已安装sentence-transformers却无法导入，torch依赖fbgemm.dll报错求助

已通过pip list确认sentence-transformers和torch均已安装且为最新版本，但运行生成SBERT嵌入的Python脚本时，始终出现无法加载fbgemm.dll或其依赖项的错误。尝试过全局环境、conda/venv虚拟环境，更换Python和包版本组合，卸载重装torch等操作，问题仍未解决。

运行脚本 #

import numpy as np
import pandas as pd
from sentence_transformers import SentenceTransformer

# Load the SBERT model
model = SentenceTransformer('all-mpnet-base-v2')

# Load the dataset
input_file = 'PJ_whole_conversations_row_per_conversation_v1.csv'
df = pd.read_csv(input_file)

# Generate embeddings for the "message" column
messages = df['message'].tolist()
embeddings = model.encode(messages, convert_to_numpy=True)

# Save embeddings to NumPy file
output_file_numpy = 'row_per_conversation_embeddings.npy'
np.save(output_file_numpy, embeddings)

print(f'Embeddings have been generated and saved to {output_file_numpy}')

错误信息 #

Exception has occurred: OSError
[WinError 126] The specified module could not be found. Error loading "C:\Users\nea\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.11_qbz5n2kfra8p0\LocalCache\local-packages\Python311\site-packages\torch\lib\fbgemm.dll" or one of its dependencies.
  File "C:\Users\saham\OneDrive\4. Semester\Independent Research Project\Python_scripts\row_per_conversation_word embedding.py", line 3, in <module>
    from sentence_transformers import SentenceTransformer
OSError: [WinError 126] The specified module could not be found. Error loading "C:\Users\nea\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.11_qbz5n2kfra8p0\LocalCache\local-packages\Python311\site-packages\torch\lib\fbgemm.dll" or one of its dependencies.

• 安装Microsoft Visual C++ Redistributable运行库：fbgemm.dll依赖VC组件，下载安装对应x64版本的VC Redistributable（2019或2022版本），重启系统后重试。
• 安装纯CPU版本torch：若没有NVIDIA GPU，卸载现有torch，执行pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu安装CPU专属版本，规避GPU相关依赖缺失问题。
• 降级Python版本：Python3.11与部分torch版本存在兼容性问题，尝试切换到Python3.10，重新创建虚拟环境并安装所有依赖。
• 排查并补全依赖项：使用Dependency Walker工具扫描fbgemm.dll，查看缺失的系统依赖dll，下载匹配x64架构的对应文件放入系统目录或torch的lib文件夹。
• 强制重装依赖：删除torch的安装目录，执行pip install --force-reinstall torch sentence-transformers清理残留文件后重新安装。

内容的提问来源于stack exchange，提问作者Nea13