bge small en v1.5

提供商BAAI
分类feature-extraction
许可证mit
下载量18.7K
星标5

简介

bge-small-en-v1.5 是由北京智源人工智能研究院（BAAI）推出的轻量级英文文本向量模型。它并非对话模型，而是专门用于将文本转化为高维向量的特征提取工具。在 RAG（检索增强生成）架构中，它扮演着“索引员”的角色，能高效地将用户问题与知识库文档进行语义匹配。该模型体积小、推理速度极快，非常适合部署在端侧设备或对响应延迟要求极高的生产环境中，是构建英文语义搜索和文档问答系统的理想底层组件。

核心亮点

  • 智源 BAAI 出品，英文语义检索能力顶尖
  • 模型体积轻量，推理延迟极低，部署成本低
  • RAG 架构核心组件，大幅提升文档检索精度
  • MIT 协议开源，企业级商业应用无压力

使用方法

安装依赖
# 安装 Hugging Face transformers
pip install transformers torch
SDK 使用
# 使用 transformers 加载模型
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("BAAI/bge-small-en-v1.5")
tokenizer = AutoTokenizer.from_pretrained("BAAI/bge-small-en-v1.5")

Hugging Face 下载

我们推荐使用命令行或者 Hugging Face Hub SDK 来进行模型的下载。

操作指引：在下载前，请先通过如下命令安装 huggingface_hub：

操作指引 
pip install -U huggingface_hub

命令行下载

下载完整模型库

下载完整模型库 
huggingface-cli download BAAI/bge-small-en-v1.5

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例）

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例） 
huggingface-cli download BAAI/bge-small-en-v1.5 config.json --local-dir ./dir

更多命令行下载选项，可参见官方文档

SDK 下载

SDK 下载 
# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('BAAI/bge-small-en-v1.5')

Git 下载

请确保 lfs 已经被正确安装

Git 下载 
git lfs install
git clone https://huggingface.co/BAAI/bge-small-en-v1.5

如果您希望跳过 lfs 大文件下载，可以使用如下命令

跳过 LFS 
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/BAAI/bge-small-en-v1.5

模型文件托管在 Hugging Face Hub，使用 HF CLI / SDK / Git 直接下载，不经过本站。

PyTorch / Transformers 使用

安装 Transformers

安装 Transformers 
pip install -U transformers torch

模型加载和推理

模型加载和推理 
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('BAAI/bge-small-en-v1.5')
tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-small-en-v1.5')

模型下载

我们推荐使用命令行或者 ModelScope SDK 来进行模型的下载。

操作指引：在下载前，请先通过如下命令安装 ModelScope：

操作指引 
pip install modelscope

命令行下载

下载完整模型库

下载完整模型库 
modelscope download --model BAAI/bge-small-en-v1.5

下载单个文件到指定本地文件夹（以下载 README.md 到当前路径下 dir 目录为例）

下载单个文件到指定本地文件夹（以下载 README.md 到当前路径下 dir 目录为例） 
modelscope download --model BAAI/bge-small-en-v1.5 README.md --local_dir ./dir

更多更丰富的命令行下载选项，可参见具体文档

SDK 下载

SDK 下载 
# 模型下载
from modelscope import snapshot_download
model_dir = snapshot_download('BAAI/bge-small-en-v1.5')

Git 下载

请确保 lfs 已经被正确安装

Git 下载 
git lfs install
git clone https://www.modelscope.cn/BAAI/bge-small-en-v1.5.git

如果您希望跳过 lfs 大文件下载，可以使用如下命令

跳过 LFS 
GIT_LFS_SKIP_SMUDGE=1 git clone https://www.modelscope.cn/BAAI/bge-small-en-v1.5.git

ModelScope 模型页直接下载模型文件；无需将模型文件放在本站服务器。

Notebook 快速开发

下载并安装 ModelScope library

下载并安装 ModelScope library 
pip install "modelscope[audio,cv,nlp,multi-modal,science]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

模型加载和推理

模型加载和推理 
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

p = pipeline('text-generation', 'BAAI/bge-small-en-v1.5')

完整文档

来源: HuggingFace

---
tags:

  • sentence-transformers

  • feature-extraction

  • sentence-similarity

  • transformers

  • mteb

model-index:
  • name: bge-small-en-v1.5

results:
- task:
type: Classification
dataset:
type: mteb/amazon_counterfactual
name: MTEB AmazonCounterfactualClassification (en)
config: en
split: test
revision: e8379541af4e31359cca9fbcf4b00f2671dba205
metrics:
- type: accuracy
value: 73.79104477611939
- type: ap
value: 37.21923821573361
- type: f1
value: 68.0914945617093
- task:
type: Classification
dataset:
type: mteb/amazon_polarity
name: MTEB AmazonPolarityClassification
config: default
split: test
revision: e2d317d38cd51312af73b3d32a06d1a08b442046
metrics:
- type: accuracy
value: 92.75377499999999
- type: ap
value: 89.46766124546022
- type: f1
value: 92.73884001331487
- task:
type: Classification
dataset:
type: mteb/amazon_reviews_multi
name: MTEB AmazonReviewsClassification (en)
config: en
split: test
revision: 1399c76144fd37290681b995c656ef9b2e06e26d
metrics:
- type: accuracy
value: 46.986
- type: f1
value: 46.55936786727896
- task:
type: Retrieval
dataset:
type: arguana
name: MTEB ArguAna
config: default
split: test
revision: None
metrics:
- type: map_at_1
value: 35.846000000000004
- type: map_at_10
value: 51.388
- type: map_at_100
value: 52.132999999999996
- type: map_at_1000
value: 52.141000000000005
- type: map_at_3
value: 47.037
- type: map_at_5
value: 49.579
- type: mrr_at_1
value: 36.558
- type: mrr_at_10
value: 51.658
- type: mrr_at_100
value: 52.402
- type: mrr_at_1000
value: 52.410000000000004
- type: mrr_at_3
value: 47.345
- type: mrr_at_5
value: 49.797999999999995
- type: ndcg_at_1
value: 35.846000000000004
- type: ndcg_at_10
value: 59.550000000000004
- type: ndcg_at_100
value: 62.596
- type: ndcg_at_1000
value: 62.759
- type: ndcg_at_3
value: 50.666999999999994
- type: ndcg_at_5
value: 55.228
- type: precision_at_1
value: 35.846000000000004
- type: precision_at_10
value: 8.542
- type: precision_at_100
value: 0.984
- type: precision_at_1000
value: 0.1
- type: precision_at_3
value: 20.389
- type: precision_at_5
value: 14.438
- type: recall_at_1
value: 35.846000000000004
- type: recall_at_10
value: 85.42
- type: recall_at_100
value: 98.43499999999999
- type: recall_at_1000
value: 99.644
- type: recall_at_3
value: 61.166
- type: recall_at_5
value: 72.191
- task:
type: Clustering
dataset:
type: mteb/arxiv-clustering-p2p
name: MTEB ArxivClusteringP2P
config: default
split: test
revision: a122ad7f3f0291bf49cc6f4d32aa80929df69d5d
metrics:
- type: v_measure
value: 47.402770198163594
- task:
type: Clustering
dataset:
type: mteb/arxiv-clustering-s2s
name: MTEB ArxivClusteringS2S
config: default
split: test
revision: f910caf1a6075f7329cdf8c1a6135696f37dbd53
metrics:
- type: v_measure
value: 40.01545436974177
- task:
type: Reranking
dataset:
type: mteb/askubuntudupquestions-reranking
name: MTEB AskUbuntuDupQuestions
config: default
split: test
revision: 2000358ca161889fa9c082cb41daa8dcfb161a54
metrics:
- type: map
value: 62.586465273207196
- type: mrr
value: 74.42169019038825
- task:
type: STS
dataset:
type: mteb/biosses-sts
name: MTEB BIOSSES
config: default
split: test
revision: d3fb88f8f02e40887cd149695127462bbcf29b4a
metrics:
- type: cos_sim_pearson
value: 85.1891186537969
- type: cos_sim_spearman
value: 83.75492046087288
- type: euclidean_pearson
value: 84.11766204805357
- type: euclidean_spearman
value: 84.01456493126516
- type: manhattan_pearson
value: 84.2132950502772
- type: manhattan_spearman
value: 83.89227298813377
- task:
type: Classification
dataset:
type: mteb/banking77
name: MTEB Banking77Classification
config: default
split: test
revision: 0fd18e25b25c072e09e0d92ab615fda904d66300
metrics:
- type: accuracy
value: 85.74025974025975
- type: f1
value: 85.71493566466381
- task:
type: Clustering
dataset:
type: mteb/biorxiv-clustering-p2p
name: MTEB BiorxivClusteringP2P
config: default
split: test
revision: 65b79d1d13f80053f67aca9498d9402c2d9f1f40
metrics:
- type: v_measure
value: 38.467181385006434
- task:
type: Clustering
dataset:
type: mteb/biorxiv-clustering-s2s
name: MTEB BiorxivClusteringS2S
config: default
split: test
revision: 258694dd0231531bc1fd9de6ceb52a0853c6d908
metrics:
- type: v_measure
value: 34.719496037339056
- task:
type: Retrieval
dataset:
type: BeIR/cqadupstack
name: MTEB CQADupstackAndroidRetrieval
config: default
split: test
revision: None
metrics:
- type: map_at_1
value: 29.587000000000003
- type: map_at_10
value: 41.114
- type: map_at_100
value: 42.532
- type: map_at_1000
value: 42.661
- type: map_at_3
value: 37.483
- type: map_at_5
value: 39.652
- type: mrr_at_1
value: 36.338
- type: mrr_at_10
value: 46.763
- type: mrr_at_100
value: 47.393
- type: mrr_at_1000
value: 47.445
- type: mrr_at_3
value: 43.538
- type: mrr_at_5
value: 45.556000000000004
- type: ndcg_at_1
value: 36.338
- type: ndcg_at_10
value: 47.658
- type: ndcg_at_100
value: 52.824000000000005
- type: ndcg_at_1000
value: 54.913999999999994
- type: ndcg_at_3
value: 41.989
- type: ndcg_at_5
value: 44.944
- type: precision_at_1
value: 36.338
- type: precision_at_10
value: 9.156
- type: precision_at_100
value: 1.4789999999999999
- type: precision_at_1000
value: 0.196
- type: precision_at_3
value: 20.076
- type: precision_at_5
value: 14.85
- type: recall_at_1
value: 29.587000000000003
- type: recall_at_10
value: 60.746
- type: recall_at_100
value: 82.157
- type: recall_at_1000
value: 95.645
- type: recall_at_3
value: 44.821
- type: recall_at_5
value: 52.819
- task:
type: Retrieval
dataset:
type: BeIR/cqadupstack
name: MTEB CQADupstackEnglishRetrieval
config: default
split: test
revision: None
metrics:
- type: map_at_1
value: 30.239
- type: map_at_10
value: 39.989000000000004
- type: map_at_100
value: 41.196
- type: map_at_1000
value: 41.325
- type: map_at_3
value: 37.261
- type: map_at_5
value: 38.833
- type: mrr_at_1
value: 37.516
- type: mrr_at_10
value: 46.177
- type: mrr_at_100
value: 46.806
- type: mrr_at_1000
value: 46.849000000000004
- type: mrr_at_3
value: 44.002
- type: mrr_at_5
value: 45.34
- type: ndcg_at_1
value: 37.516
- type: ndcg_at_10
value: 45.586
- type: ndcg_at_100
value: 49.897000000000006
- type: ndcg_at_1000
value: 51.955
- type: ndcg_at_3
value: 41.684
- type: ndcg_at_5
value: 43.617
- type: precision_at_1
value: 37.516
- type: precision_at_10
value: 8.522
- type: precision_at_100
value: 1.374
- type: precision_at_1000
value: 0.184

查看官方来源