bert base turkish cased mean nli stsb tr

提供商emrecan
分类natural-language-inference
许可证apache-2.0
下载量478.7K
星标0

简介

这是一个基于 BERT-base 架构并针对土耳其语深度优化的预训练模型。它在 NLI（自然语言推理）和 STSb（语义文本相似度）数据集上进行了微调，擅长处理文本之间的逻辑关系判断和相似度计算。对于需要处理土耳其语语料的开发者来说，它比通用多语言模型具有更强的本地语言理解能力。上手难度低，可直接集成到 Hugging Face 生态的 Pipeline 中，常用于构建土耳其语的智能问答、语义搜索或文本分类系统。

核心亮点

  • 专为土耳其语优化，语义理解比多语言模型更精准
  • 擅长 NLI 推理，可判断两段文本的蕴含或矛盾关系
  • 支持 STSb 任务，能够高效计算文本间的语义相似度
  • 兼容 Hugging Face 生态，部署简单且开发成本低

使用方法

安装依赖
# 安装 Hugging Face transformers
pip install transformers torch
SDK 使用
# 使用 transformers 加载模型
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("emrecan/bert-base-turkish-cased-mean-nli-stsb-tr")
tokenizer = AutoTokenizer.from_pretrained("emrecan/bert-base-turkish-cased-mean-nli-stsb-tr")

Hugging Face 下载

我们推荐使用命令行或者 Hugging Face Hub SDK 来进行模型的下载。

操作指引：在下载前，请先通过如下命令安装 huggingface_hub：

操作指引 
pip install -U huggingface_hub

命令行下载

下载完整模型库

下载完整模型库 
huggingface-cli download emrecan/bert-base-turkish-cased-mean-nli-stsb-tr

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例）

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例） 
huggingface-cli download emrecan/bert-base-turkish-cased-mean-nli-stsb-tr config.json --local-dir ./dir

更多命令行下载选项，可参见官方文档

SDK 下载

SDK 下载 
# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('emrecan/bert-base-turkish-cased-mean-nli-stsb-tr')

Git 下载

请确保 lfs 已经被正确安装

Git 下载 
git lfs install
git clone https://huggingface.co/emrecan/bert-base-turkish-cased-mean-nli-stsb-tr

如果您希望跳过 lfs 大文件下载，可以使用如下命令

跳过 LFS 
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/emrecan/bert-base-turkish-cased-mean-nli-stsb-tr

模型文件托管在 Hugging Face Hub，使用 HF CLI / SDK / Git 直接下载，不经过本站。

PyTorch / Transformers 使用

安装 Transformers

安装 Transformers 
pip install -U transformers torch

模型加载和推理

模型加载和推理 
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('emrecan/bert-base-turkish-cased-mean-nli-stsb-tr')
tokenizer = AutoTokenizer.from_pretrained('emrecan/bert-base-turkish-cased-mean-nli-stsb-tr')

完整文档

来源: HuggingFace

---
language:

  • tr

pipeline_tag: sentence-similarity
license: apache-2.0
tags:
  • sentence-transformers

  • feature-extraction

  • sentence-similarity

  • transformers

datasets:
  • nli_tr

  • emrecan/stsb-mt-turkish

widget:
source_sentence: "Bu çok mutlu bir kişi"
sentences:
- "Bu mutlu bir köpek"
- "Bu sevincinden havalara uçan bir insan"
- "Çok kar yağıyor"

---

emrecan/bert-base-turkish-cased-mean-nli-stsb-tr

This is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search. The model was trained on Turkish machine translated versions of NLI and STS-b datasets, using example training scripts from sentence-transformers GitHub repository.

Usage (Sentence-Transformers)

Using this model becomes easy when you have sentence-transformers installed:

code
pip install -U sentence-transformers

Then you can use the model like this:

python
from sentence_transformers import SentenceTransformer
sentences = ["Bu örnek bir cümle", "Her cümle vektöre çevriliyor"]

model = SentenceTransformer('emrecan/bert-base-turkish-cased-mean-nli-stsb-tr')
embeddings = model.encode(sentences)
print(embeddings)

Usage (HuggingFace Transformers)

Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.
python
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 = ["Bu örnek bir cümle", "Her cümle vektöre çevriliyor"]

Load model from HuggingFace Hub

tokenizer = AutoTokenizer.from_pretrained('emrecan/bert-base-turkish-cased-mean-nli-stsb-tr')
model = AutoModel.from_pretrained('emrecan/bert-base-turkish-cased-mean-nli-stsb-tr')

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)

Evaluation Results

Evaluation results on test and development sets are given below:

| Split | Epoch | cosine_pearson | cosine_spearman | euclidean_pearson | euclidean_spearman | manhattan_pearson | manhattan_spearman | dot_pearson | dot_spearman |
|------------|-------|----------------|-----------------|-------------------|--------------------|-------------------|--------------------|-------------|--------------|
| test | - | 0.834 | 0.830 | 0.820 | 0.819 | 0.819 | 0.818 | 0.799 | 0.789 |
| validation | 1 | 0.850 | 0.848 | 0.831 | 0.835 | 0.83 | 0.83 | 0.80 | 0.806 |
| validation | 2 | 0.857 | 0.857 | 0.844 | 0.848 | 0.844 | 0.848 | 0.813 | 0.810 |
| validation | 3 | 0.860 | 0.859 | 0.846 | 0.851 | 0.846 | 0.850 | 0.825 | 0.822 |
| validation | 4 | 0.859 | 0.860 | 0.846 | 0.851 | 0.846 | 0.851 | 0.825 | 0.823 |

Training

Training scripts training_nli_v2.py and training_stsbenchmark_continue_training.py were used to train the model.

The model was trained with the parameters:

DataLoader:

torch.utils.data.dataloader.DataLoader of length 360 with parameters:

code
{'batch_size': 16, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

Loss**:

sentence_transformers.losses.CosineSimilarityLoss.CosineSimilarityLoss

Parameters of the fit()-Method:

code
{
    "epochs": 4,
    "evaluation_steps": 200,
    "evaluator": "sentence_transformers.evaluation.EmbeddingSimilarityEvaluator.EmbeddingSimilarityEvaluator",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'transformers.optimization.AdamW'>",
    "optimizer_params": {
        "lr": 2e-05
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": null,
    "warmup_steps": 144,
    "weight_decay": 0.01
}

Full Model Architecture

code
SentenceTransformer(
  (0): Transformer({'max_seq_length': 75, '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})
)

Citing & Authors

<!--- Describe where people can find more information -->

查看官方来源