1、下载模型报错ConnectionError: (ProtocolError(‘Connection aborted.’, ConnectionResetError(54, ‘Connection reset by peer’))
运行代码:
topics, probabilities = model.fit_transform(docs)
报错内容:
ConnectionError: (ProtocolError(‘Connection aborted.’, ConnectionResetError(54, ‘Connection reset by peer’)), ‘(Request ID: 5a75855f-ec0f-415b-8626-cb998e73aa80)’)
解决方案:
安装OpenSSL
MAC 安装
brew install OpenSSL
2、下载模型报错ConnectionError: (MaxRetryError("HTTPSConnectionPool(host=‘huggingface.co’, port=443):
运行代码:
topics, probabilities = model.fit_transform(docs)
报错内容:
ConnectionError: (MaxRetryError("HTTPSConnectionPool(host=‘huggingface.co’, port=443)
解决方案:
- 将模型下载到本地,在huggingface 上,指定模型的Files and versions 进行下载,全部下载完成以后,设置本地模型的路径,就是下载以后放在本地的文件夹地址。
- 科学上网,切换网络后可以直接下载
3、OMP: Info #276: omp_set_nested routine deprecated, please use omp_set_max_active_levels instead."
运行代码:
topics, probabilities = model.fit_transform(docs)
报错内容:P: Info #276: omp_set_nested routine deprecated, please use omp_set_max_active_levels instead."
解决方案
conda install -c numba numba==0.56.2
4、ValueError: empty vocabulary; perhaps the documents only contain stop words
5、TypeError: object of type ‘numpy.float64’ has no len()
遇到上面两个(4 & 5)问题基本上都是因为,默认的模型是处理英文数据,即处理word+空格的语言,对中文不适用,针对问题5:网上有个解决方案,但是我自己没有解决
最终的解决方案是不用默认的模型,换成支持中文的模型
from sentence_transformers import SentenceTransformer
embedding_model = SentenceTransformer("distiluse-base-multilingual-cased-v1")