🚀 Sambert - 希伯来语嵌入模型

Sambert 是一个基于 sentence-transformers 的模型，它能够将句子和段落映射到 768 维的密集向量空间，可用于聚类或语义搜索等任务。

🚀 快速开始

安装依赖

若要使用该模型，需要安装 sentence-transformers：

pip install -U sentence-transformers

使用示例

基础用法

使用 sentence-transformers 库调用模型：

from sentence_transformers import SentenceTransformer, util
sentences = ["אמא הלכה לגן", "אבא הלך לגן", "ירקוני קונה לנו פיצות"]

model = SentenceTransformer('MPA/sambert')
embeddings = model.encode(sentences)
print(util.cos_sim(embeddings, embeddings))

高级用法

若不使用 sentence-transformers 库，可按以下步骤操作：先将输入数据传入 Transformer 模型，再对上下文词嵌入应用正确的池化操作。

from transformers import AutoTokenizer, AutoModel
import torch


#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] #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)


# Sentences we want sentence embeddings for
sentences = ["אמא הלכה לגן", "אבא הלך לגן", "ירקוני קונה לנו פיצות"]

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('MPA/sambert')
model = AutoModel.from_pretrained('MPA/sambert')

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

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

# Perform pooling. In this case, mean pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

print("Sentence embeddings:")
print(sentence_embeddings)

📚 详细文档

评估结果

若要对该模型进行自动评估，可参考 Sentence Embeddings Benchmark：https://seb.sbert.net。

训练过程

该模型分两个阶段进行训练：

1. 无监督学习：使用约 200 万个段落，在 cls 标记上采用 'MultipleNegativesRankingLoss' 损失函数。
2. 有监督学习：使用约 7 万个段落，采用 'CosineSimilarityLoss' 损失函数。

模型训练的参数如下：

数据加载器

torch.utils.data.dataloader.DataLoader，长度为 11672，参数如下：

{'batch_size': 4, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

损失函数

sentence_transformers.losses.CosineSimilarityLoss.CosineSimilarityLoss

fit() 方法的参数：

{
    "epochs": 1,
    "evaluation_steps": 1000,
    "evaluator": "sentence_transformers.evaluation.EmbeddingSimilarityEvaluator.EmbeddingSimilarityEvaluator",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
    "optimizer_params": {
        "lr": 2e-05
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": null,
    "warmup_steps": 500,
    "weight_decay": 0.01
}

完整模型架构

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False})
)

引用与作者

本模型基于以下论文：

@misc{gueta2022large,
      title={Large Pre-Trained Models with Extra-Large Vocabularies: A Contrastive Analysis of Hebrew BERT Models and a New One to Outperform Them All}, 
      author={Eylon Gueta and Avi Shmidman and Shaltiel Shmidman and Cheyn Shmuel Shmidman and Joshua Guedalia and Moshe Koppel and Dan Bareket and Amit Seker and Reut Tsarfaty},
      year={2022},
      eprint={2211.15199},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}