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

opensearch-project/opensearch-neural-sparse-encoding-v1

The model should be selected considering search relevance, model inference and retrieval efficiency(FLOPS). We benchmark models' zero-shot performance on a subset of BEIR benchmark: TrecCovid,NFCorpus,NQ,HotpotQA,FiQA,ArguAna,Touche,DBPedia,SCIDOCS,FEVER,Climate FEVER,SciFact,Quora.

Overview

Architecture
BERT
Parameters
133M
Tasks
Encode
Outputs
Sparse
Dimensions
Sparse: 30,522
Max Sequence Length
512 tokens
License
apache-2.0
Languages
en

Benchmarks

CQADupstackPhysicsRetrieval

scientific retrieval en

Duplicate question retrieval from StackExchange Physics

Corpus: 38,314 Queries: 1,039
Quality
ndcg at 10 0.3652
map at 10 0.3090
mrr at 10 0.3574
Performance L4 b1 c16
Corpus 28.5K tok/s
Corpus p50 69.1ms
Query 2.4K tok/s
Query p50 66.5ms
Reference →

CosQA

technology retrieval en

Code search with natural language queries

Corpus: 6,267 Queries: 500
Quality
ndcg at 10 0.2435
map at 10 0.1868
mrr at 10 0.1985
Performance L4 b1 c16
Corpus 15.2K tok/s
Corpus p50 52.0ms
Query 2.0K tok/s
Query p50 46.6ms
Reference →

FiQA2018

finance retrieval en

Financial opinion mining and question answering

Corpus: 57,599 Queries: 648
Quality
ndcg at 10 0.2632
map at 10 0.2004
mrr at 10 0.3273
Performance L4 b1 c16
Corpus 40.2K tok/s
Corpus p50 59.4ms
Query 3.6K tok/s
Query p50 50.0ms
Reference →

LegalBenchConsumerContractsQA

legal retrieval en

Question answering on consumer contracts

Corpus: 153 Queries: 396
Quality
ndcg at 10 0.7272
map at 10 0.6626
mrr at 10 0.6635
Performance L4 b1 c16
Corpus 89.1K tok/s
Corpus p50 86.7ms
Query 5.1K tok/s
Query p50 50.1ms
Reference →

NFCorpus

medical retrieval en

Biomedical literature search from NutritionFacts.org

Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.3600
map at 10 0.1403
mrr at 10 0.5690
Performance L4 b1 c16
Corpus 58.7K tok/s
Corpus p50 76.0ms
Query 1.5K tok/s
Query p50 51.0ms
Reference →

SCIDOCS

scientific retrieval en

Citation prediction, document classification, and recommendation for scientific papers

Corpus: 25,656 Queries: 1,000
Quality
ndcg at 10 0.1537
map at 10 0.0877
mrr at 10 0.2688
Performance L4 b1 c16
Corpus 44.8K tok/s
Corpus p50 63.2ms
Query 3.5K tok/s
Query p50 49.5ms
Reference →

SciFact

scientific retrieval en

Scientific claim verification using research literature

Corpus: 5,183 Queries: 300
Quality
ndcg at 10 0.6998
map at 10 0.6522
mrr at 10 0.6684
Performance L4 b1 c16
Corpus 54.1K tok/s
Corpus p50 75.7ms
Query 5.2K tok/s
Query p50 50.0ms
Reference →

StackOverflowQA

technology retrieval en

Programming question answering from Stack Overflow

Corpus: 19,931 Queries: 1,994
Quality
ndcg at 10 0.7703
map at 10 0.7392
mrr at 10 0.7392
Performance L4 b1 c16
Corpus 52.6K tok/s
Corpus p50 68.8ms
Query 65.6K tok/s
Query p50 71.9ms
Reference →

Open source inference for agents

Open-source inference for the models behind your agents. Run it yourself, or let us run it for you.

Github 2.1K

Contact us

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

Apply for an inference grant

Free capacity on our hosted cluster for selected projects. Tell us what you run and we reply by email.