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

intfloat/multilingual-e5-large-instruct

Multilingual E5 Text Embeddings: A Technical Report. Liang Wang, Nan Yang, Xiaolong Huang, Linjun Yang, Rangan Majumder, Furu Wei, arXiv 2024

Overview

Architecture
XLM-RoBERTa
Parameters
560M
Tasks
Encode
Outputs
Dense
Dimensions
Dense: 1,024
Max Sequence Length
512 tokens
License
mit
Languages
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

Benchmarks

CQADupstackPhysicsRetrieval

scientific retrieval en

Duplicate question retrieval from StackExchange Physics

Corpus: 38,314 Queries: 1,039
Quality
map at 10 0.3967
mrr at 10 0.4542
ndcg at 10 0.4571
Performance L4 b1 c16
Corpus 25.8K tok/s
Corpus p50 82.4ms
Query 2.8K tok/s
Query p50 57.0ms
Reference →

CosQA

technology retrieval en

Code search with natural language queries

Corpus: 6,267 Queries: 500
Quality
map at 10 0.2844
mrr at 10 0.3037
ndcg at 10 0.3616
Performance L4 b1 c16
Corpus 12.5K tok/s
Corpus p50 66.4ms
Query 1.5K tok/s
Query p50 58.6ms
Reference →

FiQA2018

finance retrieval en

Financial opinion mining and question answering

Corpus: 57,599 Queries: 648
Quality
map at 10 0.3932
mrr at 10 0.5631
ndcg at 10 0.4773
Performance L4 b1 c16
Corpus 29.4K tok/s
Corpus p50 89.2ms
Query 3.0K tok/s
Query p50 60.0ms
Reference →

LegalBenchConsumerContractsQA

legal retrieval en

Question answering on consumer contracts

Corpus: 153 Queries: 396
Quality
map at 10 0.6339
mrr at 10 0.6329
ndcg at 10 0.6961
Performance L4 b1 c16
Corpus 45.3K tok/s
Corpus p50 178.3ms
Query 4.5K tok/s
Query p50 58.0ms
Reference →

NFCorpus

medical retrieval en

Biomedical literature search from NutritionFacts.org

Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.3521
map at 10 0.1313
mrr at 10 0.5378
Performance L4 b1 c16
Corpus 35.7K tok/s
Corpus p50 137.0ms
Query 1.3K tok/s
Query p50 61.6ms
Reference →

NanoFiQA2018Retrieval

finance retrieval en

Smaller subset of the FiQA financial QA dataset

Quality
ndcg at 10 0.5539
map at 10 0.4744
mrr at 10 0.6084
Performance L4 b1 c16
Corpus 30.6K tok/s
Corpus p50 87.2ms
Query 3.3K tok/s
Query p50 51.9ms
Reference →

SCIDOCS

scientific retrieval en

Citation prediction, document classification, and recommendation for scientific papers

Corpus: 25,656 Queries: 1,000
Quality
map at 10 0.1097
mrr at 10 0.3185
ndcg at 10 0.1872
Performance L4 b1 c16
Corpus 28.8K tok/s
Corpus p50 106.9ms
Query 2.6K tok/s
Query p50 61.6ms
Reference →

SciFact

scientific retrieval en

Scientific claim verification using research literature

Corpus: 5,183 Queries: 300
Quality
map at 10 0.6749
mrr at 10 0.6867
ndcg at 10 0.7202
Performance L4 b1 c16
Corpus 31.8K tok/s
Corpus p50 137.5ms
Query 4.3K tok/s
Query p50 60.0ms
Reference →

StackOverflowQA

technology retrieval en

Programming question answering from Stack Overflow

Corpus: 19,931 Queries: 1,994
Quality
map at 10 0.8965
mrr at 10 0.8965
ndcg at 10 0.9108
Performance L4 b1 c16
Corpus 29.3K tok/s
Corpus p50 112.5ms
Query 39.4K tok/s
Query p50 124.1ms
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.