The crispy rerank family from mixedbread ai.
# mxbai-colbert-v1 This is our first English ColBERT model, which is built upon our sentence embedding model [mixedbread-ai/mxbai-embed-large-v1](https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1). You can learn more about the models in our [blog post](https://www.mixedbread.ai/blog/). ## Quickstart Currently, the best way to use it is with the official [ColBERT](https://github.com/stanford-futuredata/ColBERT) library. `python -m pip install -U colbert-ai[faiss-gpu]` Here, we provide several ways to use it. ### 1. Generate Embeddings ```python from huggingface_hub import snapshot_download from colbert.modeling.checkpoint import Checkpoint from colbert.infra import Run, RunConfig, ColBERTConfig # To ensure the total files are cached locally snapshot_download(repo_id="mixedbread-ai/mxbai-colbert-v1") # load mixedbread colbert ckpt = Checkpoint("mixedbread-ai/mxbai-colbert-v1", colbert_config=ColBERTConfig()) # encode query and documents query = "Who wrote 'To Kill a Mockingbird'?" documents = [ "'To Kill a Mockingbird' is a novel by Harper Lee published in 1960. It was immediately successful, winning the Pulitzer Prize, and has become a classic of modern American literature.", "The novel 'Moby-Dick' was written by Herman Melville and first published in 1851. It is considered a masterpiece of American literature and deals with complex themes of obsession, revenge, and the conflict between good and evil.", "Harper Lee, an American novelist widely known for her novel 'To Kill a Mockingbird', was born in 1926 in Monroeville, Alabama. She received the Pulitzer Prize for Fiction in 1961.", "Jane Austen was an English novelist known primarily for her six major novels, which interpret, critique and comment upon the British landed gentry at the end of the 18th century.", "The 'Harry Potter' series, which consists of seven fantasy novels written by British author J.K. Rowling, is among the most popular and critically acclaimed books of the modern era.", "'The Great Gatsby', a novel written by American author F. Scott Fitzgerald, was published in 1925. The story is set in the Jazz Age and follows the life of millionaire Jay Gatsby and his pursuit of Daisy Buchanan." ] query_vectors = ckpt.queryFromText([query], bsize=16) doc_vectors = ckpt.docFromText(documents, bsize=16) ``` ### 2. Index & Search 1) Index ```python from huggingface_hub import snapshot_download from colbert import Indexer from colbert.infra import Run, RunConfig, ColBERTConfig # To ensure the total files are cached locally snapshot_download(repo_id="mixedbread-ai/mxbai-colbert-v1") gpu_count = 1 documents = [ "'To Kill a Mockingbird' is a novel by Harper Lee published in 1960. It was immediately successful, winning the Pulitzer Prize, and has become a classic of modern American literature.", "The novel 'Moby-Dick' was written by Herman Melville and first published in 1851. It is considered a masterpiece of American literature and deals with complex themes of obsession, revenge, and the conflict between good and evil.", "Harper Lee, an American novelist widely known for her novel 'To Kill a Mockingbird', was born in 1926 in Monroeville, Alabama. She received the Pulitzer Prize for Fiction in 1961.", "Jane Austen was an English novelist known primarily for her six major novels, which interpret, critique and comment upon the British landed gentry at the end of the 18th century.", "The 'Harry Potter' series, which consists of seven fantasy novels written by British author J.K. Rowling, is among the most popular and critically acclaimed books of the modern era.", "'The Great Gatsby', a novel written by American author F. Scott Fitzgerald, was published in 1925. The story is set in the Jazz Age and follows the life of millionaire Jay Gatsby and his pursuit of Daisy Buchanan." ] with Run().context(RunConfig(nranks=gpu_count, gpus=gpu_count, experiment='experiments')): config = ColBERTConfig( doc_maxlen=512 ) indexer = Indexer( checkpoint="mixedbread-ai/mxbai-colbert-v1", config=config, ) indexer.index(name='demo', collection=documents) ``` 2) Search ```python from colbert import Searcher from colbert.infra import Run, RunConfig, ColBERTConfig gpu_count = 1 with Run().context(RunConfig(nranks=1, experiment='experiments')): config = ColBERTConfig( query_maxlen=128 ) searcher = Searcher( index='demo', config=config ) query = "Who wrote 'To Kill a Mockingbird'?" results = searcher.search(query, k=3) ``` ## Using API You’ll be able to use the models through our API as well. The API is coming soon and will have some exciting features. Stay tuned! ## Evaluation ### 1. Reranking Performance **Setup:** we use BM25 as the first-stage retrieval model, and then use ColBERT for reranking. Following common practice, we report NDCG@10 as the metrics. Here, we compare our model with two widely used ColBERT models, as follows: | Model | ColBERTv2 | Jina-ColBERT-V1 | Mxbai-ColBERT-V1| | ---------------------- | -------- | ---------- | ---------- | | dbpedia-entity | 31.8 | **42.2** | 40.6 | | fiqa | 23.6 | 35.6 | **35.9** | | nfcorpus | 33.8 | **36.7** | 36.4 | | nq | 30.6 | 51.3 | **51.4** | | scidocs | 14.9 | 15.4 | **17.0** | | scifact | 67.9 | 70.2 | **71.5** | | trec-covid | 59.5 | 75.0 | **81.0** | | webis-touche2020 | 44.2 | 32.1 | 31.7 | | signal1m | **33.2** | 30.9 | 33.1 | | trec-news | 46.0 | 45.2 | **47.1** | | robust04 | 47.5 | **47.7** | 47.5 | | avg | 39.4 | 43.8 | **44.8** | Find more in our [blog-post](https://www.mixedbread.ai/blog/) and on this [spreadsheet](https://docs.google.com/spreadsheets/d/1ZT_KN40PnHQa21hTdrk4_9GCnqm916lJJz3W83mo1og/edit?usp=sharing). ### 2. Retrieval Performance ColBERT is mainly used for reranking. Here, we also test our model's performance on retrieval tasks. Due to resource limitations, we only test our model on three beir tasks. NDCG@10 servers as the main metric. | Model | ColBERTv2 | Jina-ColBERT-V1 | Mxbai-ColBERT-V1| | ---------------------- | -------- | ---------- | ---------- | | scifact | 68.9 | 70.1 | **71.3** | | nfcorpus | 33.7 | 33.8 | **36.5** | | trec-covid | 72.6 | 75.0 | **80.5** | Although our ColBERT also performs well on retrieval, we recommend using our embedding model [mixedbread-ai/mxbai-embed-large-v1](https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1) for retrieval. ## Community Please join our [Discord Community](https://discord.gg/jDfMHzAVfU) and share your feedback and thoughts! We are here to help and also always happy to chat. ## License Apache 2.0