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

intfloat/e5-base-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
109M
Tasks
Encode
Outputs
Dense
Dimensions
Dense: 768
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 0.3560
mrr at 10 0.4064
ndcg at 10 0.4109
Performance L4 b1 c16
Corpus 40.2K tok/s
Corpus p50 48.6ms
Query 3.0K tok/s
Query p50 45.9ms
Reference →

CosQA

technology retrieval en

Code search with natural language queries

Corpus: 6,267 Queries: 500
Quality
map at 10 0.2299
mrr at 10 0.2607
ndcg at 10 0.3083
Performance L4 b1 c16
Corpus 18.8K tok/s
Corpus p50 43.2ms
Query 1.9K tok/s
Query p50 43.0ms
Reference →

FiQA2018

finance retrieval en

Financial opinion mining and question answering

Corpus: 57,599 Queries: 648
Quality
map at 10 0.2758
mrr at 10 0.4220
ndcg at 10 0.3512
Performance L4 b1 c16
Corpus 45.2K tok/s
Corpus p50 52.3ms
Query 3.4K tok/s
Query p50 45.4ms
Reference →

LegalBenchConsumerContractsQA

legal retrieval en

Question answering on consumer contracts

Corpus: 153 Queries: 396
Quality
map at 10 0.6609
mrr at 10 0.6596
ndcg at 10 0.7200
Performance L4 b1 c16
Corpus 109.6K tok/s
Corpus p50 68.3ms
Query 5.1K tok/s
Query p50 45.0ms
Reference →

NFCorpus

medical retrieval en

Biomedical literature search from NutritionFacts.org

Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.3541
map at 10 0.1311
mrr at 10 0.5549
Performance L4 b1 c16
Corpus 79.2K tok/s
Corpus p50 57.9ms
Query 1.5K tok/s
Query p50 42.8ms
Reference →

NanoFiQA2018Retrieval

finance retrieval en

Smaller subset of the FiQA financial QA dataset

Quality
ndcg at 10 0.4603
map at 10 0.3791
mrr at 10 0.5059
Performance L4 b1 c16
Corpus 33.6K tok/s
Corpus p50 63.5ms
Query 3.5K tok/s
Query p50 49.2ms
Reference →

SCIDOCS

scientific retrieval en

Citation prediction, document classification, and recommendation for scientific papers

Corpus: 25,656 Queries: 1,000
Quality
map at 10 0.0651
mrr at 10 0.2070
ndcg at 10 0.1175
Performance L4 b1 c16
Corpus 53.2K tok/s
Corpus p50 53.2ms
Query 3.4K tok/s
Query p50 45.6ms
Reference →

SciFact

scientific retrieval en

Scientific claim verification using research literature

Corpus: 5,183 Queries: 300
Quality
map at 10 0.6381
mrr at 10 0.6512
ndcg at 10 0.6840
Performance L4 b1 c16
Corpus 67.5K tok/s
Corpus p50 59.2ms
Query 4.6K tok/s
Query p50 47.7ms
Reference →

StackOverflowQA

technology retrieval en

Programming question answering from Stack Overflow

Corpus: 19,931 Queries: 1,994
Quality
map at 10 0.8593
mrr at 10 0.8593
ndcg at 10 0.8781
Performance L4 b1 c16
Corpus 55.7K tok/s
Corpus p50 59.0ms
Query 63.4K tok/s
Query p50 63.1ms
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.