Skip to content

Latest commit

 

History

History
53 lines (53 loc) · 2.92 KB

Report-of-cuDNN-Installation.md

File metadata and controls

53 lines (53 loc) · 2.92 KB

Errors and instructions of openpose installation

1. CUDA8.0安装

参考:http://blog.csdn.net/zhyj3038/article/details/52494892

  • NVIDIA官网下载.deb安装,.run安装有下面问题:
    (1) Error: X server is running; 解决:tty下$ init 3命令可以kill X server再进行安装
    (2) 各种关闭X server尝试过后继续Error: Reboot required to continue;重启后什么也没发生......
    • 待研究...
  • NVIDIA官网下载安装.deb: 成功
  • 校验测试是否安装成功
    (1) 编译samples时目录与教程稍有不同,注意
    (2) 个别sample编译出错,没深究,大部分过了:(
    (3) 不死心,重新搜了检查CUDA是否安装成功 参考:http://blog.csdn.net/u010480194/article/details/54287335
    (4) Result=pass

2. cuDNN安装

参考:http://blog.csdn.net/zhyj3038/article/details/52494892

  • 顺利

3. caffe安装

参考:http://www.linuxidc.com/Linux/2016-09/135034.htm

  • 最后一步import caffe 出错 ImportError: No module named google.protobuf.internal; 解决:依赖库不够
$ sudo apt-get install python-numpy python-scipy python-matplotlib python-sklearn python-skimage python-h5py python-protobuf python-leveldb python-networkx python-nose python-pandas python-gflags Cython ipython  
$ sudo apt-get update

4. openpose安装

参考:https://github.com/CMU-Perceptual-Computing-Lab/openpose

  • caffeopenpose可以同时安装,或单独安装,见参考安装说明
  • 运行demo 出错 
Starting pose estimation demo.
Starting thread(s)
F0508 10:39:36.013202  6215 cudnn_conv_layer.cpp:53] Check failed: status == CUDNN_STATUS_SUCCESS (6 vs. 0)  CUDNN_STATUS_ARCH_MISMATCH
*** Check failure stack trace: ***
    @     0x7f24cadd45cd  google::LogMessage::Fail()
    @     0x7f24cadd6433  google::LogMessage::SendToLog()
    @     0x7f24cadd415b  google::LogMessage::Flush()
    @     0x7f24cadd6e1e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7f24c8eeef73  caffe::CuDNNConvolutionLayer<>::LayerSetUp()
    @     0x7f24c9051b83  caffe::Net<>::Init()
    @     0x7f24c90549b0  caffe::Net<>::Net()
    @     0x7f24cb0927ae  op::NetCaffe::initializationOnThread()
    @     0x7f24cb0624ef  op::PoseExtractorCaffe::netInitializationOnThread()
    @     0x7f24cb10dd31  op::SubThread<>::initializationOnThread()
    @     0x7f24cb10e320  op::Thread<>::initializationOnThread()
    @     0x7f24cb11318d  op::Thread<>::threadFunction()
    @     0x7f24ca8dec80  (unknown)
    @     0x7f24ca3fa6ba  start_thread
    @     0x7f24ca13082d  clone
    @              (nil)  (unknown)
已放弃 (核心已转储)
  • 搜了一圈,简单讲:就是GPU的加速性能不够,CUDNN只支持CUDA Capability 3.0以上的GPU加速;笔记本上的GT520M的CUDA Capability是2.1,而官方的cudnn加速是不支持3.0以下的版本的.
  • 解决:换电脑!!!最后还是败给了穷.

总结:年轻人不要用这么老的本!