/gym-example

An example to use Reinforcement Learning technology on AlphaRTC-Gym training a bandwidth estimator

Primary LanguagePython

AlphaRTC Gym example

This repository is an example for AlphaRTC Gym. By converting the raw stats of packet traces to some features and leveraging the PPO algorithms, this example trained a simple bandwidth estimator.

Try this example

  1. Fetch all submodule
git submodule init
git submodule update

  1. Please visit Gym link for the instructions of AlphaRTC Gym to install AlphaRTC Gym in alphartc_gym

  2. Install example dependencies OpenAI GYM

python3 -m pip install -r requirements.txt
  1. Run this example
python3 main.py

If you see something like

Episode 0        Average policy loss, value loss, reward -0.001012914622997811, 1749.7713505035483, -0.5917726188465685
Episode 1        Average policy loss, value loss, reward -0.003164666119424294, 1643.378693441069, -0.5696511401323291
Episode 2        Average policy loss, value loss, reward -0.0006242794975365944, 1503.4368403712722, -0.5418709073877055
Episode 3        Average policy loss, value loss, reward -0.0013577909024748813, 1396.8935836247986, -0.5149282318393551
Episode 4        Average policy loss, value loss, reward -0.0002334891452391077, 1349.4928827780047, -0.5091586053527624

which means this example has readied in your environment. And you can find your model under the folder data.