Google Colab中UMAP导入失败求助：ImportError无法从umap导入UMAP

1. 安装包名称混淆 #

你安装的umap是无关第三方包，真正提供UMAP算法的是umap-learn。卸载错误包并安装正确版本：

!pip uninstall -y umap
!pip install umap-learn

2. 版本兼容性问题 #

BERTopic对UMAP版本有严格依赖，版本不匹配会导致导入失败。建议安装BERTopic官方推荐的兼容版本：

!pip install umap-learn==0.5.3

3. Colab预装依赖冲突 #

Colab默认预装部分库，可能与手动安装的包产生冲突。按以下顺序在干净运行时安装依赖：

• 重启Colab运行时
• 依次执行安装命令：

!pip install sentence_transformers
!pip install umap-learn
!pip install bertopic

4. 导入路径修正 #

部分UMAP版本需要从umap.umap_模块导入UMAP类，替换原有导入语句：

from umap.umap_ import UMAP

from bertopic import BERTopic
from sentence_transformers import SentenceTransformer
from umap.umap_ import UMAP

docs = abstracts
sentence_model = SentenceTransformer("neuml/pubmedbert-base-embeddings")
embeddings = sentence_model.encode(docs, show_progress_bar = False)

topic_model = BERTopic()
topics, probs = topic_model.fit_transform(docs, embeddings = embeddings)

topic_model.visualize_documents(docs, embeddings = embeddings)

内容的提问来源于stack exchange，提问作者konichiwow