关于Import pykeops时出现warnning No cuda detect

最新推荐文章于 2024-07-17 21:52:34 发布

isunLt

最新推荐文章于 2024-07-17 21:52:34 发布

阅读量942

点赞数 2

分类专栏：环境配置文章标签： linux cuda pykeops

本文链接：https://blog.csdn.net/isunlt/article/details/115269143

版权

环境配置专栏收录该内容

5 篇文章 0 订阅

订阅专栏

本文探讨了在Python库PyKeOps中遇到的CUDA找不到问题，通过查找并修复libcuda.so路径，提供了两种可能的解决方案：在libnames中添加特定版本或在LD_LIBRARY_PATH中设置。

摘要由CSDN通过智能技术生成

具体问题在pykeops.common.gpu_utils的函数get_gpu_number()中

    libnames = ("libcuda.so", "libcuda.dylib", "cuda.dll")
    for libname in libnames:
        try:
            cuda = ctypes.CDLL(libname)
        except OSError:
            continue
        else:
            break
    else:
        print("[pyKeOps]: Warning, no cuda detected. Switching to cpu only.")
        return 0  # raise

    nGpus = ctypes.c_int()
    error_str = ctypes.c_char_p()

    result = cuda.cuInit(0)

ctypes.CDLL(libname)找不到libcuda.so，使用命令locate libcuda.so得到结果

/usr/lib/x86_64-linux-gnu/libcuda.so.1
/usr/lib/x86_64-linux-gnu/libcuda.so.450.51.06
/usr/lib/x86_64-linux-gnu/stubs/libcuda.so
/usr/local/cuda-10.1/doc/man/man7/libcuda.so.7
/usr/local/cuda-10.1/targets/x86_64-linux/lib/stubs/libcuda.so
/usr/local/cuda-10.2/doc/man/man7/libcuda.so.7
/usr/local/cuda-10.2/targets/x86_64-linux/lib/stubs/libcuda.so
/usr/local/cuda-11.0/doc/man/man7/libcuda.so.7
/usr/local/cuda-11.0/targets/x86_64-linux/lib/stubs/libcuda.so
/usr/local/cuda-9.0/doc/man/man7/libcuda.so.7
/usr/local/cuda-9.0/lib64/stubs/libcuda.so
/usr/share/man/man7/libcuda.so.7
/usr/share/man/man7/libcuda.so.7.gz

根据nvidia-smi显示驱动的版本是450.51.06所以正确的cuda应该是/usr/lib/x86_64-linux-gnu目录下的3项，那么修改的方法有两种：
1. 在libnames里添加libcuda.so.1

# libnames = ("libcuda.so", "libcuda.dylib", "cuda.dll")
libnames = ("libcuda.so.1", "libcuda.so", "libcuda.dylib", "cuda.dll")

2. 在LD_LIBRARY_PATH里添加/usr/lib/x86_64-linux-gnu/stubs

vim ~/.profile
export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu/stubs${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
:wq