The core implementation is contained in the following files:
actor.py - implements the actor class
algorithms.py - implements TRPO and NPG, as well as target algorithms (MC returns and GAE)
experiment_class - implements an experiment class and wrapper class to run experiments with different parameters more smoothly
utils.py - implements useful functions, such as sampling memory or updating pytorch model parameters.
Minimal working examples can be found in
min_example_actor.py
demo.py
experiment_demo.py
For the results, we refer to the actual report
Report.pdf