Lunerland.py

import numpy as np
import gym
from utils import *
from agent import *
from config import *

def train(env, agent, num_episode, eps_init, eps_decay, eps_min, max_t):
    rewards_log = []
    average_log = []
    eps = eps_init

    for i in range(1, 1 + num_episode):

        episodic_reward = 0
        done = False
        state = env.reset()
        state = state[0]
        t = 0
        while not done and t < max_t:

            t += 1
         
            action = agent.act(state, eps)
           
            next_state, reward, done, _, _ = env.step(action)
      
            agent.memory.append((state, action, reward, next_state, done))
            if t % 4 == 0 and len(agent.memory) >= agent.bs:
                agent.learn()
                agent.soft_update(agent.tau)
            state = next_state.copy()
            episodic_reward += reward
        rewards_log.append(episodic_reward)
        average_log.append(np.mean(rewards_log[-100:]))
        print('\rEpisode {}, Reward {:.3f}, Average Reward {:.3f}'.format(i, episodic_reward, average_log[-1]), end='')
        eps = max(eps * eps_decay, eps_min)
        if i % 100 == 0:
            print()
    return rewards_log


if __name__ == '__main__':
    print(RAM_ENV_NAME)
    env = gym.make(RAM_ENV_NAME)
    print(env.observation_space.shape[0], env.action_space.n)
    agent = Agent(env.observation_space.shape[0], env.action_space.n, BATCH_SIZE, LEARNING_RATE, TAU, GAMMA, DEVICE, False)
    rewards_log = train(env, agent, RAM_NUM_EPISODE, EPS_INIT, EPS_DECAY, EPS_MIN, MAX_T)
    np.save('{}_0rewards_ram.npy'.format(RAM_ENV_NAME), rewards_log)
    agent.Q_local.to('cpu')
    torch.save(agent.Q_local.state_dict(), '{}_0weights_ram.pth'.format(RAM_ENV_NAME))