train.py

from unityagents import UnityEnvironment
import numpy as np
import random
import torch
import matplotlib.pyplot as plt
from ddpg_agent import Agent
from collections import deque

env = UnityEnvironment(file_name='Reacher.app')

# get the default brain
brain_name = env.brain_names[0]
brain = env.brains[brain_name]

# reset the environment
env_info = env.reset(train_mode=True)[brain_name]

# number of agents
num_agents = len(env_info.agents)
print('Number of agents:', num_agents)

# size of each action
action_size = brain.vector_action_space_size
print('Size of each action:', action_size)

# examine the state space
states = env_info.vector_observations
state_size = states.shape[1]
print('There are {} agents. Each observes a state with length: {}'.format(states.shape[0], state_size))
print('The state for the first agent looks like:', states[0])

agent = Agent(state_size=state_size, action_size=action_size, random_seed=1)

def ddpg(n_episodes=2000, max_t=20000):

    scores_deque = deque(maxlen=100)
    total_scores = []

    for i_episode in range(1, n_episodes+1):
        env_info = env.reset(train_mode=True)[brain_name]      # reset the environment
        states = env_info.vector_observations                  # get the current state (for each agent)
        agent.reset()
        scores = np.zeros(num_agents)                          # initialize the score (for each agent)

        while True:
            action = agent.act(states)
            env_info = env.step(action)[brain_name]
            next_states = env_info.vector_observations         # get next state (for each agent)
            rewards = env_info.rewards                         # get reward (for each agent)
            dones = env_info.local_done                        # see if episode finished
            agent.step(states, action, rewards, next_states, dones)
            scores += rewards                                  # update the score (for each agent)
            states = next_states                               # roll over states to next time step
            if np.any(dones):                                  # exit loop if episode finished
                break

        scores_deque.append(np.mean(scores))
        total_scores.append(np.mean(scores))

        print('\rEpisode: \t{} \tScore: \t{:.2f} \tAverage Score: \t{:.2f}'.format(i_episode, np.mean(scores), np.mean(scores_deque)), end="")

        if i_episode % 100 == 0:
            torch.save(agent.actor_local.state_dict(), 'checkpoint_actor.pth')
            torch.save(agent.critic_local.state_dict(), 'checkpoint_critic.pth')

        if np.mean(scores_deque)>=30.0:  # consider done when the average score reaches 30 or more
            print('\nEnvironment solved in {:d} episodes!\tAverage Score: {:.2f}'.format(i_episode-100, np.mean(scores_deque)))
            torch.save(agent.actor_local.state_dict(), 'checkpoint_actor.pth')
            torch.save(agent.critic_local.state_dict(), 'checkpoint_critic.pth')
            break

    plt.plot(np.arange(1, len(total_scores)+1), total_scores)
    plt.ylabel('Score')
    plt.xlabel('Episode #')
    plt.show()

ddpg()