Python常见报错总结_typeerror: np.matrix is not supported. please conv

持续更新中......

TypeError: np.matrix is not supported. Please convert to a numpy array with np.asarray

解决方案：https://github.com/XingyanLiu/CAME/issues/20

numpy                     1.23.2                   
scikit-learn              1.1.2 
scipy                     1.9.1  

使用CUDA进行模型训练时出现错误RuntimeError: No CUDA GPUs are available

解决方案：1.先检测是否有GPU

print(torch.cuda.is_available())

返回0则是pytorch CUDA安装的问题

2.如果返回1，注意代码中是否有“os.environ”，“CUDA_VISIBLE_DEVICE”，是否是代码设置了显卡编号，默认编号是0

AttributeError: module 'numpy' has no attribute 'float'

解决方案：从numpy1.24起，删除了numpy.bool、numpy.int、numpy.fioat、numpy.complex, numpy.object, numpy.str、numpy.long、numpy.unicode类型的支持

pip uninstall numpy
pip install numpy==1.23.5

找不到指定的模块fbgemm.dll

pytorch｜找不到 fbgemm.dll 问题处理-CSDN博客

OSError: [WinError 126] 找不到指定的模块。 Error loading “E:\python310\lib\site-packages\torch\lib\fbgemm.dll-CSDN博客

解决 python出现的[WinError 126] 找不到指定的模块。fbgemm.dll“ or one of its dependencies.-CSDN博客 torch fbgemm.dll 报错-CSDN博客

Failed to import pytorch fbgemm.dll or one of its dependencies is missing - #27 by traits - windows - PyTorch Forums

 从零开始搭建LLaMA-Factory_libomp140-CSDN博客

This is still unresolved. I have pip installations with
 
torch-2.4.0+cu121
VC_redist.x64.exe (14.40.33810.0)
NVIDIA drivers 560.70
 
on two machines
 
Win10:
 
with additional full Visual Studio 2022 installation
dedicated RTX 3070
Win11
 
no Visual Studio
eGPU GTX 1060
The Win10 version is working, the Win11 version not.
 
edit: Win10 has libomp140.x86_64.dll in c:/Windows/System32 (and a debug version elsewhere for VS 2022), the Win11 version has nothing on the whole drive. So, this dll seems to be no longer a part of vcredist. I loosely remember some recent problems in downloading OpenMP from Microsoft, but cannot find the associated defunctional link.

 这个环境看出win11的libomp140跟win10不太一样，我的环境也是win11，不知道是不是这个原因

 libomp140.x86_64.dll文件不知道是不是网络的问题太难下载，放在顶部资源

把libomp140.x86_64.dll放在C:\Windows\SysWOW64下，import torch不在报错！！