The Jupyter notebook in this repository converts the ARPABET transcriptions in cmudict-0.7b to Unicode IPA symbols.
The notebook uses Python 3. (Note that Python /3/ makes Unicode much more straightforward and painless to deal with than Python 2.) There are no other salient dependencies.