def setUp(self) -> None:
self.buffer = MultiStepBuffer(buffer_size=10, n_step=2)
self.state = np.zeros([32, 32])
self.state_02 = np.ones([32, 32])
self.next_state = np.zeros([32, 32])
self.next_state_02 = np.ones([32, 32])
self.action = np.zeros([1])
self.action_02 = np.ones([1])
self.reward = np.zeros([1])
self.reward_02 = np.ones([1])
self.done = np.zeros([1])
self.done_02 = np.zeros([1])
self.experience01 = Experience(self.state, self.action, self.reward, self.done, self.next_state)
self.experience02 = Experience(self.state_02, self.action_02, self.reward_02, self.done_02, self.next_state_02)
self.experience03 = Experience(self.state_02, self.action_02, self.reward_02, self.done_02, self.next_state_02)
def setUp(self) -> None:
self.buffer = PERBuffer(10)
self.state = np.random.rand(32, 32)
self.next_state = np.random.rand(32, 32)
self.action = np.ones([1])
self.reward = np.ones([1])
self.done = np.zeros([1])
self.experience = Experience(self.state, self.action, self.reward, self.done, self.next_state)
def setUp(self) -> None:
self.net = Mock()
self.agent = DummyAgent(net=self.net)
self.env = gym.make("CartPole-v0")
self.n_step = 2
self.source = NStepExperienceSource(self.env, self.agent, Mock(), n_steps=self.n_step)
self.state = np.zeros([32, 32])
self.state_02 = np.ones([32, 32])
self.next_state = np.zeros([32, 32])
self.next_state_02 = np.ones([32, 32])
self.action = np.zeros([1])
self.action_02 = np.ones([1])
self.reward = np.zeros([1])
self.reward_02 = np.ones([1])
self.done = np.zeros([1])
self.done_02 = np.zeros([1])
self.experience01 = Experience(self.state, self.action, self.reward, self.done, self.next_state)
self.experience02 = Experience(self.state_02, self.action_02, self.reward_02, self.done_02, self.next_state_02)
self.experience03 = Experience(self.state_02, self.action_02, self.reward_02, self.done_02, self.next_state_02)
def setUp(self) -> None:
self.state = np.random.rand(32, 32)
self.next_state = np.random.rand(32, 32)
self.action = np.ones([1])
self.reward = np.ones([1])
self.done = np.zeros([1])
self.experience = Experience(self.state, self.action, self.reward, self.done, self.next_state)
self.source = Mock()
self.source.step = Mock(return_value=(self.experience, torch.tensor(0), False))
self.warm_start = 10
self.buffer = ReplayBuffer(20)
for _ in range(self.warm_start):
self.buffer.append(self.experience)
def setUp(self) -> None:
self.state = np.random.rand(4, 84, 84)
self.next_state = np.random.rand(4, 84, 84)
self.action = np.ones([1])
self.reward = np.ones([1])
self.done = np.zeros([1])
self.experience = Experience(self.state, self.action, self.reward, self.done, self.next_state)
self.source = Mock()
self.source.step = Mock(return_value=(self.experience, torch.tensor(0), False))
self.batch_size = 8
self.buffer = Buffer(8)
for _ in range(self.batch_size):
self.buffer.append(self.experience)
def step(self) -> Tuple[Experience, float, bool]:
"""Takes a single step through the environment"""
action = self.agent(self.state, self.device)
new_state, reward, done, _ = self.env.step(action)
experience = Experience(state=self.state,
action=action,
reward=reward,
new_state=new_state,
done=done)
self.state = new_state
if done:
self.state = self.env.reset()
return experience, reward, done
def step(self) -> Experience:
"""Carries out a single step in the environment"""
action = self.agent(self.state, self.device)
new_state, reward, done, _ = self.env.step(action)
experience = Experience(state=self.state,
action=action,
reward=reward,
new_state=new_state,
done=done)
self.state = new_state
if done:
self.state = self.env.reset()
return experience
def step(self) -> Tuple[Experience, float, bool]:
"""
Takes an n-step in the environment
Returns:
Experience
"""
exp = self.single_step()
while len(self.n_step_buffer) < self.n_steps:
self.single_step()
reward, next_state, done = self.get_transition_info()
first_experience = self.n_step_buffer[0]
multi_step_experience = Experience(first_experience.state,
first_experience.action, reward,
done, next_state)
return multi_step_experience, exp.reward, exp.done