tensorflow 1.4 could not create cudnn handle: CUDNN_STATUS_INTERNAL_ERROR 錯誤的解決辦法
阿新 • • 發佈:2019-01-27
tensorflow 1.4 cuda8.0 cudnn 6.0
執行gpu版本的 MNIst 時候報錯:
E tensorflow/stream_executor/cuda/cuda_dnn.cc:371] could not create cudnn handle: CUDNN_STATUS_INTERNAL_ERROR
E tensorflow/stream_executor/cuda/cuda_dnn.cc:338] could not destroy cudnn handle: CUDNN_STATUS_BAD_PARAM
F tensorflow/core/kernels/conv_ops.cc:672] Check failed: stream->parent()->GetConvolveAlgorithms( conv_parameters.ShouldIncludeWinogradNonfusedAlgo<T>(), &algorithms)
試了很多辦法
(1):
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))這個方法對我來說不管用,還是之前的錯誤;
(2):執行 sudo rm -f ~/.nv 解決問題
參考:https://github.com/tensorflow/tensorflow/issues/6698
歡迎大家都交流;