Setup Environment for Berkeley CS294(Deep Reinforcement Learning) Homework

Posted on by 一言以概

# Set up development environment for Berkeley CS294 Homework: http://rll.berkeley.edu/deeprlcourse/

# Platform: Ubuntu 17.04, Python 3.5.3

#294 Repo: https://github.com/berkeleydeeprlcourse/homework

Readme here: https://github.com/berkeleydeeprlcourse/homework/blob/master/hw1/README.md

Dependencies: TensorFlow, MuJoCo version 1.31, OpenAI Gym

  1. MuJoCo: Download version 1.3.1. mjpro131 linux Get a student licence and install folder is ~/.mujoco Place the licence file mjkey.txt under /.mujoco and ~/.mujoco/mjpro131/bin
  2. You can run ~/.mujoco/mjpro131/bin/simulate to test, but it needs libglfw.so.3 under ~/.mujoco/mjpro131/bin, run `export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/yuan8421/.mujoco/mjpro131/bin
    `
  3. Clone gym repo at: https://github.com/openai/gym#examples, follow the instruction in Install section
  4. git clone https://github.com/openai/gym.git
    cd gym
    pip3 install -e .
  5. mujoco-py Because we use MuJoCo 1.3.1, the coresponding mujoco-py version is 0.5.7
  6. pip3 install mujoco-py==0.5.7
  7. Navigate to homework/hw1, modify script demo.bash python to python3. Run it and check the result