Summary of changes

grep -r "python2" and replace all python2 to just python
check with open(filename, 'r') and with open(filename, 'w')... if they are opening pickle or other binary files, they need to be opened as 'wb' or 'rb'.
fix urllib compatibility (response.info().getheader breaks)
the pickles from the Model Zoo (as of January 2018) seem to be encoded with latin1, so use pickle.load(f, encoding='latin1'); see also tflearn/tflearn#57 However, Python 2 doesn't have an encoding argument to pickle, so a check is needed (try/except or sys version) for compatibility.
Be careful when processing values of type bytes, they may be intended to be ascii strings (as they were in python 2)... e.g. if isinstance(somestr,bytes): somestr.decode('ascii'), which is what bytes2string does in lib/utils/py3compat.py.
Also some types appear in Python 2 as unicode... need to decode them using latin1. 2to3 will change unicode to str, which will leave an error if using unicode_str.decode(), so that has to be checked too.
file is no longer a keyword in python3; opened text files appear as io.TextIOWrapper type.

Git notes

2to3 makes too many changes, and the git changelog would be overwhelming.

Instead this repo remains Python2 code, and fixes issues that 2to3 won't fix by itself.

This is tested on Ubuntu 16.04 with Python 2.7 and 3.5; it should work with tools/infer_simple.py and the tests in tests_to_pass.sh.

To convert to python 3, run this from the root directory: ./python2_to_python3_conversion_automated.sh
Let #!/usr/bin/env python map to python 3 (maybe a python virtual environment, bash alias, or Docker)