facebookresearch/DPR

Health v2 · maintenance only

Flags: no_releases archived no_license

Adoption not part of the score

1869 stars · 311 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

DPR is a set of tools and pretrained models for dense passage retrieval in open-domain question answering, based on the EMNLP 2020 paper from Facebook AI Research. It provides bi-encoder retriever training, an extractive reader/ranker, FAISS-based inference, and data processing utilities.

Use cases

• retrieve relevant passages for open-domain question answering
• train a dense bi-encoder retriever on QA datasets
• build semantic search over Wikipedia with FAISS
• run extractive QA with a reader and ranker pipeline
• mine hard negatives for retrieval model training
• reproduce DPR paper results with pretrained checkpoints

When to choose

• you need state-of-the-art dense passage retrieval for open-domain QA research
• you want pretrained DPR checkpoints and Wikipedia embeddings
• you are reproducing or extending the DPR paper's experiments

When to avoid

• you need a production-ready, actively maintained retrieval library
• you want a simple plug-and-play RAG pipeline rather than research tooling
• you need multi-vector or late-interaction retrieval beyond bi-encoders

Facets

library · maturity maintenance

rag search-engine machine-learning nlp machine-learning python dense-retrieval question-answering bi-encoder faiss research-code open-domain-qa natural-language-processing retrieval-augmented-generation search

Member repositories

For agents

markdown · JSON · MCP: product_card(name="facebookresearch/DPR")