---
title: contriever
canonical_url: "https://www.modelscope.cn/models/facebook/contriever"
md_url: "https://www.modelscope.cn/models/facebook/contriever.md"
repository: facebook/contriever
last_updated: 2025-05-20
model_type:
  - bert
architectures:
  - Contriever
library_name:
  - pytorch
frameworks:
  - pytorch
inference_backends:
  - "deploy_task emb"
  - "sglang 0.5.2"
downloads: 973
stars: 0
---

# contriever

> contriever - facebook 在 ModelScope 开源的模型。This model has been trained without supervision following the approach described in Towards Unsupervised Dense Information Retrieval with Contrastive Learning. The associated GitHub repository is available here…

facebook/contriever 是 ModelScope 魔搭社区上的机器学习模型，可用 deploy_task emb、sglang 0.5.2 部署。

- **Repository**: facebook/contriever
- **Inference backends**: deploy_task emb, sglang 0.5.2
- **Downloads**: 973
- **Stars**: 0
- **Last updated**: 2025-05-20

Source: https://www.modelscope.cn/models/facebook/contriever

---

This model has been trained without supervision following the approach described in [Towards Unsupervised Dense Information Retrieval with Contrastive Learning](https://arxiv.org/abs/2112.09118). The associated GitHub repository is available here https://github.com/facebookresearch/contriever.

## Usage (HuggingFace Transformers)
Using the model directly available in HuggingFace transformers requires to add a mean pooling operation to obtain a sentence embedding.

```python
import torch
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained('facebook/contriever')
model = AutoModel.from_pretrained('facebook/contriever')

sentences = [
    "Where was Marie Curie born?",
    "Maria Sklodowska, later known as Marie Curie, was born on November 7, 1867.",
    "Born in Paris on 15 May 1859, Pierre Curie was the son of Eugène Curie, a doctor of French Catholic origin from Alsace."
]

# Apply tokenizer
inputs = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
outputs = model(**inputs)

# Mean pooling
def mean_pooling(token_embeddings, mask):
    token_embeddings = token_embeddings.masked_fill(~mask[..., None].bool(), 0.)
    sentence_embeddings = token_embeddings.sum(dim=1) / mask.sum(dim=1)[..., None]
    return sentence_embeddings
embeddings = mean_pooling(outputs[0], inputs['attention_mask'])
```