Why did we open-source our inference engine? Read the post

intfloat/e5-small-v2

Text Embeddings by Weakly-Supervised Contrastive Pre-training. Liang Wang, Nan Yang, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang, Rangan Majumder, Furu Wei, arXiv 2022

Overview

Architecture
BERT
Parameters
33M
Tasks
Encode
Outputs
Dense
Dimensions
Dense: 384
Max Sequence Length
512 tokens
License
mit
Languages
en

Benchmarks

CQADupstackPhysicsRetrieval

scientific retrieval en

Duplicate question retrieval from StackExchange Physics

Corpus: 38,314 Queries: 1,039
Quality
map at 10 0.3521
mrr at 10 0.4049
ndcg at 10 0.4052
Performance L4 b1 c16
Corpus 42.5K tok/s
Corpus p50 47.8ms
Query 3.1K tok/s
Query p50 48.7ms
Reference →

CosQA

technology retrieval en

Code search with natural language queries

Corpus: 6,267 Queries: 500
Quality
map at 10 0.2161
mrr at 10 0.2341
ndcg at 10 0.2864
Performance L4 b1 c16
Corpus 16.7K tok/s
Corpus p50 49.7ms
Query 1.6K tok/s
Query p50 50.3ms
Reference →

FiQA2018

finance retrieval en

Financial opinion mining and question answering

Corpus: 57,599 Queries: 648
Quality
map at 10 0.2840
mrr at 10 0.4390
ndcg at 10 0.3560
Performance L4 b1 c16
Corpus 41.5K tok/s
Corpus p50 57.8ms
Query 3.3K tok/s
Query p50 48.5ms
Reference →

LegalBenchConsumerContractsQA

legal retrieval en

Question answering on consumer contracts

Corpus: 153 Queries: 396
Quality
map at 10 0.6679
mrr at 10 0.6687
ndcg at 10 0.7278
Performance L4 b1 c16
Corpus 115.7K tok/s
Corpus p50 63.0ms
Query 4.0K tok/s
Query p50 57.9ms
Reference →

NFCorpus

medical retrieval en

Biomedical literature search from NutritionFacts.org

Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.3195
map at 10 0.1142
mrr at 10 0.5174
Performance L4 b1 c16
Corpus 95.9K tok/s
Corpus p50 45.8ms
Query 1.5K tok/s
Query p50 43.2ms
Reference →

NanoFiQA2018Retrieval

finance retrieval en

Smaller subset of the FiQA financial QA dataset

Quality
ndcg at 10 0.4299
map at 10 0.3531
mrr at 10 0.4611
Performance L4 b1 c16
Corpus 58.3K tok/s
Corpus p50 40.3ms
Query 4.1K tok/s
Query p50 35.0ms
Reference →

SCIDOCS

scientific retrieval en

Citation prediction, document classification, and recommendation for scientific papers

Corpus: 25,656 Queries: 1,000
Quality
map at 10 0.0643
mrr at 10 0.2128
ndcg at 10 0.1162
Performance L4 b1 c16
Corpus 57.1K tok/s
Corpus p50 49.9ms
Query 3.2K tok/s
Query p50 47.2ms
Reference →

SciFact

scientific retrieval en

Scientific claim verification using research literature

Corpus: 5,183 Queries: 300
Quality
map at 10 0.6308
mrr at 10 0.6498
ndcg at 10 0.6753
Performance L4 b1 c16
Corpus 85.1K tok/s
Corpus p50 47.6ms
Query 4.6K tok/s
Query p50 48.1ms
Reference →

StackOverflowQA

technology retrieval en

Programming question answering from Stack Overflow

Corpus: 19,931 Queries: 1,994
Quality
map at 10 0.8113
mrr at 10 0.8113
ndcg at 10 0.8349
Performance L4 b1 c16
Corpus 63.0K tok/s
Corpus p50 53.7ms
Query 55.0K tok/s
Query p50 75.1ms
Reference →

Self-hosted inference for search & document processing

Cut API costs by 50x, boost quality with 85+ SOTA models, and keep your data in your own cloud.

Github 2.0K

Contact us

Tell us about your use case and we'll get back to you shortly.