Minicpm Embedding

language:

• zh
• en base_model: openbmb/MiniCPM-2B-sft-bf16 model-index:
• name: MiniCPM-Embedding results:
  • task: type: Retrieval dataset: type: mteb/arguana name: MTEB ArguAna config: default split: test revision: c22ab2a51041ffd869aaddef7af8d8215647e41a metrics:
    • type: ndcg_at_10 value: 64.65
  • task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackRetrieval config: default split: test revision: 4ffe81d471b1924886b33c7567bfb200e9eec5c4 metrics:
    • type: ndcg_at_10 value: 46.53
  • task: type: Retrieval dataset: type: mteb/climate-fever name: MTEB ClimateFEVER config: default split: test revision: 47f2ac6acb640fc46020b02a5b59fdda04d39380 metrics:
    • type: ndcg_at_10 value: 35.55
  • task: type: Retrieval dataset: type: mteb/dbpedia name: MTEB DBPedia config: default split: test revision: c0f706b76e590d620bd6618b3ca8efdd34e2d659 metrics:
    • type: ndcg_at_10 value: 47.82
  • task: type: Retrieval dataset: type: mteb/fever name: MTEB FEVER config: default split: test revision: bea83ef9e8fb933d90a2f1d5515737465d613e12 metrics:
    • type: ndcg_at_10 value: 90.76
  • task: type: Retrieval dataset: type: mteb/fiqa name: MTEB FiQA2018 config: default split: test revision: 27a168819829fe9bcd655c2df245fb19452e8e06 metrics:
    • type: ndcg_at_10 value: 56.64
  • task: type: Retrieval dataset: type: mteb/hotpotqa name: MTEB HotpotQA config: default split: test revision: ab518f4d6fcca38d87c25209f94beba119d02014 metrics:
    • type: ndcg_at_10 value: 78.11
  • task: type: Retrieval dataset: type: mteb/msmarco name: MTEB MSMARCO config: default split: dev revision: c5a29a104738b98a9e76336939199e264163d4a0 metrics:
    • type: ndcg_at_10 value: 43.93
  • task: type: Retrieval dataset: type: mteb/nfcorpus name: MTEB NFCorpus config: default split: test revision: ec0fa4fe99da2ff19ca1214b7966684033a58814 metrics:
    • type: ndcg_at_10 value: 39.77
  • task: type: Retrieval dataset: type: mteb/nq name: MTEB NQ config: default split: test revision: b774495ed302d8c44a3a7ea25c90dbce03968f31 metrics:
    • type: ndcg_at_10 value: 69.29
  • task: type: Retrieval dataset: type: mteb/quora name: MTEB QuoraRetrieval config: default split: test revision: None metrics:
    • type: ndcg_at_10 value: 89.97
  • task: type: Retrieval dataset: type: mteb/scidocs name: MTEB SCIDOCS config: default split: test revision: None metrics:
    • type: ndcg_at_10 value: 22.38
  • task: type: Retrieval dataset: type: mteb/scifact name: MTEB SciFact config: default split: test revision: 0228b52cf27578f30900b9e5271d331663a030d7 metrics:
    • type: ndcg_at_10 value: 86.6
  • task: type: Retrieval dataset: type: mteb/trec-covid name: MTEB TRECCOVID config: default split: test revision: None metrics:
    • type: ndcg_at_10 value: 81.32
  • task: type: Retrieval dataset: type: mteb/touche2020 name: MTEB Touche2020 config: default split: test revision: a34f9a33db75fa0cbb21bb5cfc3dae8dc8bec93f metrics:
    • type: ndcg_at_10 value: 25.08
  • task: type: Retrieval dataset: type: C-MTEB/CmedqaRetrieval name: MTEB CmedqaRetrieval config: default split: dev revision: cd540c506dae1cf9e9a59c3e06f42030d54e7301 metrics:
    • type: ndcg_at_10 value: 46.05
  • task: type: Retrieval dataset: type: C-MTEB/CovidRetrieval name: MTEB CovidRetrieval config: default split: dev revision: 1271c7809071a13532e05f25fb53511ffce77117 metrics:
    • type: ndcg_at_10 value: 92.01
  • task: type: Retrieval dataset: type: C-MTEB/DuRetrieval name: MTEB DuRetrieval config: default split: dev revision: a1a333e290fe30b10f3f56498e3a0d911a693ced metrics:
    • type: ndcg_at_10 value: 90.98
  • task: type: Retrieval dataset: type: C-MTEB/EcomRetrieval name: MTEB EcomRetrieval config: default split: dev revision: 687de13dc7294d6fd9be10c6945f9e8fec8166b9 metrics:
    • type: ndcg_at_10 value: 70.21
  • task: type: Retrieval dataset: type: C-MTEB/MMarcoRetrieval name: MTEB MMarcoRetrieval config: default split: dev revision: 539bbde593d947e2a124ba72651aafc09eb33fc2 metrics:
    • type: ndcg_at_10 value: 85.55
  • task: type: Retrieval dataset: type: C-MTEB/MedicalRetrieval name: MTEB MedicalRetrieval config: default split: dev revision: 2039188fb5800a9803ba5048df7b76e6fb151fc6 metrics:
    • type: ndcg_at_10 value: 63.91
  • task: type: Retrieval dataset: type: C-MTEB/T2Retrieval name: MTEB T2Retrieval config: default split: dev revision: 8731a845f1bf500a4f111cf1070785c793d10e64 metrics:
    • type: ndcg_at_10 value: 87.33
  • task: type: Retrieval dataset: type: C-MTEB/VideoRetrieval name: MTEB VideoRetrieval config: default split: dev revision: 58c2597a5943a2ba48f4668c3b90d796283c5639 metrics:
    • type: ndcg_at_10 value: 78.05 pipeline_tag: feature-extraction tags:
• mteb
• sentence-transformers library_name: transformers

MiniCPM-Embedding

MiniCPM-Embedding 是面壁智能与清华大学自然语言处理实验室（THUNLP）、东北大学信息检索小组（NEUIR）共同开发的中英双语言文本嵌入模型，有如下特点：

• 出色的中文、英文检索能力。
• 出色的中英跨语言检索能力。

MiniCPM-Embedding 基于 MiniCPM-2B-sft-bf16 训练，结构上采取双向注意力和 Weighted Mean Pooling [1]。采取多阶段训练方式，共使用包括开源数据、机造数据、闭源数据在内的约 600 万条训练数据。

欢迎关注 RAG 套件系列：

• 检索模型：MiniCPM-Embedding
• 重排模型：MiniCPM-Reranker
• 面向 RAG 场景的 LoRA 插件：MiniCPM3-RAG-LoRA

MiniCPM-Embedding is a bilingual & cross-lingual text embedding model developed by ModelBest Inc. , THUNLP and NEUIR , featuring:

• Exceptional Chinese and English retrieval capabilities.
• Outstanding cross-lingual retrieval capabilities between Chinese and English.

MiniCPM-Embedding is trained based on MiniCPM-2B-sft-bf16 and incorporates bidirectional attention and Weighted Mean Pooling [1] in its architecture. The model underwent multi-stage training using approximately 6 million training examples, including open-source, synthetic, and proprietary data.

We also invite you to explore the RAG toolkit series:

• Retrieval Model: MiniCPM-Embedding
• Re-ranking Model: MiniCPM-Reranker
• LoRA Plugin for RAG scenarios: MiniCPM3-RAG-LoRA

[1] Muennighoff, N. (2022). Sgpt: Gpt sentence embeddings for semantic search. arXiv preprint arXiv:2202.08904.

模型信息 Model Information

• 模型大小：2.4B

• 嵌入维度：2304

• 最大输入token数：512

• Model Size: 2.4B

• Embedding Dimension: 2304

• Max Input Tokens: 512

使用方法 Usage

输入格式 Input Format

本模型支持 query 侧指令，格式如下：

MiniCPM-Embedding supports query-side instructions in the following format:

Instruction: {{ instruction }} Query: {{ query }}

例如：

For example:

Instruction: 为这个医学问题检索相关回答。Query: 咽喉癌的成因是什么？

Instruction: Given a claim about climate change, retrieve documents that support or refute the claim. Query: However the warming trend is slower than most climate models have forecast.

也可以不提供指令，即采取如下格式：

MiniCPM-Embedding also works in instruction-free mode in the following format:

Query: {{ query }}

我们在 BEIR 与 C-MTEB/Retrieval 上测试时使用的指令见 instructions.json，其他测试不使用指令。文档侧直接输入文档原文。

When running evaluation on BEIR and C-MTEB/Retrieval, we use instructions in instructions.json. For other evaluations, we do not use instructions. On the document side, we directly use the bare document as the input.

环境要求 Requirements

transformers==4.37.2

示例脚本 Demo

Huggingface Transformers

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

model_name = "openbmb/MiniCPM-Embedding"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModel.from_pretrained(model_name, trust_remote_code=True, torch_dtype=torch.float16).to("cuda")
# You can also use the following line to enable the Flash Attention 2 implementation
# model = AutoModel.from_pretrained(model_name, trust_remote_code=True, attn_implementation="flash_attention_2", torch_dtype=torch.float16).to("cuda")
model.eval()

# 由于在 `model.forward` 中缩放了最终隐层表示，此处的 mean pooling 实际上起到了 weighted mean pooling 的作用
# As we scale hidden states in `model.forward`, mean pooling here actually works as weighted mean pooling
def mean_pooling(hidden, attention_mask):
    s = torch.sum(hidden * attention_mask.unsqueeze(-1).float(), dim=1)
    d = attention_mask.sum(dim=1, keepdim=True).float()
    reps = s / d
    return reps

@torch.no_grad()
def encode(input_texts):
    batch_dict = tokenizer(input_texts, max_length=512, padding=True, truncation=True, return_tensors='pt', return_attention_mask=True).to("cuda")
    
    outputs = model(**batch_dict)
    attention_mask = batch_dict["attention_mask"]
    hidden = outputs.last_hidden_state

    reps = mean_pooling(hidden, attention_mask)   
    embeddings = F.normalize(reps, p=2, dim=1).detach().cpu().numpy()
    return embeddings

queries = ["中国的首都是哪里？"]
passages = ["beijing", "shanghai"]


INSTRUCTION = "Query: "
queries = [INSTRUCTION + query for query in queries]

embeddings_query = encode(queries)
embeddings_doc = encode(passages)

scores = (embeddings_query @ embeddings_doc.T)
print(scores.tolist())  # [[0.3535913825035095, 0.18596848845481873]]

Sentence Transformers

import torch
from sentence_transformers import SentenceTransformer

model_name = "openbmb/MiniCPM-Embedding"
model = SentenceTransformer(model_name, trust_remote_code=True, model_kwargs={ "torch_dtype": torch.float16})
# You can also use the following line to enable the Flash Attention 2 implementation
# model = SentenceTransformer(model_name, trust_remote_code=True, attn_implementation="flash_attention_2", model_kwargs={ "torch_dtype": torch.float16})

queries = ["中国的首都是哪里？"]
passages = ["beijing", "shanghai"]

INSTRUCTION = "Query: "

embeddings_query = model.encode(queries, prompt=INSTRUCTION)
embeddings_doc = model.encode(passages)

scores = (embeddings_query @ embeddings_doc.T)
print(scores.tolist())  # [[0.35365450382232666, 0.18592746555805206]]

实验结果 Evaluation Results

中文与英文检索结果 CN/EN Retrieval Results

中英跨语言检索结果 CN-EN Cross-lingual Retrieval Results

许可证 License

• 本仓库中代码依照 Apache-2.0 协议开源。
• MiniCPM-Embedding 模型权重的使用则需要遵循 MiniCPM 模型协议。
• MiniCPM-Embedding 模型权重对学术研究完全开放。如需将模型用于商业用途，请填写此问卷。

• The code in this repo is released under the Apache-2.0 License.
• The usage of MiniCPM-Embedding model weights must strictly follow MiniCPM Model License.md.
• The models and weights of MiniCPM-Embedding are completely free for academic research. After filling out a "questionnaire" for registration, MiniCPM-Embedding weights are also available for free commercial use.