msmarco-MiniLM-L12-cos-v5

这是一个 sentence-transformers 模型：它将句子和段落映射到 768 维的密集向量空间，专为语义搜索设计。它基于 MS MARCO Passages 数据集 的 50 万个（查询，答案）对进行训练。关于语义搜索的入门介绍，请参阅：SBERT.net - Semantic Search

使用方法 (Sentence-Transformers)

安装 sentence-transformers 后，使用此模型变得非常简单：

pip install -U sentence-transformers

然后你可以这样使用该模型：

from sentence_transformers import SentenceTransformer, util

query = "How many people live in London?"
docs = ["Around 9 Million people live in London", "London is known for its financial district"]

#Load the model
model = SentenceTransformer('sentence-transformers/msmarco-MiniLM-L12-cos-v5')

#Encode query and documents
query_emb = model.encode(query)
doc_emb = model.encode(docs)

#Compute dot score between query and all document embeddings
scores = util.dot_score(query_emb, doc_emb)[0].cpu().tolist()

#Combine docs & scores
doc_score_pairs = list(zip(docs, scores))

#Sort by decreasing score
doc_score_pairs = sorted(doc_score_pairs, key=lambda x: x[1], reverse=True)

#Output passages & scores
for doc, score in doc_score_pairs:
    print(score, doc)

使用方法 (HuggingFace Transformers)

如果没有使用 sentence-transformers，你可以这样使用该模型：首先将输入通过 transformer 模型，然后需要在上下文词嵌入之上应用正确的池化操作。

from transformers import AutoTokenizer, AutoModel
import torch
import torch.nn.functional as F

#Mean Pooling - Take average of all tokens
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output.last_hidden_state #First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


#Encode text
def encode(texts):
    # Tokenize sentences
    encoded_input = tokenizer(texts, padding=True, truncation=True, return_tensors='pt')

    # Compute token embeddings
    with torch.no_grad():
        model_output = model(**encoded_input, return_dict=True)

    # Perform pooling
    embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

    # Normalize embeddings
    embeddings = F.normalize(embeddings, p=2, dim=1)
	
    return embeddings


# Sentences we want sentence embeddings for
query = "How many people live in London?"
docs = ["Around 9 Million people live in London", "London is known for its financial district"]

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained("sentence-transformers/msmarco-MiniLM-L12-cos-v5")
model = AutoModel.from_pretrained("sentence-transformers/msmarco-MiniLM-L12-cos-v5")

#Encode query and docs
query_emb = encode(query)
doc_emb = encode(docs)

#Compute dot score between query and all document embeddings
scores = torch.mm(query_emb, doc_emb.transpose(0, 1))[0].cpu().tolist()

#Combine docs & scores
doc_score_pairs = list(zip(docs, scores))

#Sort by decreasing score
doc_score_pairs = sorted(doc_score_pairs, key=lambda x: x[1], reverse=True)

#Output passages & scores
for doc, score in doc_score_pairs:
    print(score, doc)

技术细节

以下是关于如何使用此模型的一些技术细节：

注意：当使用 sentence-transformers 加载时，此模型会产生长度为 1 的归一化嵌入。在这种情况下，dot-product 和 cosine-similarity 是等价的。建议使用 dot-product，因为它更快。Euclidean 距离与 dot-product 成正比，也可以使用。

引用 & 作者

此模型由 sentence-transformers 训练。

如果你觉得这个模型有帮助，欢迎引用我们的论文 Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks：

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "http://arxiv.org/abs/1908.10084",
}