI/BGE-VL-large">
+        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-BGE_VL_large-yellow">
     </a>
-    <a href="https://huggingface.co/JUNJIE99/MMRet-MLLM-S1">
-        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-MMRet_MLLM_S1-yellow">
+    <a href="https://huggingface.co/BAAI/BGE-VL-MLLM-S1">
+        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-BGE_VL_MLLM_S1-yellow">
     </a>
-    <a href="https://huggingface.co/JUNJIE99/MMRet-MLLM-S2">
-        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-MMRet_MLLM_S2-yellow">
+    <a href="https://huggingface.co/BAAI/BGE-VL-MLLM-S2">
+        <img alt="Build" src="https://img.shields.io/badge/🤗 Model-BGE_VL_MLLM_S2-yellow">
     </a>
 </p>
 
 ## News
-```2024-3-4``` 🚀🚀 We have released the MMRet-MLLM models on Hugging Face: [MMRet-MLLM-S1](https://huggingface.co/JUNJIE99/MMRet-MLLM-S1) and [MMRet-MLLM-S2](https://huggingface.co/JUNJIE99/MMRet-MLLM-S2). **MMRet-MLLM-S1** is trained exclusively on our MegaPairs dataset, achieving outstanding performance in composed image retrieval, with an 8.1% improvement on the CIRCO benchmark (mAP@5) over the previous state-of-the-art. **MMRet-MLLM-S2** builds on MMRet-MLLM-S1 with an additional epoch of fine-tuning on the MMEB benchmark training set, delivering enhanced performance across a broader range of multimodal embedding tasks.
+```2024-3-4``` 🚀🚀 We have released the BGE-VL-MLLM models on Huggingface: [BGE-VL-MLLM-S1](https://huggingface.co/BAAI/BGE-VL-MLLM-S1) and [BGE-VL-MLLM-S2](https://huggingface.co/BAAI/BGE-VL-MLLM-S2). **BGE-VL-MLLM-S1** is trained exclusively on our MegaPairs dataset, achieving outstanding performance in composed image retrieval, with an 8.1% improvement on the CIRCO benchmark (mAP@5) over the previous state-of-the-art. **BGE-VL-MLLM-S2** builds on BGE-VL-MLLM-S1 with an additional epoch of fine-tuning on the MMEB benchmark training set, delivering enhanced performance across a broader range of multimodal embedding tasks.
 
-```2024-12-27``` 🚀🚀 MMRet-CLIP models are released in Huggingface: [MMRet-base](https://huggingface.co/JUNJIE99/MMRet-base) and [MMRet-large](https://huggingface.co/JUNJIE99/MMRet-large).
+```2024-12-27``` 🚀🚀 BGE-VL-CLIP models are released on Huggingface: [BGE-VL-base](https://huggingface.co/BAAI/BGE-VL-base) and [BGE-VL-large](https://huggingface.co/BAAI/BGE-VL-large).
 
 ```2024-12-19``` 🎉🎉 Release our paper: [MegaPairs: Massive Data Synthesis For Universal Multimodal Retrieval](https://arxiv.org/pdf/2412.14475).
 
 ## Release Plan
 - [x] Paper
-- [x] MMRet-base and MMRet-large models
-- [x] MMRet-MLLM model
+- [x] BGE-VL-base and BGE-VL-large models
+- [x] BGE-VL-MLLM model
 - [ ] MegaPairs Dataset
 - [ ] Evaluation code
 - [ ] Fine-tuning code
 
 
 ## Introduction
-In this project, we introduce **MegaPairs**, a novel data synthesis method that leverages open-domain images to create *heterogeneous KNN triplets* for universal multimodal retrieval. Our MegaPairs dataset contains over 26 million triplets, and we have trained a series of multimodal retrieval models, **MMRets**, including MMRet-CLIP (base and large) and MMRet-MLLM.
+In this work, we introduce **MegaPairs**, a novel data synthesis method that leverages open-domain images to create *heterogeneous KNN triplets* for universal multimodal retrieval. Our MegaPairs dataset contains over 26 million triplets, and we have trained a series of multimodal retrieval models, **BGE-VL**, including BGE-VL-CLIP (base and large) and BGE-VL-MLLM.
 
-MMRets achieve state-of-the-art performance on four popular zero-shot composed image retrieval benchmarks and the massive multimodal embedding benchmark (MMEB). Extensive experiments demonstrate the ***efficiency, scalability, and generalization*** features of MegaPairs. Please refer to our [paper](https://arxiv.org/abs/2412.14475) for more details.
+BGE-VL achieve state-of-the-art performance on four popular zero-shot composed image retrieval benchmarks and the massive multimodal embedding benchmark (MMEB). Extensive experiments demonstrate the ***efficiency, scalability, and generalization*** features of MegaPairs. Please refer to our [paper](https://arxiv.org/abs/2412.14475) for more details.
 
 ## Model Usage
 
-### 1. MMRet-CLIP Models
-You can easily use MMRet-CLIP models based on ```transformers```
+### 1. BGE-VL-CLIP Models
+You can easily use BGE-VL-CLIP models based on ```transformers```
 ```python
 import torch
 from transformers import AutoModel
 
-MODEL_NAME = "JUNJIE99/MMRet-base" # or "JUNJIE99/MMRet-large"
+MODEL_NAME = "BAAI/BGE-VL-base" # or "BAAI/BGE-VL-large"
 
 model = AutoModel.from_pretrained(MODEL_NAME, trust_remote_code=True) # You must set trust_remote_code=True
 model.set_processor(MODEL_NAME)
@@ -86,13 +87,18 @@ with torch.no_grad():
 print(scores)
 ```
 
-### 2. MMRet-MLLM Models
+See the [demo](./retrieval_demo.ipynb) for a complete example of using BGE-VL for multimodel retrieval.
+
+
+### 2. BGE-VL-MLLM Models
+
+
 ```python
 import torch
 from transformers import AutoModel
 from PIL import Image
 
-MODEL_NAME= "JUNJIE99/MMRet-MLLM-S1"
+MODEL_NAME= "BAAI/BGE-VL-MLLM-S1"
 
 model = AutoModel.from_pretrained(MODEL_NAME, trust_remote_code=True)
 model.eval()
@@ -127,30 +133,30 @@ print(scores)
 ## Model Performance
 ### Zero-Shot Composed Image Retrieval
 
-MMRet sets a new performance benchmark in zero-shot composed image retrieval tasks. On the CIRCO benchmark, our MMRet-base model, with only 149 million parameters, surpasses all previous models, including those with 50 times more parameters. Additionally, MMRet-MLLM achieves an 8.1% improvement over the previous state-of-the-art model.
+BGE-VL sets a new performance benchmark in zero-shot composed image retrieval tasks. On the CIRCO benchmark, our BGE-VL-base model, with only 149 million parameters, surpasses all previous models, including those with 50 times more parameters. Additionally, BGE-VL-MLLM achieves an 8.1% improvement over the previous state-of-the-art model.
 
 <img src="./assets/res-zs-cir.png" width="800">
 
 ### Zero-Shot Performance on MMEB
 
-MMRet-MLLM achieves state-of-the-art zero-shot performance on the Massive Multimodal Embedding Benchmark (MMEB), despite being trained only on the ImageText-to-Image paradigm. This demonstrates the excellent generalization capability of MegaPairs for multimodal embedding.
+BGE-VL-MLLM achieves state-of-the-art zero-shot performance on the Massive Multimodal Embedding Benchmark (MMEB), despite being trained only on the ImageText-to-Image paradigm. This demonstrates the excellent generalization capability of MegaPairs for multimodal embedding.
 
 <img src="./assets/res-zs-mmeb.png" width="800">
 
 ### Fine-Tuning Performance on MMEB
 
-After fine-tuning on downstream tasks, MMRet-MLLM maintains its leading performance. Notably, it surpasses the previous state-of-the-art by 7.1% on the MMEB out-of-distribution (OOD) set. These results demonstrate the robust generalization capability of MMRet-MLLM and highlight the potential of MegaPairs as foundational training data for universal multimodal embedding.
+After fine-tuning on downstream tasks, BGE-VL-MLLM maintains its leading performance. Notably, it surpasses the previous state-of-the-art by 7.1% on the MMEB out-of-distribution (OOD) set. These results demonstrate the robust generalization capability of BGE-VL-MLLM and highlight the potential of MegaPairs as foundational training data for universal multimodal embedding.
 
 <img src="./assets/res-ft-mmeb.png" width="800">
 
 ### Performance Scaling
-MegaPairs showcases **scalability**: MMRet-base improves as training data increases. It also demonstrates **efficiency**: with just 0.5M training samples, MMRet-base significantly outperforms MagicLens, which uses the same CLIP-base backbone and was trained on 36.7M samples.
+MegaPairs showcases **scalability**: BGE-VL-base improves as training data increases. It also demonstrates **efficiency**: with just 0.5M training samples, BGE-VL-base significantly outperforms MagicLens, which uses the same CLIP-base backbone and was trained on 36.7M samples.
 
 <img src="./assets/res-scaling.png" width="800">
 
 
 ## License
-The annotations for MegaPairs and the MMRet models are released under the [MIT License](LICENSE). The images in MegaPairs originate from the [Recap-Datacomp](https://huggingface.co/datasets/UCSC-VLAA/Recap-DataComp-1B), which is released under the CC BY 4.0 license.
+The annotations for MegaPairs and the BGE-VL models are released under the [MIT License](LICENSE). The images in MegaPairs originate from the [Recap-Datacomp](https://huggingface.co/datasets/UCSC-VLAA/Recap-DataComp-1B), which is released under the CC BY 4.0 license.
 
 
 
@@ -164,4 +170,4 @@ If you find this repository useful, please consider giving a star ⭐ and citati
   journal={arXiv preprint arXiv:2412.14475},
   year={2024}
 }
-```
+```