Skip to content
This repository has been archived by the owner on Aug 24, 2019. It is now read-only.

Latest commit

 

History

History
289 lines (267 loc) · 9.4 KB

setup.md

File metadata and controls

289 lines (267 loc) · 9.4 KB
layout title permalink
page
Setup
/setup/

NIH HPC setup

For this course we will use Felix, an NIMH-dedicated high performance computing resource or, for non-NIMH participants at the NIH, Helix. If you don't have an NIH HPC account, please request one. We will access these computers using the NoMachine remote desktop client. To install NoMachine:

  • Windows
  • Mac OSX
  • Linux (Ubuntu example)
    • Download the installer to your downloads directory.
    • In the terminal run:
cd ~/Downloads
sudo dpkg -i nomachine_5.2.11_1_amd64.deb # the name of the downloaded version

{: bash}

After NoMachine is installed, create a new connection to felix.nimh.nih.gov (for NIMH users) or helix.nih.gov (for non-NIMH users). Maintain the default configuration selections (NX port 4000, no proxy, password authentication).

After using NoMachine to start a remote desktop session on felix (or helix):

Open the terminal application and type:

source /data/DSST/scripts/repro_env_setup.sh

{: bash}

You're ready to go after you open the jupyter qtconsole by typing:

jupyter qtconsole &

{: bash}

You are now setup for the HPC portion of the class. Some exercises will be conducted directly on your laptop. Read on below the boxed text on how to install the needed software.

Under the hood (for those who are interested)

The above script executes the commands listed below. These commands load modules on felix to provide access to additional software. Specifically conda, git, and the text editor atom. Conda is associated with environments (packages of software isolated from each other). In this case we are using the environment "py3.5" pre-installed on the NIH HPC resources. The source activate command selects this conda environment. Since we wish to use jupyter qtconsole as our interface to IPython we generate a configuration file and then use the unix "sed" text processing utility to add atom, as the associated text editor for our work.

See script commands

module load Anaconda
module load git
source activate py3.5
module load atom
jupyter qtconsole --generate-config
sed -i "/#c.JupyterWidget.editor = ''/a c.JupyterWidget.editor = 'atom'" ~/.jupyter/jupyter_qtconsole_config.py

{: .bash} {: .solution}

{: .challenge}

A guide to local installation

Click here for Windows

  • Please install the text editor Atom.

  • Download and install miniconda if you do not already have conda installed on your system

  • Once installed open the Anaconda prompt and type:

  conda create -n py3.5 python=3.5 jupyter pandas seaborn git

{: .source}

  • To use the version of python along with the packages that we installed we must activate the py3.5 conda environment. We do this by typing the following at the Anaconda prompt:
activate py3.5

{: .source}

  • To configure jupyter qtconsole to use atom type the following at the Anaconda prompt:
jupyter qtconsole --generate-config

{: .source}

Then open atom and navigate to the config file listed at the Anaconda prompt.

Edit the line containing the text:

#c.JupyterWidget.editor = ''

{: .source}

Change this line to

 c.JupyterWidget.editor = 'atom'

{: .source}

  • Close atom and the Anaconda prompt. Your system is now set up.

{: .solution}

Click here for Mac OS X

A note to administrators

  • Upon installation Anacaconda prepends itself to the PATH (command placed in ~/.bash_profile). This won't work cleanly on many NIH laptops as their login shell is set to /bin/sh. Note "echo $SHELL" will still say /bin/bash. The login shell should be set to /bin/bash for everything to work.

  • If conda already exists on the laptop (eg. in bash check for output from ls ~|grep conda) then this can be used to construct a working environment for the class:

conda create -n py3.5 python=3.5 jupyter git pandas seaborn

{: .bash}

And then modify the qtconsole config file as described below.

  • Many users wish to have tcsh as they're default shell. Our current solution for this is the following: In order to maintain tcsh as their login shell and separate the setup for this course as cleanly as possible for the user we have installed the terminal application iterm2. In the preferences for this application, under the profile tab, is a section called "Command". Changing the selection from "Login shell" to "Command" and typing /bin/bash here will set all iterm sessions to interactive bash sessions. All modifications to the path for these sessions must be made in ~/.bashrc instead of ~/.bash_profile. The change comprises the package manager conda (an path with anaconda3 in it in this case) and /usr/local/bin if it is not already on the path so that the text editor atom works from the command line.

  • Installation of atom has posed some problems for people. One easily solvable problem is that /usr/local/bin is not on the path. The other problem has been that for computers where the user does not have administrative privileges. The solution to this is the following:

  • Download Atom, extract it, move it to applications, select "Authenticate", type in the admin password. In bash:

su - admin_user # change as appropriate
sudo /Applications/Atom.app/Contents/Resources/app/atom.sh
~~~
{: .bash}

  After this, in a new iTerm window the editor should be opened by
   executing the command "atom".

{: .solution}

  • Please install the text editor Atom. Do this by downloading it and moving it to the Applications folder. You will require administrator privileges for this. Once installed open Atom and select "Atom> Install Shell Commands" in the menu bar. Once this is done, typing "atom" in a bash terminal should open the application. If not you may have to add "/usr/local/bin" to your path.

  • Download and install miniconda if you do not already have conda installed on your system

  • Once installed the command "which conda" at a terminal should return a path to the installation. Then type:

  conda create -n py3.5 python=3.5 jupyter pandas seaborn git

{: .source}

  • To use the version of python along with the packages that we installed we must activate the py3.5 conda environment. We do this by typing the following at bash terminal:
source activate py3.5

{: .source}

  • To configure jupyter qtconsole to use atom type the following at the terminal prompt:
jupyter qtconsole --generate-config

{: .source}

Then open atom and navigate to the config file listed (probably "~/.jupyter/jupyter_qtconsole_config.py")

Edit the line containing the text:

#c.JupyterWidget.editor = ''

{: .source}

Change this line to

 c.JupyterWidget.editor = 'atom'

{: .source} {: .solution}

Click here for Linux

  • Please install the text editor Atom.

  • Download and install miniconda if you do not already have conda installed on your system

  • Once installed the command "which conda" at a terminal should return a path to the installation. Then type:

  conda create -n py3.5 python=3.5 jupyter pandas seaborn git

{: .source}

  • To use the version of python along with the packages that we installed we must activate the py3.5 conda environment. We do this by typing the following at bash terminal:
source activate py3.5

{: .source}

  • To configure jupyter qtconsole to use atom type the following at the terminal prompt:
jupyter qtconsole --generate-config
sed -i "/#c.JupyterWidget.editor = ''/a c.JupyterWidget.editor = 'atom'" ~/.jupyter/jupyter_qtconsole_config.py

{: .source} {: .solution}

Once the software is installed open a terminal (Linux or Mac OSX) or Anaconda prompt (Windows). Now type (same for all operating systems):

  jupyter qtconsole &

{: .source}