Python Reinforcement Learning Projects

train.py

Finally, train.py is where all the functions we implemented in the controller are called and coordinated. More specifically, we execute each step as subprocess:

import subprocessimport sysfrom utils import timerimport osfrom constants import PATHSimport logginglogger = logging.getLogger(__name__)def main():    if not os.path.exists(PATHS.SELFPLAY_DIR):        with timer("Initialize"):            logger.info('==========================================')            logger.info("============ Initializing...==============")            logger.info('==========================================')            res = subprocess.call("python controller.py initialize-random-model", shell=True)        with timer('Initial Selfplay'):            logger.info('=======================================')            logger.info('============ ...