linux - TensorFlow from sources problems -
i'm on fedora 26 distro
- kernel:
4.12.5
- gcc:
5.4
(5.3.1 recommended, couldn't find @ gnu's) - bazel:
0.5.3
(which bazel
outputs/usr/local/bin/bazel
) - path:
....:/usr/local/bin:.....
- cuda 8.0
i git cloned tensorflow's repository, ran ./configure
following (kept essntials):
lease specify location cuda 8.0 toolkit installed. refer readme.md more details. [default /usr/local/cuda]: /usr please specify location cudnn 5 library installed. refer readme.md more details. [default /usr]:/usr/local/cudnn please specify gcc should used nvcc host compiler. [default /usr/bin/gcc]: /home/elior/gcc54/bin/gcc
if more needed let me know , i'll post it. configuration seems finished , when run
bazel build --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package
i following error
...... cuda configuration error: repository command failed find: ‘/usr/nvvm’: no such file or directory
now few things have gone wrong think of.
- using
which nvcc
outputs/usr/bin/nvcc
when configuration asks me path cuda compiler , reply/usr/bin/nvcc
says/usr/bin/nvcc/lib64/libcudart.so.8.0
not found, did search , found file @/usr/lib64/libcudart.so.8.0
that's why i've put path/usr
- pretty same thing cudnn, downloaded 5.1 site, , i've extracted
/usr/local/cudnn
when put in cudnn version want use 5.1 can't find/usr/local/cudnn/libcudnn.so.5.1
have there 5.0, "5" version , works out
that's come now... want installation done, appreciated.
please specify location cuda 8.0 toolkit installed. refer readme.md more details. [default /usr/local/cuda]: /usr
here need se path cuda installation directory; /usr/local/cuda
(as it's default); set /usr
wrong; either leave default or set /usr/local/cuda
please specify location cudnn 5 library installed. refer readme.md more details. [default /usr]:/usr/local/cudnn
here usual path need set /usr/local/cuda/
cudnn install
cp cudnn/lib64/cudnn* /usr/local/cuda/lib64 cp cudnn/include/* /usr/local/cuda/include
Comments
Post a Comment