如何使用BERT对相似句子进行聚类

from transformers import BertTokenizer, BertModel
from sklearn.cluster import KMeans
import torch
import numpy as np

# 加载BERT模型和分词器
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
model = BertModel.from_pretrained('bert-base-uncased')

# 示例句子列表
sentences = ["Hello, how are you?", "Hi, how are you doing?", "Good morning!", "Have a nice day!"]

# 对句子进行编码
encoded_inputs = tokenizer(sentences, return_tensors='pt', padding=True, truncation=True)
with torch.no_grad():
    outputs = model(**encoded_inputs)
sentence_embeddings = outputs.last_hidden_state[:, 0, :].numpy()

# 使用K-means进行聚类
kmeans = KMeans(n_clusters=2)
clusters = kmeans.fit_predict(sentence_embeddings)

# 输出聚类结果
for sentence, cluster in zip(sentences, clusters):
    print(f"Sentence: '{sentence}' - Cluster: {cluster}")