Serafim 100m Portuguese Pt Sentence Encoder

🚀 塞拉菲姆100m葡萄牙語（PT）句子編碼器

塞拉菲姆100m葡萄牙語（PT）句子編碼器是一個基於 sentence-transformers 的模型，它可以將句子和段落映射到一個768維的密集向量空間，可用於聚類或語義搜索等任務。

🚀 快速開始

📦 安裝指南

若要使用此模型，你需要安裝 sentence-transformers：

pip install -U sentence-transformers

💻 使用示例

基礎用法

使用 sentence-transformers 庫調用模型：

from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]

model = SentenceTransformer('PORTULAN/serafim-100m-portuguese-pt-sentence-encoder')
embeddings = model.encode(sentences)
print(embeddings)

高級用法

在不使用 sentence-transformers 庫的情況下，你可以通過以下方式使用該模型：首先，將輸入數據傳入Transformer模型，然後對上下文詞嵌入應用正確的池化操作。

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 = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('PORTULAN/serafim-100m-portuguese-pt-sentence-encoder')
model = AutoModel.from_pretrained('PORTULAN/serafim-100m-portuguese-pt-sentence-encoder')

# 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)

📚 詳細文檔

評估結果

若要對該模型進行自動評估，請參考 Sentence Embeddings Benchmark：https://seb.sbert.net

訓練詳情

該模型的訓練參數如下：

數據加載器

torch.utils.data.dataloader.DataLoader，長度為296，參數如下：

{'batch_size': 64, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

損失函數

sentence_transformers.losses.CoSENTLoss.CoSENTLoss，參數如下：

{'scale': 20.0, 'similarity_fct': 'pairwise_cos_sim'}

訓練方法參數

{
    "epochs": 20,
    "evaluation_steps": 30,
    "evaluator": "sentence_transformers.evaluation.EmbeddingSimilarityEvaluator.EmbeddingSimilarityEvaluator",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
    "optimizer_params": {
        "lr": 1e-05
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": 296,
    "warmup_steps": 592,
    "weight_decay": 0.01
}

完整模型架構

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, '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, 'include_prompt': True})
)

引用與作者

該文章已在EPIA 2024會議上發表，並由Springer出版：

@InProceedings{epia2024serafim,
    title={Open Sentence Embeddings for Portuguese with the Serafim PT* encoders family}, 
    author={Luís Gomes and António Branco and João Silva and João Rodrigues and Rodrigo Santos},
    editor={Manuel Filipe Santos and José Machado and Paulo Novais and Paulo Cortez and Pedro Miguel Moreira},
    booktitle={Progress in Artificial Intelligence},
    doi={doi.org/10.1007/978-3-031-73503-5_22},
    year={2024},
    publisher={Springer Nature Switzerland},
    address={Cham},
    pages={267--279},
    isbn={978-3-031-73503-5}
}

在Springer出版之前，預印本可在arXiv上獲取：

@misc{gomes2024opensentenceembeddingsportuguese,
    title={Open Sentence Embeddings for Portuguese with the Serafim PT* encoders family}, 
    author={Luís Gomes and António Branco and João Silva and João Rodrigues and Rodrigo Santos},
    year={2024},
    eprint={2407.19527},
    archivePrefix={arXiv},
    primaryClass={cs.CL},
    url={https://arxiv.org/abs/2407.19527}, 
}

📄 許可證

本項目採用MIT許可證。