UC Berkeley Deep RL in Pytorch

Pytorch starter code for UC Berkeley's CS285 Deep RL course. The code is meant to be used as a direct alternative to the official HW repository for those who would rather complete the course assignments in pytorch. Solutions to this starter code is available.

Changes

All tensorflow in the starter code was converted to pytorch and or numpy, and all solutions are to be written in pytorch. Overall structure of the HW starter code was kept mostly the same, although with the move to pytorch it made sense to delete some files and move thier contents elsewhere. The README.txt in each HW folder has been modified where necessary but the pdf has not - refer to the README.txt for any changes made in the pytorch version. Although tensorflow is not needed within the main code, the logging is still done with tensorboard and thus tensorflow is still needed to easily use and view tensorboard in your browser.

Please note that while this starter code has been shown to produce reasonable results when filled in correctly there may still exist small bugs/errors.