davidADSP

import gym

from stable_baselines import PPO1
from stable_baselines.common.policies import MlpPolicy
from stable_baselines.common.callbacks import EvalCallback

env = gym.make('Pendulum-v0')
model = PPO1(MlpPolicy, env)

# Separate evaluation env

davidADSP

class SelfPlayEnv(env):
    # ...
    
    def reset(self):
        super(SelfPlayEnv, self).reset()
        self.setup_opponents()

        if self.current_player_num != self.agent_player_num:   
            self.continue_game()

davidADSP

class SelfPlayEnv(env):
  # ...

    def continue_game(self):
        while self.current_player_num != self.agent_player_num:
            self.render()
            action = self.current_agent.choose_action(self, choose_best_action = False, mask_invalid_actions = False)
            observation, reward, done, _ = super(SelfPlayEnv, self).step(action)
            logger.debug(f'Rewards: {reward}')
            logger.debug(f'Done: {done}')

davidADSP

class SelfPlayEnv(env):
  # ...

    def step(self, action):
        self.render()
        observation, reward, done, _ = super(SelfPlayEnv, self).step(action)
        logger.debug(f'Action played by agent: {action}')
        logger.debug(f'Rewards: {reward}')
        logger.debug(f'Done: {done}')