model = T.load('model/model.pt', map_location=T.device('cpu'))
env = gym.make('LunarLander-v2')
# env.monitor.start('cartpole-experiment-1', force=True)
agent = Agent(gamma=0.99,
epsilon=0.01,
batch_size=64,
n_actions=4,
eps_end=0.01,
input_dims=[8],
lr=0.001)
agent.Q_eval.load_state_dict(model)
for i in range(13):
done = False
observation = env.reset()
k = 0
tot = 0
while not done:
time.sleep(.005)
action = agent.choose_action(observation)
observation_, reward, done, info = env.step(action)
observation = observation_
env.render()
k += 1
tot += reward
print('episode:', i, '; steps:', k, '; tot reward:', tot, "; sub:",
tot - k)
env.close()
batch_size=64,
eps_end=0.01,
fc1_dims=128,
fc2_dims=128,
replace=100,
n_actions=3)
scores, eps_history = [], []
for i in range(n_games):
done = False
score = 0
game.reset()
while not done:
state = game.get_state()
action = agent.choose_action(state)
reward, done, score = game.play_step(action)
state_ = game.get_state()
score += reward
agent.store_transition(state, action, reward, state_, done)
agent.learn()
eps_history.append(agent.epsilon)
scores.append(score)
avg_score = np.mean(scores[-100:])
print('episode ', i, 'score %.1f' % score,
'average score %.1f' % avg_score, 'epsilon %.2f' % agent.epsilon)
filename = 'snakeAI_Results.png'
x = [i + 1 for i in range(n_games)]
agent.load_models()
np.random.seed(0)
score_history = []
for i in range(750):
done = False
score = 0
obs = env.reset()
obs = state_preproces(obs)
obs = np.reshape(obs, (1,1,32,32))
for j in range(1000):
if done:
break
act = agent.choose_action(obs)
new_state, reward, done, info = env.step_discrete(act)
new_state = state_preproces(new_state)
new_state = np.reshape(new_state, (1,1,32,32))
agent.remember(obs, act, reward, new_state, int(done))
agent.learn()
score+= reward
obs = new_state
score_history.append(score)
print('episode ', i, ' score %.2f' % score," steps: ",j, '100 game average %.2f' %np.mean(score_history[-100:]))
plt.plot(score_history)
plt.xlabel("episode")
plt.ylabel("score")
if i % 25 ==0:
action = env.action_space.sample()
observation_, reward, done, info = env.step(action)
agent.remember(observation, action, reward, observation_, done)
n_steps += 1
agent.learn()
agent.load_models()
evaluate = True
else:
evaluate = False
for i in range(n_games):
observation = env.reset()
done = False
score = 0
while not done:
action = agent.choose_action(observation, evaluate)
observation_, reward, done, info = env.step(action)
score += reward
agent.remember(observation, action, reward, observation_, done)
if not load_checkpoint:
agent.learn()
observation = observation_
score_history.append(score)
avg_score = np.mean(score_history[-100:])
if avg_score > best_score:
best_score = avg_score
print('episode ', i, 'score %.1f' % score,
'avg score %.1f' % avg_score)