Ubuntu 18.04 CUDA10.1 Anaconda 安装tensorflow-gpu 1.13.1
由于想使用https://github.com/matterport/Mask_RCNN
这个代码使用的是比较老版本的tensorflow 对应的cuda版本是10.0或9.0或9.2，我系统安装的cuda是10.1版本
cudatoolkit是库不是驱动，库意味着不是必须安装到系统目录，可以安装到用户目录，具体链接哪些库，怎么链接由Makefile决定，可执行文件执行的时候，在LD_LIBRARY_PATH路径或其他路径里能找到需要的库文件即可。
查看LD_LIBRARY_PATH方法
echo $LD_LIBRARY_PATH查看nvidia驱动是否安装好
nvidia-smi
nvcc -v在系统安装好驱动后，anaconda安装cudatoolkit是很方便的，需要配置好源，我使用的源如下(~/.condarc):
ssl_verify: true
channels:
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
- https://mirrors6.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/
- https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
- https://mirrors6.tuna.tsinghua.edu.cn/anaconda/cloud//pytorch/
show_channel_urls: true安装cudatoolkit命令
conda install cudatoolkit=10.0安装tensorflow-gpu命令
#1
conda install tensorflow-gpu=1.13.1
#2
pip install tensorflow-gpu=1.13.1/home/kitty/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:526: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
_np_qint8 = np.dtype([("qint8", np.int8, 1)])
出现上面警告是由于，tensorflow-gpu 1.13.1要求numpy版本不能太高，可以使用下面命令解决。
pip uninstall numpy
pip install numpy==1.16.4测试GPU对安装的tensorflow是否可用
import tensorflow as tf
print(tf.test.is_gpu_available())
如果返回True，接下来tensorflow-gpu就可以正常使用了。
安装好后python库列表
(py36) kitty@heal-PowerEdge-T630:~$ conda list
# packages in environment at /home/kitty/anaconda3/envs/py36:
#
# Name Version Build Channel
_libgcc_mutex 0.1 main https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
blas 1.1 openblas https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
ca-certificates 2020.6.20 hecda079_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
certifi 2020.6.20 py36h9f0ad1d_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
cudatoolkit 10.0.130 0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
future 0.18.2 pypi_0 pypi
keras-preprocessing 1.1.2 pypi_0 pypi
ld_impl_linux-64 2.33.1 h53a641e_7 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
libedit 3.1.20191231 h14c3975_1 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
libffi 3.3 he6710b0_2 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
libgcc-ng 9.1.0 hdf63c60_0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
libgfortran-ng 7.5.0 hdf63c60_14 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
libstdcxx-ng 9.1.0 hdf63c60_0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
mock 4.0.2 pypi_0 pypi
ncurses 6.2 he6710b0_1 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
numpy 1.16.4 pypi_0 pypi
openblas 0.3.3 h9ac9557_1001 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
openssl 1.1.1g h516909a_1 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
pip 20.2.1 py36_0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
python 3.6.10 h7579374_2 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
python_abi 3.6 1_cp36m https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
readline 8.0 h7b6447c_0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
setuptools 49.2.1 pypi_0 pypi
sqlite 3.32.3 h62c20be_0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
tensorflow-gpu 1.13.1 pypi_0 pypi
tk 8.6.10 hbc83047_0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
wheel 0.34.2 pypi_0 pypi
xz 5.2.5 h7b6447c_0 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
zlib 1.2.11 h7b6447c_3 https://mirrors6.tuna.tsinghua.edu.cn/anaconda/pkgs/main
参考文献：
1.FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.解决办法
2.官网经过测试的构建配置
3.非ROOT用户安装CUDA和CUDNN
4.如何查看tensorflow-gpu是否可用