SergioSV96 / sfii-challenge

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

MAD_RL_ SFII Challenge

  • ROM name = 'Street Fighter II' - Special Champion Edition (USA).md

  • Agent base model = Asynchronous Advantage Actor-Critic (A3C)

How to test it

  1. Download SFII Challenge Engine docker:
docker pull madrl/sfii-challenge-engine:0.3.2
  1. Execute agent with the engine docker:
docker run -v $PWD/roms:/roms/ -v $PWD/models/:/models/ -v $PWD/:/requirements -v $PWD/sfii_agent_base/:/sfii_agent_base/ -e AGENT_MODULE=sfii_agent_base.agent madrl/sfii-challenge-engine:0.3.2

How to play with your agent

You can modify some parameters using environment variables:

ENGINE_PARAMETERS 

    "EPISODES_TRAINING" = 10,
    "EPISODES_TESTING"= 5,
    "NUM_PROCESSES" = 5,
    "OUTPUT_MODELS_PATH" = "models",
    "DELAY_FRAMES" = 50,
    "ENGINE_MODULE" = "src.environments.gym_retro.engine",
    "ENGINE_CLASS" = "Engine"

Default agent parameters that you can modify inside the agent.py:

AGENT_PARAMETERS

    'frames': 16,
    'cnn_channels': 32,
    'n_outputs': 49,
    'screen_height': 256,
    'screen_width': 200,
    'width': 80,
    'height': 80,
    'start_from_model': "models/sf2_a3c.pth",
    'module': "src.environments.gym_retro.my_agent.agent",
    "class": "Agent"

About

License:MIT License


Languages

Language:Python 100.0%