tensorflow: Error Building Docker Image

Using Dockerfile.devel-gpu branched from 459c2fed498530b794c4871892fd68d1e6834ac6.

Configuration finished
INFO: Reading 'startup' options from /root/.bazelrc: --batch
ERROR: com.google.devtools.build.lib.packages.BuildFileContainsErrorsException: error loading package '': Encountered error while reading extension file 'cuda/build_defs.bzl': no such package '@local_config_cuda//cuda': Traceback (most recent call last):
    File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 406
        _create_cuda_repository(repository_ctx)
    File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 340, in _create_cuda_repository
        _find_cudnn_lib_path(repository_ctx, cudnn_install_base..., ...)
    File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 249, in _find_cudnn_lib_path
        fail("Cannot find %s or %s under %s" ...))
Cannot find lib64/libcudnn.so.5 or libcudnn.so.5 under /usr/local/cuda.
____Elapsed time: 1.274s
The command '/bin/sh -c ./configure &&     bazel build --local_resources 3072,3.0,1.0 -c opt --config=cuda tensorflow/tools/pip_package:build_pip_package &&     bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/pip &&     pip install --upgrade /tmp/pip/tensorflow-*.whl' returned a non-zero code: 1

./build-all.sh returned exit code 1

About this issue

Original URL
State: closed
Created 8 years ago
Comments: 16 (13 by maintainers)

Most upvoted comments

Yes, from the discussion in NVIDIA/nvidia-docker#172 and #3269, the line:

ENV CUDNN_INSTALL_PATH /usr/local/cuda

seems to be incorrect since in the Docker image, libcudnn.so is installed under /usr/lib/x86_64-linux-gnu and not /usr/local/cuda/lib64. The line should be changed to:

ENV CUDNN_INSTALL_PATH /usr/lib/x86_64-linux-gnu

as we have done so in tensorflow/tools/ci_build/Dockerfile.gpu.

davidzchen on Aug 24, 2016