Skip to content
Learn to Move with Deep Reinforcement Learning - Cmpe462 Machine Learning Spring 2018 @ Boğaziçi University
Jupyter Notebook Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
notebook_files
osim-rl-with-farming @ 8c4f56c
rlkit-with-farming @ f76eca2
scripts
trained_params
training_statistics
.gitmodules
Notebook.ipynb
Poster.pdf
README.md

README.md

Cmpe462-Machine Learning Term Project

"Learn to Move with Deep Reinforcement Learning"

NIPS 2017 Challenge: Learning to Run

Folders & Submodules inside this repository

  • osim-rl -> submodule containing simulation environment
  • rlkit -> submodule containing deep reinforcement learning framework
  • notebook_files -> files necessary to run the notebook
  • scripts -> scripts used to manage training & simulation
  • trained_params -> files composed of trained parameters
  • training_statistics -> files composed of training statistics

Installation

Installation of the submodules are told inside the submodules.

How to simulate with trained parameters?

Run osim-rl-with-farming/sim_farm/farm.py on the opensim-rl conda environment and run scripts/simulate_policy.py with parameter file argument on the rlkit conda environment:

On the first terminal:

source activate opensim-rl
python2 osim-rl-with-farming/sim_farm/farm.py

On the second terminal:

source activate rlkit
python3 scripts/simulate_policy --file <path to parameter file>

How to train models?

Note: To change hyperparameter, see and change method script files.

Run osim-rl-with-farming/farming_scripts/farm.py on the opensim-rl conda environment and run method(ddpg/sac/td3) script on the rlkit conda environment: On the first terminal:

source activate opensim-rl
python2 osim-rl-with-farming/farming_scripts/farm.py

On the second terminal:

source activate rlkit
python3 <path to method(ddpg/sac/td3) script>

Components

Downloading only this repository

Run following command:

git clone https://github.com/simitii/Learn-to-Move-with-Deep-Reinforcement-Learning.git

Downloading All Source Code

Run following command:

git clone --recursive https://github.com/simitii/Learn-to-Move-with-Deep-Reinforcement-Learning.git

Plan

  • Develop training system - DONE
  • Train models(DDPG, SAC, TD3) - DONE
  • Get a lot of visual material - DONE
  • Write an IPython Notebook about the project - DONE
You can’t perform that action at this time.