• conda cheatsheet

• Publish documentation created with sphinx to github. Note that you don’t need to install the extra package. You can bypass Jekyll by adding a .nojekyll file to the repo

• NumPy style documentation

• Miniconda: Quick Installing full scientific python on a new unix machine. Instructions for setting it up on Travis-CI

• python subprocess example in a unittest class

def setUp(self):
self.server_log_file = open('.tsdb_server.log.test','w')
self.server_proc = subprocess.Popen(['python', 'go_server.py']
,stdout=self.server_log_file,stderr=subprocess.STDOUT)
time.sleep(1)

def tearDown(self):
self.server_proc.terminate()
self.server_log_file.close()
time.sleep(1)

• Check if any python jobs are running in the background (of course this doesn’t have to be python)

jobs
ps aux | grep 'python'

• End-to-End Memory Networks for Question Answering in Python with very nice visualizations

• nngraph resources

• Oxford ML Practical

• Sample Code (harvardnlp) here and here and some more that I wrote. Important thing that I found was that there is little to no thought given to the initial values of model params and gradParams. Set them yourself.

• Convert .dot files (graphviz) between formats:

• PDF dot -Tpdf filename.dot -o filename.pdf

• PostScript dot -Tps filename.dot -o filename.ps

• bash shell conditionals

#!/bin/bash

if [ $1 == "-h" ]; then
echo "Syntax::sh scriptname output-file classifer-name num-epochs"
exit
fi

# check outputfile
if [ -z $1 ]; then
echo "Output file not specified. Using 'output.all' "
OUTFILE="output.all"
else
OUTFILE=$1
if [ -e $OUTFILE ]; then
echo "Emptying " $OUTFILE
rm $OUTFILE
fi
echo "Output File = " $OUTFILE
fi

• Get a specific file from a different branch on git

git checkout branch2 file.py

• Get a specific file from a different branch on git and check changes line by line

git checkout --patch branch2 file.py