JupyterLab远程访问配置方法（CentOS7）

下载 Anaconda3安装包，并执行安装命令：

bash Anaconda3-2019.07-Linux-x86_64.sh

确定安装并初始化：

Do you wish the installer to initialize Anaconda3
by running conda init? [yes|no]
[no] >>> yes

安装完Anaconda3后用conda命令安装jupyterlab：

conda install jupyterlab

需要用ipython命令生成秘钥，启动ipython：

ipython

执行命令生成秘钥：

In []: from notebook.auth import passwd
In []: passwd()
Enter password:
Verify password:
Out[]: 'sha1:f704b702aea2:01e2bd991f9c7208ba177b46f4d10b6907810927'

产生jupyterlab配置文件：

jupyter lab --generate-config

修改配置文件：

vi /root/.jupyter/jupyter_notebook_config.py

更改内容如下：

# 将ip设置为*，意味允许任何IP访问
c.NotebookApp.ip = '*'
# 这里的密码就是上边我们生成的那一串
c.NotebookApp.password = # 允许远程访问 c.NotebookApp.allow_remote_access = True

接下来输入jupyter lab启动jupyter服务即可：

jupyter lab --allow-root

返回如下信息：

[W ::12.159 LabApp] WARNING: The notebook server is listening on all IP addresses and not using encryption. This is not recommended.
[I ::/site-packages/jupyterlab
[I ::12.166 LabApp] JupyterLab application directory is /root/anaconda3/share/jupyter/lab
[I ::12.169 LabApp] Serving notebooks from local directory: /root
[I ::12.169 LabApp] The Jupyter Notebook is running at:
[I ::12.169 LabApp] http://dn07:8888/
[I ::12.169 LabApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
[I :: GET /?token=fe7d07705f34f7e0d81d0c169f06f5cb11039cd13092b4d6 (.53ms
[I :: GET /lab?token=fe7d07705f34f7e0d81d0c169f06f5cb11039cd13092b4d6 (.55ms
[E ::47.463 LabApp] Could not open static file ''
[W :: GET /static/components/react/react-dom.production.min.js (.64ms referer=http://10.200.101.112:8888/login?next=%2Flab%3Ftoken%3Dfe7d07705f34f7e0d81d0c169f06f5cb11039cd13092b4d6
[W :: GET /static/components/react/react-dom.production.min.js (.58ms referer=http://10.200.101.112:8888/login?next=%2Flab%3Ftoken%3Dfe7d07705f34f7e0d81d0c169f06f5cb11039cd13092b4d6

打开页面查看：

直接点击“Log in”登录主界面：

解释：Ipython把输入的密码转换成sha，并用于认证JupyterLab，本文在Ipython输入密码和确认密码时直接回车，相当于不设密码，因此登录JupyterLab时可以不输入密码直接点击登录。

安装好Tensorflow：

安装好Keras：

查找site-packages文件目录：

find / -name 'site-packages' -type d

经比较排除，确定具体地址为：

/root/anaconda3/lib/python3./site-packages

下载mnist.npz数据，修改mnist.py文件，参考： https://www.cnblogs.com/ratels/p/11144881.html 和 https://www.cnblogs.com/shinny/p/9283372.html ，如下：

复制并执行CNN代码，参考：https://www.cnblogs.com/ratels/p/11144881.html

服务器相比便携本速度提高了不少，平均每个Epoch的训练时间由7s缩短到4s。

复制并执行Seq2Seq代码，参考：https://www.cnblogs.com/ratels/p/11192045.html

服务器相比便携本速度提高了不少，平均每个Epoch的训练时间由140s缩短到57s。

参考：

https://blog.csdn.net/qixizhuang/article/details/82793442

https://www.cnblogs.com/xiao-apple36/p/9052102.html

https://www.cnblogs.com/ratels/p/11144881.html

https://www.cnblogs.com/shinny/p/9283372.html

https://blog.csdn.net/hhhzua/article/details/80395352