已加载运行时 CuDNN 库:5005(兼容版本 5000)但源代码是用 5103(兼容版本 5100)编译的

问题描述

我有以下错误.我正在使用 tensorflow 的 conda 安装.我正在努力尝试将它与我的 GPU 一起使用.

I've the following error. I'm using a conda installation of tensorflow. I'm struggling to try to use it with my GPU.

加载的运行时 CuDNN 库:5005(兼容版本 5000)但源代码是用 5103(兼容版本 5100)编译的.如果使用二进制安装，请升级您的 CuDNN 库以匹配.如果从源代码构建，请确保在运行时加载的库与编译配置期间指定的兼容版本相匹配.F tensorflow/core/kernels/conv_ops.cc:526] 检查失败:stream->parent()->GetConvolveAlgorithms(&algorithms)中止(核心转储)

哪个 nvcc 返回/usr/local/cuda-7.5/bin/nvcc

nvcc 版本返回Cuda 编译工具，release 7.5，V7.5.17

我尝试下载 CuDNN v5.1 并执行以下操作，但也不起作用``须藤cp lib*/usr/local/cuda-7.5/lib64/须藤cp include/cudnn.h/usr/local/cuda-7.5/include/须藤ldconfig

I tried downloading CuDNN v5.1 and did the following but it didn't work either```sudo cp lib* /usr/local/cuda-7.5/lib64/sudo cp include/cudnn.h /usr/local/cuda-7.5/include/sudo ldconfig

```

我也试过另一个文件夹须藤cp lib*/usr/local/cuda/lib64/须藤cp include/cudnn.h/usr/local/cuda/include/须藤ldconfig

推荐答案

这里有一个很好的解释 - 错误是什么:`加载的运行时CuDNN库:5005但源代码是用5103编译的`是什么意思?

There's a good explanation of what it means here - What does the error: `Loaded runtime CuDNN library: 5005 but source was compiled with 5103` mean?

简短的回答是你有 CuDNN 5.0 但你应该安装 CuDNN 5.1

Short answer is that you have CuDNN 5.0 but you should install CuDNN 5.1

看起来这就是您正在尝试做的事情.我只需按照此处的说明进行操作即可 - https://www.tensorflow.org/get_started/os_setup#optional_install_cuda_gpus_on_linux

It looks like that's what you are attempting to do. It worked for me to just follow the the instructions here - https://www.tensorflow.org/get_started/os_setup#optional_install_cuda_gpus_on_linux

在这样做之前，我检查了 /usr/local/cuda/include/cudnn.h 的内容，确实它在顶部有这些行，表明它是 5.0.5 版

Before doing this, I checked the contents of /usr/local/cuda/include/cudnn.h and indeed it had these lines towards the top indicating it was version 5.0.5

#define CUDNN_MAJOR      5
#define CUDNN_MINOR      0
#define CUDNN_PATCHLEVEL 5

如果您的 /usr/local/cuda/include/cudnn.h 已经是 5.1，那么在另一个目录中还有另一个 CuDNN 正在被引用.我的 .bashrc 中有以下内容 - 也许尝试添加它或检查 Tensorflow 说明添加什么.

If your /usr/local/cuda/include/cudnn.h is already 5.1, then there is another CuDNN in another directory that is getting referenced. I have the following in my .bashrc -- perhaps try adding this or checking the Tensorflow instructions for what to add.

export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"
export CUDA_HOME="/usr/local/cuda"

这篇关于已加载运行时 CuDNN 库:5005(兼容版本 5000)但源代码是用 5103(兼容版本 5100)编译的的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！