https://www.crowdai.org/challenges/nips-2017-learning-to-run

Install Anaconda from https://www.continuum.io/downloads

In the root directory of this repository, input the following commands.

1. Create Conda Environment (We assume the opensim-rl environment has been activated for the rest steps.)

    conda create -n opensim-rl -c kidzik opensim git python=2.7
    source activate opensim-rl    // activate the environment, need to execute this every time
   

2. Install Opensim

    conda install -c conda-forge lapack git
    git clone https://github.com/stanfordnmbl/osim-rl.git
    cd osim-rl
    python setup.py install
    cd ..
    python -c "import opensim"    // confirm it runs smoothly
   

3. Install Tensorflow

    pip install tensorflow==1.2.0    // tensorflow-gpu if GPU is available
   

4. Install Keras

    git clone https://github.com/fchollet/keras.git
    cd keras
    python setup.py install
    cd ..
   

5. Install Other Packages (for plotting, mail notification, model saving, etc)

    pip install pandas h5py matplotlib
   

Make directory for trials

mkdir trials    // all trial results are saved in this folder

Training

python dev/run.py --train
python dev/run.py --train --visualize   // if you want to visualize the process
python dev/run.py --train --trial_dir=trials/xxxx   // if you want to continue training

Test a trained model. It runs for 10 episodes locally.

python dev/run.py --test --trial_dir=trials/xxxx
python dev/run.py --test --trial_dir=trials/xxxx --visualize  // if you like to visualize the process
python dev/run.py --test --trial_dir=trials/xxxx --submit  // to submit to the grader, need a token