Skip to content

My master's thesis: Custom MuZero implementation in PyTorch + Thesis paper in Typst

License

Notifications You must be signed in to change notification settings

raphCode/master-thesis-muzero

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 

Repository files navigation

My Master's Thesis

Deep Reinforcement Learning with MuZero: Theoretical Foundations, Variants, and Implementation for a Collaborative Game

Repo organisation

This repo contains the source code of:

The repo is mostly a show-off of the work I did for my thesis, do not expect any more value or useable software.

The final compiled thesis PDF can be viewed in thesis.pdf and the slides in slides.pdf.

Other branches

  • paper-*: old versions of the paper source code, in different typesetting systems. I started in LaTeX (branch), then tried Sile (branch), before I discovered Typst.
  • wip*: Various work-in-progress side branches of the MuZero implementation, most to all work should also be in master. Numbered according to age.
  • all other branches: Various older checkpoints of the MuZero implementation, not really interesting except for understanding the development history.

Commit History

I made heavy use of history-rewriting to tidy up commit diffs and messages. That means commit dates may be all over the place (but author dates are accurate).

Near the end, I biseced some performance regressions, and had to fix bugs in very old commits. Fixing these bugs required rewriting the history of the master branch, and I never rebased the wip branches, so their history does not line up anymore nicely.

About

My master's thesis: Custom MuZero implementation in PyTorch + Thesis paper in Typst

Resources

License

Stars

Watchers

Forks