FlagEmbedding holds a whole curriculum for retrieval, embedding models, RAG, etc. This section is currently being actively updated. No matter you are new to NLP or a veteran, we hope you can find something helpful!
If you are new to embedding and retrieval, check out the 5 minute quick start!
Tutorial roadmap
This module includes tutorials and demos showing how to use BGE and Sentence Transformers, as well as other embedding related topics.
- Intro to embedding model
- BGE series
- Usage of BGE
- BGE-M3
- BGE-ICL
- ...
In this part, we show popular similarity functions and techniques about searching.
- Similarity metrics
- ...
Although not included in the quick start, indexing is a very important part in practical cases. This module shows how to use popular libraries like Faiss and Milvus to do indexing.
- Intro to Faiss
- Using GPU in Faiss
- Indexes
- Quantizers
- Milvus
- ...
In this module, we'll show the full pipeline of evaluating an embedding model, as well as popular benchmarks like MTEB and C-MTEB.
- Evaluate MSMARCO
- Intro to MTEB
- MTEB Leaderboard Eval
- C-MTEB
- ...
To balance accuracy and efficiency tradeoff, many retrieval system use a more efficient retriever to quickly narrow down the candidates. Then use more accurate models do reranking for the final results.
- Intro to reranker
- ...