multilingual e5 small

提供商intfloat
分类sentence-similarity
许可证mit
下载量49.0K
星标3

简介

multilingual-e5-small 是一款轻量级且高性能的多语言文本向量化模型。它将文本转化为高维向量，通过计算余弦相似度来衡量语义相关性，而非简单的关键词匹配。该模型在 100 多种语言上表现均衡，非常适合中文用户构建 RAG（检索增强生成）系统，用于实现高效的文档检索、语义去重或知识库问答。由于参数量小，它可以在 CPU 上流畅运行，部署成本极低，是替代大型 Embedding 模型的理想选择。

核心亮点

  • 支持百种语言，中文语义检索效果出色
  • 模型体积小，推理速度快，部署门槛极低
  • RAG 架构中理想的向量化组件，提升检索精度
  • MIT 协议开源，可自由用于商业项目

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("intfloat/multilingual-e5-small")
tokenizer = AutoTokenizer.from_pretrained("intfloat/multilingual-e5-small")

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 intfloat/multilingual-e5-small

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 intfloat/multilingual-e5-small 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('intfloat/multilingual-e5-small')

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/intfloat/multilingual-e5-small

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/intfloat/multilingual-e5-small

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('intfloat/multilingual-e5-small')
tokenizer = AutoTokenizer.from_pretrained('intfloat/multilingual-e5-small')

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 intfloat/multilingual-e5-small

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 intfloat/multilingual-e5-small 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('intfloat/multilingual-e5-small')

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/intfloat/multilingual-e5-small.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/intfloat/multilingual-e5-small.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', 'intfloat/multilingual-e5-small')

完整文档

来源: HuggingFace

---
language:

  • multilingual

  • af

  • am

  • ar

  • as

  • az

  • be

  • bg

  • bn

  • br

  • bs

  • ca

  • cs

  • cy

  • da

  • de

  • el

  • en

  • eo

  • es

  • et

  • eu

  • fa

  • fi

  • fr

  • fy

  • ga

  • gd

  • gl

  • gu

  • ha

  • he

  • hi

  • hr

  • hu

  • hy

  • id

  • is

  • it

  • ja

  • jv

  • ka

  • kk

  • km

  • kn

  • ko

  • ku

  • ky

  • la

  • lo

  • lt

  • lv

  • mg

  • mk

  • ml

  • mn

  • mr

  • ms

  • my

  • ne

  • nl

  • 'no'

  • om

  • or

  • pa

  • pl

  • ps

  • pt

  • ro

  • ru

  • sa

  • sd

  • si

  • sk

  • sl

  • so

  • sq

  • sr

  • su

  • sv

  • sw

  • ta

  • te

  • th

  • tl

  • tr

  • ug

  • uk

  • ur

  • uz

  • vi

  • xh

  • yi

  • zh

license: mit
model-index:
  • name: intfloat/multilingual-e5-small

results:
- dataset:
config: en
name: MTEB AmazonCounterfactualClassification (en)
revision: e8379541af4e31359cca9fbcf4b00f2671dba205
split: test
type: mteb/amazon_counterfactual
metrics:
- type: accuracy
value: 73.79104477611939
- type: ap
value: 36.9996434842022
- type: f1
value: 67.95453679103099
task:
type: Classification
- dataset:
config: de
name: MTEB AmazonCounterfactualClassification (de)
revision: e8379541af4e31359cca9fbcf4b00f2671dba205
split: test
type: mteb/amazon_counterfactual
metrics:
- type: accuracy
value: 71.64882226980728
- type: ap
value: 82.11942130026586
- type: f1
value: 69.87963421606715
task:
type: Classification
- dataset:
config: en-ext
name: MTEB AmazonCounterfactualClassification (en-ext)
revision: e8379541af4e31359cca9fbcf4b00f2671dba205
split: test
type: mteb/amazon_counterfactual
metrics:
- type: accuracy
value: 75.8095952023988
- type: ap
value: 24.46869495579561
- type: f1
value: 63.00108480037597
task:
type: Classification
- dataset:
config: ja
name: MTEB AmazonCounterfactualClassification (ja)
revision: e8379541af4e31359cca9fbcf4b00f2671dba205
split: test
type: mteb/amazon_counterfactual
metrics:
- type: accuracy
value: 64.186295503212
- type: ap
value: 15.496804690197042
- type: f1
value: 52.07153895475031
task:
type: Classification
- dataset:
config: default
name: MTEB AmazonPolarityClassification
revision: e2d317d38cd51312af73b3d32a06d1a08b442046
split: test
type: mteb/amazon_polarity
metrics:
- type: accuracy
value: 88.699325
- type: ap
value: 85.27039559917269
- type: f1
value: 88.65556295032513
task:
type: Classification
- dataset:
config: en
name: MTEB AmazonReviewsClassification (en)
revision: 1399c76144fd37290681b995c656ef9b2e06e26d
split: test
type: mteb/amazon_reviews_multi
metrics:
- type: accuracy
value: 44.69799999999999
- type: f1
value: 43.73187348654165
task:
type: Classification
- dataset:
config: de
name: MTEB AmazonReviewsClassification (de)
revision: 1399c76144fd37290681b995c656ef9b2e06e26d
split: test
type: mteb/amazon_reviews_multi
metrics:
- type: accuracy
value: 40.245999999999995
- type: f1
value: 39.3863530637684
task:
type: Classification
- dataset:
config: es
name: MTEB AmazonReviewsClassification (es)
revision: 1399c76144fd37290681b995c656ef9b2e06e26d
split: test
type: mteb/amazon_reviews_multi
metrics:
- type: accuracy
value: 40.394
- type: f1
value: 39.301223469483446
task:
type: Classification
- dataset:
config: fr
name: MTEB AmazonReviewsClassification (fr)
revision: 1399c76144fd37290681b995c656ef9b2e06e26d
split: test
type: mteb/amazon_reviews_multi
metrics:
- type: accuracy
value: 38.864
- type: f1
value: 37.97974261868003
task:
type: Classification
- dataset:
config: ja
name: MTEB AmazonReviewsClassification (ja)
revision: 1399c76144fd37290681b995c656ef9b2e06e26d
split: test
type: mteb/amazon_reviews_multi
metrics:
- type: accuracy
value: 37.682
- type: f1
value: 37.07399369768313
task:
type: Classification
- dataset:
config: zh
name: MTEB AmazonReviewsClassification (zh)
revision: 1399c76144fd37290681b995c656ef9b2e06e26d
split: test
type: mteb/amazon_reviews_multi
metrics:
- type: accuracy
value: 37.504
- type: f1
value: 36.62317273874278
task:
type: Classification
- dataset:
config: default
name: MTEB ArguAna
revision: None
split: test
type: arguana
metrics:
- type: map_at_1
value: 19.061
- type: map_at_10
value: 31.703
- type: map_at_100
value: 32.967
- type: map_at_1000
value: 33.001000000000005
- type: map_at_3
value: 27.466
- type: map_at_5
value: 29.564
- type: mrr_at_1
value: 19.559
- type: mrr_at_10
value: 31.874999999999996
- type: mrr_at_100
value: 33.146
- type: mrr_at_1000
value: 33.18
- type: mrr_at_3
value: 27.667
- type: mrr_at_5
value: 29.74
- type: ndcg_at_1
value: 19.061
- type: ndcg_at_10
value: 39.062999999999995
- type: ndcg_at_100
value: 45.184000000000005
- type: ndcg_at_1000
value: 46.115
- type: ndcg_at_3
value: 30.203000000000003
- type: ndcg_at_5
value: 33.953
- type: precision_at_1
value: 19.061
- type: precision_at_10
value: 6.279999999999999
- type: precision_at_100
value: 0.9129999999999999
- type: precision_at_1000
value: 0.099
- type: precision_at_3
value: 12.706999999999999
- type: precision_at_5
value: 9.431000000000001
- type: recall_at_1
value: 19.061
- type: recall_at_10
value: 62.802
- type: recall_at_100
value: 91.323
- type: recall_at_1000
value: 98.72
- type: recall_at_3
value: 38.122
- type: recall_at_5
value: 47.155
task:
type: Retrieval
- dataset:
config: default
name: MTEB ArxivClusteringP2P
revision: a122ad7f3f0291bf49cc6f4d32aa80929df69d5d
split: test
type: mteb/arxiv-clustering-p2p
metrics:
- type: v_measure
value: 39.22266660528253
task:
type: Clustering
- dataset:
config: default
name: MTEB ArxivClusteringS2S
revision: f910caf1a6075f7329cdf8c1a6135696f37dbd53
split: test
type: mteb/arxiv-clustering-s2s
metrics:
- type: v_measure
value: 30.79980849482483
task:
type: Clustering
- dataset:
config: default
name: MTEB AskUbuntuDupQuestions
revision: 2000358ca161889fa9c082cb41daa8dcfb161a54
split: test
type: mteb/askubuntudupquestions-reranking
metrics:
- type: map
value: 57.8790068352054
- type: mrr
value: 71.78791276436706
task:
type: Reranking
- dataset:
config: default
name: MTEB BIOSSES
revision: d3fb88f8f02e40887cd149695127462bbcf29b4a
split: test
type: mteb/biosses-sts
metrics:
- type: cos_sim_pearson
value: 82.36328364043163
- type: cos_sim_spearman
value: 82.26211536195868
- type: euclidean_pearson
value: 80.3183865039173
- type: euclidean_spearman
value: 79.88495276296132
- type: manhattan_pearson
value: 80.14484480692127
- type: manhattan_spearman
value: 80.39279565980743
task:
type: STS
- dataset:
config: de-en
name: MTEB BUCC (de-en)
revision: d51519689f32196a32af33b075a01d0e7c51e252
split: test
type: mteb/bucc-bitext-mining
metrics:
- type: accuracy
value: 98.0375782881002
- type: f1
value: 97.86012526096033
- type: precision
value: 97.77139874739039
- type: recall
value: 98.0375782881002
task:
type: Bi

查看官方来源