STAC Learning Algorithms

Setting up "Learning"

Note: these install instructions are based on the installation structions for OpenAI's spinningup found here: https://spinningup.openai.com/en/latest/user/installation.html.

Also note that I've only tested this on Mac. I suspect it will work with Linux, but am less sure about Windows. (Mostly because OpenAI says they're not sure whether it works with Windows!)

1. First, we will create a virtual environment for learning: virtualenv -p `which python3` stac

2. Then, activate this environment: source stac/bin/activate

Always activate this environment before using this repo. You can exit the environment by typing deactivate when you're done working.

3. Install OpenMPI [This section is OS-dependent]:

• Ubuntu: sudo apt-get update && sudo apt-get install libopenmpi-dev
• Mac: brew install openmpi

4. If you haven't already cloned this git repo, clone it and navigate into it:

  git clone https://github.com/space-technologies-at-california/learning
  cd learning

5. Use pip to install dependencies: pip install -e .

Note that you must be in the learning directory to do this.

6. Check that installation was successful. From within the learning directory, run python test.py. This should begin training an agent for a LunarLander environment.

This will run for a while; once it's done, you can run python -m spinup.run plot data/test1 to see a plot of the results of this training.

You can also run `python -m spinup.run test_policy data/test1 to see a video of the trained policy.