【发布时间】：2016-08-04 17:02:57
【问题描述】：
我正在尝试使用以下配置从源安装 Tensorflow：
- NVIDIA GTX 1070
- UBUNTU 16.04
- CUDA 8.0
- Cudnn v5.0
我已按照here 的以下步骤进行操作：
- 已安装 bazel
- 已安装的依赖项
- 已安装 CUDA 支持
-
./configure支持 CUDA 8.0 bazel build -c opt --config=cuda //tensorflow/tools/pip_package:build_pip_package
在这一步之后，据我所知，应该有一个bazel-bin 目录，以便我随后可以执行
$ bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
# The name of the .whl file will depend on your platform.
$ sudo pip install /tmp/tensorflow_pkg/tensorflow-0.10.0rc0-py2-none-any.whl
但是，没有这样的目录。
我感觉这个错误信息可能与它有关？
ERROR: /usr/local/lib/python2.7/dist-packages/tensorflow_clone/tensorflow/contrib/rnn/BUILD:45:1: error while parsing .d file: /home/volcart/.cache/bazel/_bazel_volcart/62dff5ffffc63bcd8a9350984645e0be/execroot/tensorflow_clone/bazel-out/local_linux-opt/bin/tensorflow/contrib/rnn/_objs/python/ops/_lstm_ops_gpu/tensorflow/contrib/rnn/kernels/lstm_ops_gpu.cu.pic.d (No such file or directory).
nvcc warning : option '--relaxed-constexpr' has been deprecated and replaced by option '--expt-relaxed-constexpr'.
In file included from third_party/gpus/cuda/include/cuda_runtime.h:78:0,
from <command-line>:0:
third_party/gpus/cuda/include/host_config.h:115:2: error: #error -- unsupported GNU version! gcc versions later than 5.3 are not supported!
#error -- unsupported GNU version! gcc versions later than 5.3 are not supported!
在重新执行bazel build ... 后，我发现了这个...
WARNING: /usr/local/lib/python2.7/dist-packages/tensorflow/util/python/BUILD:11:16: in includes attribute of cc_library rule //util/python:python_headers: 'python_include' resolves to 'util/python/python_include' not in 'third_party'. This will be an error in the future.
我也应该添加这个...
$ bazel version
Build label: 0.3.1
Build target: bazel-out/local-fastbuild/bin/src/main/java/com/google/devtools/build/lib/bazel/BazelServer_deploy.jar
Build time: Fri Jul 29 09:09:52 2016 (1469783392)
Build timestamp: 1469783392
Build timestamp as int: 1469783392
【问题讨论】：
-
"不支持 5.3 以上的 gcc 版本！"听起来对我来说是一个非常明确的错误信息.....
-
@talonmies 同意，我唯一的疑问是它包含了
contrib目录的路径。虽然，我想这仍然可能导致整个事情失败，现在尝试不同的版本 -
@talonmies 原来是权限问题——Doh！似乎总是一个权限问题
sudo bazel build ...
标签： tensorflow ubuntu-16.04 cudnn