cud张量流量误差

Question

当我使用tensorflow时，我遇到了一个错误：

    [W 09:27:49.213 NotebookApp] 404 GET /api/kernels/4e889506-2258-481c-b18e-d6a8e920b606/channels?session_id=0665F3F07C004BBAA7CDF6601B6E2BA1 (127.0.0.1): Kernel does not exist: 4e889506-2258-481c-b18e-d6a8e920b606
    [W 09:27:49.266 NotebookApp] 404 GET /api/kernels/4e889506-2258-481c-b18e-d6a8e920b606/channels?session_id=0665F3F07C004BBAA7CDF6601B6E2BA1 (127.0.0.1) 340.85ms referer=None
    [W 09:27:50.337 NotebookApp] /home/dxq/g++ doesn't exist
    [W 09:27:50.514 NotebookApp] /home/dxq/gcc doesn't exist
    [I 09:28:03.159 NotebookApp] Kernel started: aa5e56b4-df58-4e74-8dc1-96a4cee847aa
    [I 09:28:04.032 NotebookApp] Adapting to protocol v5.1 for kernel aa5e56b4-df58-4e74-8dc1-96a4cee847aa
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcurand.so locally
E tensorflow/core/common_runtime/direct_session.cc:132] Internal: failed initializing StreamExecutor for CUDA device ordinal 0: Internal: failed call to cuCtxCreate: CUDA_ERROR_OUT_OF_MEMORY; total memory reported: 18446744071514750976

这里怎么了？在

以下是完整规格：

^{pr2}$

Answer 1

在终端中运行以下命令：

nvidia-smi

你会得到这样的输出。

您将获得占用GPU内存的进程的摘要。在笔记本电脑中，即使当前没有运行任何单元，但以前正在运行并且本地服务器仍在运行，内存也会被占用。您必须停止占用更多内存的进程，以便为当前进程分配一些带宽。在

Answer 2

检查cuDNN版本。应该是5.1