GPU云服务器配置keras环境

Ubuntu16.04中anaconda2和anaconda3并存的解决办法

1.在终端输入

conda create -n py27 python=2.7 anaconda
激活py2.7环境
source activate py27
关闭2.7环境
source deactivate.

1.2 jupyter notebook切换内核
$python2 -m pip install ipykernel
$python2 -m ipykernel install --user

1.3  依次安装各种包
conda install nibabel
conda install nipype
sckit-image == 0.13.0
sckit-learn == 0.18.1


2.安装theano,tensorflow,keras,
https://blog.csdn.net/m984789463/article/details/77074451

error: To use MKL 2018 with Theano you MUST set "MKL_THREADING_LAYER=GNU" in your environement.
解决办法
conda install mkl=2017
export MKL_THREADING_LAYER=GNU


3.安装tensorflow-gpu
conda install tensorflow-gpu

3.1 安装后，查看tensorflow-gpu版本

输入python，进入python命令行
import tensorflow as tf
tf.__version__
print('default gpu device {}'.format(tf.test.gpu_device_name()))

查询tensorflow安装路径为:
tf.__path__

检测tensorflow是否使用gpu进行计算
import tensorflow as tf
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))


4.数据
 解压数据从/data到/input
unzip /data/datasets.zip -d /input/datasets

5. 解决MKL2018
conda install mkl=2017

6.keras import error
import tensorflow.python.ops.tensor_array_ops

7.ubuntu 递归解压文件夹内gz文件
 gunzip -r /your/path

翻译中...