-
Clone this repository
git clone https://github.com/ashnvael/FProject_Team26.git
-
ALso in addition to the standard packages contained in the Colab envrionment, please install those
pip install keras-tcn pip install ray['rllib'] pip install nengolib
-
IMPORTANT: In the source files of
ray['rllib']
just installed, please replaceray/rllib/models/modelv2.py
by themodelv2.py
file contained in this project. This file fixes (supposedly) a bug in one of the RLLIB functions.
Jupyter Notebooks are named after the section of the paper their results appear in.
-
Part 1 - In this part we build a DRL agent and provide it with 2 NN models.
-
Part 2 - In this part we label the data and build several models aimed to generate features.
-
Part 3 - In this part we were supposed to extract the features and feed them into the model. The first part shows how to load weights and extract the features. However, due to ineer intricacies of RLLIB I failed to make RLLIB accept the extracted features, and so this section is left without significant results.
To run evaluations, please navigate to the sections of the jupyter Notebook containing '''Evaluation''' in their titles, skipping the training part to directly proceed to the evluation.