🚀 robbert-2022-dutch-sentence-transformers - Onnx

本项目提供了一个用于荷兰语句子相似度计算的Onnx模型，它基于robbert-2022-dutch-sentence-transformers模型转换而来，可用于聚类、语义搜索等任务。

🚀 快速开始

模型信息

数据集

该模型在以下荷兰语翻译数据集上进行训练：

• NetherlandsForensicInstitute/AllNLI-translated-nl
• NetherlandsForensicInstitute/altlex-translated-nl
• NetherlandsForensicInstitute/coco-captions-translated-nl
• NetherlandsForensicInstitute/flickr30k-captions-translated-nl
• NetherlandsForensicInstitute/msmarco-translated-nl
• NetherlandsForensicInstitute/quora-duplicates-translated-nl
• NetherlandsForensicInstitute/sentence-compression-translated-nl
• NetherlandsForensicInstitute/simplewiki-translated-nl
• NetherlandsForensicInstitute/stackexchange-duplicate-questions-translated-nl
• NetherlandsForensicInstitute/wiki-atomic-edits-translated-nl

模型描述

这个Onnx模型是robbert-2022-dutch-sentence-transformers的转换版本，使用了 这里 找到的transformers.js脚本进行转换。

示例展示

• 示例标题：荷兰语
  • 源句子：Deze week ga ik naar de kapper
  • 对比句子：
    • Ik ga binnenkort mijn haren laten knippen
    • Morgen wil ik uitslapen
    • Gisteren ging ik naar de bioscoop

✨ 主要特性

这是一个 sentence-transformers 模型，它可以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。该模型基于 KU Leuven的RobBERT模型，并在 Paraphrase数据集 上进行了微调，该数据集已被机器翻译成荷兰语。

📦 安装指南

若要使用此模型，你需要安装 sentence-transformers：

pip install -U sentence-transformers

💻 使用示例

基础用法

使用 sentence-transformers 库的简单示例：

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

model = SentenceTransformer('NetherlandsForensicInstitute/robbert-2022-dutch-sentence-transformers')
embeddings = model.encode(sentences)
print(embeddings)

高级用法

不使用 sentence-transformers 库，手动处理输入并进行池化操作的示例：

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('NetherlandsForensicInstitute/robbert-2022-dutch-sentence-transformers')
model = AutoModel.from_pretrained('NetherlandsForensicInstitute/robbert-2022-dutch-sentence-transformers')

# 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)

🔧 技术细节

训练参数

模型训练时使用的参数如下：

数据加载器

MultiDatasetDataLoader.MultiDatasetDataLoader，长度为414262，参数如下：

{'batch_size': 1}

损失函数

sentence_transformers.losses.MultipleNegativesRankingLoss.MultipleNegativesRankingLoss，参数如下：

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

训练方法参数

{
    "epochs": 1,
    "evaluation_steps": 50000,
    "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': 128, 'do_lower_case': False}) with Transformer model: RobertaModel 
  (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})
)

📚 详细文档

引用与作者

如果你使用了该模型，请参考相关文档进行引用。关于更多信息，可在模型创建者的页面 Netherlands Forensic Institute 中查找。