欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页

could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED

程序员文章站 2022-05-26 23:44:15
...

今天我在运行keras版本的cnn rnn classifier的时候,报了如下的错误:

totalMemory: 10.91GiB freeMemory: 10.20GiB
2018-05-08 21:27:29.962822: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1435] Adding visible gpu devices: 0
2018-05-08 21:27:30.176092: I tensorflow/core/common_runtime/gpu/gpu_device.cc:923] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-05-08 21:27:30.176127: I tensorflow/core/common_runtime/gpu/gpu_device.cc:929]      0 
2018-05-08 21:27:30.176135: I tensorflow/core/common_runtime/gpu/gpu_device.cc:942] 0:   N 
2018-05-08 21:27:30.176364: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1053] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 9868 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1080 Ti, pci bus id: 0000:03:00.0, compute capability: 6.1)
2018-05-08 21:27:32.141585: E tensorflow/stream_executor/cuda/cuda_dnn.cc:455] could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED
2018-05-08 21:27:32.141664: E tensorflow/stream_executor/cuda/cuda_dnn.cc:463] possibly insufficient driver version: 384.130.0
2018-05-08 21:27:32.141682: F tensorflow/core/kernels/conv_ops.cc:713] Check failed: stream->parent()->GetConvolveAlgorithms( conv_parameters.ShouldIncludeWinogradNonfusedAlgo<T>(), &algorithms) 
Aborted (core dumped)

百思不得其解,查了很多资料都行不通,我的环境是cuda 9.0,ubuntu 16.04,然后我看到cudnn错误的地方好像跟cudnn有关,我的cuda是9.0,于是下载了一个cudnn 7.0,

tar zxvf cudnn-9.0-linux-x64-v7.tgz
sudo cp cuda/include/cudnn.h /usr/local/cuda/include
sudo cp cuda/lib64/* /usr/local/cuda/lib64

然后发现程序能够正常的运行了,我分析这可能是我在更新cuda后,忘记把cudnn也更新到7.0引起的,我这里记录下来,希望帮助到大家。


参考文献

[1].cuDNN.https://blog.csdn.net/guoyunfei20/article/details/78191599