CUDA安装

问题

PaddlePaddle GPU安装后，检查出错

在研究PaddlePaddle GPU版本安装的过程中，在解释器中运行

>>> import paddle
>>> paddle.utils.run_check()

出现了错误，表明CUDA、CUDNN未安装成功，下面则是介绍如何安装CUDA、CUDNN

Running verify PaddlePaddle program ...
W1126 15:44:35.675730   540 device_context.cc:447] Please NOTE: device: 0, GPU Compute Capability: 6.1, Driver API Version: 11.0, Runtime API Version: 10.2
W1126 15:44:35.676738   540 dynamic_loader.cc:258] Note: [Recommend] copy cudnn into CUDA installation directory.
 For instance, download cudnn-10.0-windows10-x64-v7.6.5.32.zip from NVIDIA's official website,
then, unzip it and copy it into C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0
You should do this according to your CUDA installation directory and CUDNN version.
Traceback (most recent call last):
  File "", line 1, in <module>
  File "C:\ProgramData\Miniconda3\lib\site-packages\paddle\utils\install_check.py", line 196, in run_check
    _run_static_single(use_cuda)
  File "C:\ProgramData\Miniconda3\lib\site-packages\paddle\utils\install_check.py", line 124, in _run_static_single
    exe.run(startup_prog)
  File "C:\ProgramData\Miniconda3\lib\site-packages\paddle\fluid\executor.py", line 1246, in run
    six.reraise(*sys.exc_info())
  File "C:\ProgramData\Miniconda3\lib\site-packages\six.py", line 703, in reraise
    raise value
  File "C:\ProgramData\Miniconda3\lib\site-packages\paddle\fluid\executor.py", line 1234, in run
    return self._run_impl(
  File "C:\ProgramData\Miniconda3\lib\site-packages\paddle\fluid\executor.py", line 1366, in _run_impl
    return self._run_program(
  File "C:\ProgramData\Miniconda3\lib\site-packages\paddle\fluid\executor.py", line 1463, in _run_program
    self._default_executor.run(program.desc, scope, 0, True, True,
RuntimeError: (PreconditionNotMet) The third-party dynamic library (cudnn64_7.dll) that Paddle depends on is not configured correctly. (error code is 126)
  Suggestions:
  1. Check if the third-party dynamic library (e.g. CUDA, CUDNN) is installed correctly and its version is matched with paddlepaddle you installed.
  2. Configure third-party dynamic library environment variables as follows:
  - Linux: set LD_LIBRARY_PATH by `export LD_LIBRARY_PATH=...`
  - Windows: set PATH by `set PATH=XXX; (at ..\paddle\fluid\platform\dynload\dynamic_loader.cc:285)

安装CUDA

下载CUDA

连接 https://developer.nvidia.com/cuda-downloads

下载历史版本

链接 https://developer.nvidia.com/cuda-toolkit-archive

下载

我的电脑安装的是10.2版本的，

具体安装哪个版本要看你的显卡型号，可在设备管理器中查看

可以在NVIDIA的控制面板中查看CUDA最大安装版本

下载截图如下：

下载后一路安装即可。

检查CUDA是否安装成功

在终端中输入nvcc --version查看CUDA版本，如下，则证明安装成功

安装CUDNN

cudnn安装需要会员，会员比较好注册，需要一个邮箱，其他自行填写就可。

登陆后，进入下载页 https://developer.nvidia.com/rdp/cudnn-download

安装哪个版本

这个需要看PaddlePaddle 的官网，需要你安装哪个版本，如果版本错了，那么后面会报错，我就是在这个地方卡住，而且不知道为什么会出错，原来是没有按照文档安装相应的版本。
PaddlePaddle 的官网版本的链接：https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/install/pip/windows-pip.html
此处标记了安装哪个版本的要求

我的CUDA是10.2的，那么需要安装cuDNN7。那么找到cuDNN7

下载cuDNN这是个压缩包，解压后，复制所有文件，到CUDA的安装目录下，bin目录下的文件复制到CUDA的bin，lib到lib，include到include

我的CUDA的安装目录是CUDA的默认目录：
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2

检查是否成功

在您的Python解释器中，输入

>>> import paddle
>>> paddle.utils.run_check()

如果结果如下，那么则安装成功了：

Running verify PaddlePaddle program ...
W1126 16:33:38.690778  3240 device_context.cc:447] Please NOTE: device: 0, GPU Compute Capability: 6.1, Driver API Version: 11.0, Runtime API Version: 10.2
W1126 16:33:38.747905  3240 device_context.cc:465] device: 0, cuDNN Version: 7.6.
PaddlePaddle works well on 1 GPU.
PaddlePaddle works well on 1 GPUs.
PaddlePaddle is installed successfully! Let's start deep learning with PaddlePaddle now.