ldwang commited on
Commit
f03589c
1 Parent(s): fcd022c
Files changed (1) hide show
  1. README.md +34 -16
README.md CHANGED
@@ -35,25 +35,36 @@ FlagEmbedding can map any text to a low-dimensional dense vector which can be us
35
  And it also can be used in vector databases for LLMs.
36
 
37
  ************* 🌟**Updates**🌟 *************
38
- - 09/15/2023: Release [paper](https://arxiv.org/pdf/2309.07597.pdf) and [dataset](https://data.baai.ac.cn/details/BAAI-MTP).
39
- - 09/12/2023: New Release:
 
 
40
  - **New reranker model**: release cross-encoder models `BAAI/bge-reranker-base` and `BAAI/bge-reranker-large`, which are more powerful than embedding model. We recommend to use/fine-tune them to re-rank top-k documents returned by embedding models.
41
  - **update embedding model**: release `bge-*-v1.5` embedding model to alleviate the issue of the similarity distribution, and enhance its retrieval ability without instruction.
42
- - 09/07/2023: Update [fine-tune code](https://github.com/FlagOpen/FlagEmbedding/blob/master/FlagEmbedding/baai_general_embedding/README.md): Add script to mine hard negatives and support adding instruction during fine-tuning.
 
 
 
 
 
 
43
  - 08/09/2023: BGE Models are integrated into **Langchain**, you can use it like [this](#using-langchain); C-MTEB **leaderboard** is [available](https://huggingface.co/spaces/mteb/leaderboard).
44
- - 08/05/2023: Release base-scale and small-scale models, **best performance among the models of the same size 🤗**
45
- - 08/02/2023: Release `bge-large-*`(short for BAAI General Embedding) Models, **rank 1st on MTEB and C-MTEB benchmark!** :tada: :tada:
46
- - 08/01/2023: We release the [Chinese Massive Text Embedding Benchmark](https://github.com/FlagOpen/FlagEmbedding/blob/master/C_MTEB) (**C-MTEB**), consisting of 31 test dataset.
 
 
47
 
48
 
49
  ## Model List
50
 
51
  `bge` is short for `BAAI general embedding`.
52
 
53
- | Model | Language | | Description | query instruction for retrieval\* |
54
  |:-------------------------------|:--------:| :--------:| :--------:|:--------:|
55
- | [BAAI/bge-reranker-large](https://huggingface.co/BAAI/bge-reranker-large) | Chinese and English | [Inference](#usage-for-reranker) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker) | a cross-encoder model which is more accurate but less efficient \** | |
56
- | [BAAI/bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base) | Chinese and English | [Inference](#usage-for-reranker) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker) | a cross-encoder model which is more accurate but less efficient \** | |
 
57
  | [BAAI/bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
58
  | [BAAI/bge-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
59
  | [BAAI/bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
@@ -68,11 +79,15 @@ And it also can be used in vector databases for LLMs.
68
  | [BAAI/bge-small-zh](https://huggingface.co/BAAI/bge-small-zh) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | a small-scale model but with competitive performance | `为这个句子生成表示以用于检索相关文章:` |
69
 
70
 
71
- \*: If you need to search the relevant passages to a query, we suggest to add the instruction to the query; in other cases, no instruction is needed, just use the original query directly. In all cases, **no instruction** needs to be added to passages.
72
 
73
- \**: Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. To balance the accuracy and time cost, cross-encoder is widely used to re-rank top-k documents retrieved by other simple models.
74
  For examples, use bge embedding model to retrieve top 100 relevant documents, and then use bge reranker to re-rank the top 100 document to get the final top-3 results.
75
 
 
 
 
 
76
  ## Frequently asked questions
77
 
78
  <details>
@@ -109,7 +124,11 @@ please select an appropriate similarity threshold based on the similarity distri
109
  <summary>3. When does the query instruction need to be used</summary>
110
 
111
  <!-- ### When does the query instruction need to be used -->
112
-
 
 
 
 
113
  For a retrieval task that uses short queries to find long related documents,
114
  it is recommended to add instructions for these short queries.
115
  **The best method to decide whether to add instructions for queries is choosing the setting that achieves better performance on your task.**
@@ -369,7 +388,7 @@ which is more accurate than embedding model (i.e., bi-encoder) but more time-con
369
  Therefore, it can be used to re-rank the top-k documents returned by embedding model.
370
  We train the cross-encoder on a multilingual pair data,
371
  The data format is the same as embedding model, so you can fine-tune it easily following our [example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker).
372
- More details pelease refer to [./FlagEmbedding/reranker/README.md](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/reranker)
373
 
374
 
375
  ## Contact
@@ -379,7 +398,8 @@ You also can email Shitao Xiao(stxiao@baai.ac.cn) and Zheng Liu(liuzheng@baai.ac
379
 
380
  ## Citation
381
 
382
- If you find our work helpful, please cite us:
 
383
  ```
384
  @misc{bge_embedding,
385
  title={C-Pack: Packaged Resources To Advance General Chinese Embedding},
@@ -394,5 +414,3 @@ If you find our work helpful, please cite us:
394
  ## License
395
  FlagEmbedding is licensed under the [MIT License](https://github.com/FlagOpen/FlagEmbedding/blob/master/LICENSE). The released models can be used for commercial purposes free of charge.
396
 
397
-
398
-
 
35
  And it also can be used in vector databases for LLMs.
36
 
37
  ************* 🌟**Updates**🌟 *************
38
+ - 10/12/2023: Release [LLM-Embedder](./FlagEmbedding/llm_embedder/README.md), a unified embedding model to support diverse retrieval augmentation needs for LLMs. [Paper](https://arxiv.org/pdf/2310.07554.pdf) :fire:
39
+ - 09/15/2023: The [technical report](https://arxiv.org/pdf/2309.07597.pdf) of BGE has been released
40
+ - 09/15/2023: The [masive training data](https://data.baai.ac.cn/details/BAAI-MTP) of BGE has been released
41
+ - 09/12/2023: New models:
42
  - **New reranker model**: release cross-encoder models `BAAI/bge-reranker-base` and `BAAI/bge-reranker-large`, which are more powerful than embedding model. We recommend to use/fine-tune them to re-rank top-k documents returned by embedding models.
43
  - **update embedding model**: release `bge-*-v1.5` embedding model to alleviate the issue of the similarity distribution, and enhance its retrieval ability without instruction.
44
+
45
+
46
+ <details>
47
+ <summary>More</summary>
48
+ <!-- ### More -->
49
+
50
+ - 09/07/2023: Update [fine-tune code](https://github.com/FlagOpen/FlagEmbedding/blob/master/FlagEmbedding/baai_general_embedding/README.md): Add script to mine hard negatives and support adding instruction during fine-tuning.
51
  - 08/09/2023: BGE Models are integrated into **Langchain**, you can use it like [this](#using-langchain); C-MTEB **leaderboard** is [available](https://huggingface.co/spaces/mteb/leaderboard).
52
+ - 08/05/2023: Release base-scale and small-scale models, **best performance among the models of the same size 🤗**
53
+ - 08/02/2023: Release `bge-large-*`(short for BAAI General Embedding) Models, **rank 1st on MTEB and C-MTEB benchmark!** :tada: :tada:
54
+ - 08/01/2023: We release the [Chinese Massive Text Embedding Benchmark](https://github.com/FlagOpen/FlagEmbedding/blob/master/C_MTEB) (**C-MTEB**), consisting of 31 test dataset.
55
+
56
+ </details>
57
 
58
 
59
  ## Model List
60
 
61
  `bge` is short for `BAAI general embedding`.
62
 
63
+ | Model | Language | | Description | query instruction for retrieval [1] |
64
  |:-------------------------------|:--------:| :--------:| :--------:|:--------:|
65
+ | [BAAI/llm-embedder](https://huggingface.co/BAAI/llm-embedder) | English | [Inference](./FlagEmbedding/llm_embedder/README.md) [Fine-tune](./FlagEmbedding/llm_embedder/README.md) | a unified embedding model to support diverse retrieval augmentation needs for LLMs | See [README](./FlagEmbedding/llm_embedder/README.md) |
66
+ | [BAAI/bge-reranker-large](https://huggingface.co/BAAI/bge-reranker-large) | Chinese and English | [Inference](#usage-for-reranker) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker) | a cross-encoder model which is more accurate but less efficient [2] | |
67
+ | [BAAI/bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base) | Chinese and English | [Inference](#usage-for-reranker) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker) | a cross-encoder model which is more accurate but less efficient [2] | |
68
  | [BAAI/bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
69
  | [BAAI/bge-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
70
  | [BAAI/bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
 
79
  | [BAAI/bge-small-zh](https://huggingface.co/BAAI/bge-small-zh) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | a small-scale model but with competitive performance | `为这个句子生成表示以用于检索相关文章:` |
80
 
81
 
82
+ [1\]: If you need to search the relevant passages to a query, we suggest to add the instruction to the query; in other cases, no instruction is needed, just use the original query directly. In all cases, **no instruction** needs to be added to passages.
83
 
84
+ [2\]: Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. To balance the accuracy and time cost, cross-encoder is widely used to re-rank top-k documents retrieved by other simple models.
85
  For examples, use bge embedding model to retrieve top 100 relevant documents, and then use bge reranker to re-rank the top 100 document to get the final top-3 results.
86
 
87
+ All models have been uploaded to Huggingface Hub, and you can see them at https://huggingface.co/BAAI.
88
+ If you cannot open the Huggingface Hub, you also can download the models at https://model.baai.ac.cn/models .
89
+
90
+
91
  ## Frequently asked questions
92
 
93
  <details>
 
124
  <summary>3. When does the query instruction need to be used</summary>
125
 
126
  <!-- ### When does the query instruction need to be used -->
127
+
128
+ For the `bge-*-v1.5`, we improve its retrieval ability when not using instruction.
129
+ No instruction only has a slight degradation in retrieval performance compared with using instruction.
130
+ So you can generate embedding without instruction in all cases for convenience.
131
+
132
  For a retrieval task that uses short queries to find long related documents,
133
  it is recommended to add instructions for these short queries.
134
  **The best method to decide whether to add instructions for queries is choosing the setting that achieves better performance on your task.**
 
388
  Therefore, it can be used to re-rank the top-k documents returned by embedding model.
389
  We train the cross-encoder on a multilingual pair data,
390
  The data format is the same as embedding model, so you can fine-tune it easily following our [example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker).
391
+ More details please refer to [./FlagEmbedding/reranker/README.md](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/reranker)
392
 
393
 
394
  ## Contact
 
398
 
399
  ## Citation
400
 
401
+ If you find this repository useful, please consider giving a star :star: and citation
402
+
403
  ```
404
  @misc{bge_embedding,
405
  title={C-Pack: Packaged Resources To Advance General Chinese Embedding},
 
414
  ## License
415
  FlagEmbedding is licensed under the [MIT License](https://github.com/FlagOpen/FlagEmbedding/blob/master/LICENSE). The released models can be used for commercial purposes free of charge.
416