Local Machine

Our software packages are developed mainly in python on Linux machines. For deep learning related functionality, GPUs are required for speed.

• OS

  • Unix-based OS (e.g. Linux, Mac OS X).

  • Recommendation: Ubuntu LTS

• Install Miniconda (python 3.7): create a new conda env for each project

  • download

  • create new env:

    conda create -n ${conda-env-name}
    

  • activate new env:

    source activate ${conda-env-name}
    

  • deactivate env:

    source deactivate
    

• Install Jupyter notebook for interactive result display

  • create new kernel (first install ipython, ipykernel):

    ipython kernel install --user --name ${conda-env-name} --display-name "${display-name}"
    

• Mount pfister_lab2 file system to local machine

  • Install packages:

    sudo apt-get install cifs-utils
    

  • Get your gid on your local machine: id

  • Create the folder if it doesn’t exist sudo mkdir /mnt/pfister_lab2

  • Mount it with your rc username and local machine gid:

    sudo mount -t cifs -o vers=1.0,workgroup=rc,username=${1},gid=${2} \
    //coxfs01.rc.fas.harvard.edu/pfister_lab2 /mnt/pfister_lab2
    

• Unix Tips

  • Terminal (split screen)

    • On mac: try iterm2

    • On Linux: try terminator or tmux

  • ssh

    • Automatic login in new bashes (after the login in a bash)

    • Create a file with the following content: vim ~/.ssh/config:

      Host *
        ControlMaster auto
        ControlPath ~/.ssh/master-%r@%h:%p
      

  • bash

    • Add useful alias: vim ~/.bashrc

      alias csh='ssh ${USERNAME}@login.rc.fas.harvard.edu'