def setUp(self):
self.memory_params = MemoryParams({"verbose": 0})
self.memory_params.window_length = 5
self.memory_params.memory_size = 1000
self.memory_params.seed = 123
self.memory_params.combined_with_last = False
self.memory = ReplayBuffer(self.memory_params)
class TestMemoryOverwrite(unittest.TestCase):
def setUp(self):
self.memory_params = MemoryParams({"verbose": 0})
self.memory_params.window_length = 1
self.memory_params.memory_size = 24
self.memory_params.seed = 123
self.memory_params.combined_with_last = False
states = np.arange(50).reshape(-1, 2)
next_states = np.arange(2, 52).reshape(-1, 2)
actions = np.arange(0, 25, 1)**2
rewards = np.linspace(-1, 2, 25)
terms = np.zeros(25)
self.memory = ReplayBuffer(self.memory_params)
for s, a, r, s2, d in zip(states, actions, rewards, next_states,
terms):
self.memory.append(s, a, r, s2, d)
self.memory.append_recent(s, d)
def test_rolling(self):
self.assertTrue(
np.all(np.array([2, 3]) == self.memory.memory[0].state))
def test_overwritting(self):
self.assertTrue(
np.all(np.array([48, 49]) == self.memory.memory[-1].state))
def setUp(self):
self.memory_params = MemoryParams({"verbose": 0})
self.memory_params.window_length = 2
self.memory_params.memory_size = 1000
self.memory_params.seed = 123
self.memory_params.combined_with_last = False
states = np.arange(50).reshape(-1, 2)
next_states = np.arange(2, 52).reshape(-1, 2)
actions = np.arange(0, 25, 1)**2
rewards = np.linspace(-1, 2, 25)
terms = np.zeros(25)
self.memory = ReplayBuffer(self.memory_params)
for s, a, r, s2, d in zip(states, actions, rewards, next_states,
terms):
self.memory.append(s, a, r, s2, d)
self.memory.append_recent(s, d)
class TestMemoryCombinedWithLast(unittest.TestCase):
def setUp(self):
self.memory_params = MemoryParams({"verbose": 0})
self.memory_params.window_length = 1
self.memory_params.memory_size = 24
self.memory_params.seed = 123
self.memory_params.combined_with_last = True
states = np.arange(50).reshape(-1, 2)
next_states = np.arange(2, 52).reshape(-1, 2)
actions = np.arange(0, 25, 1)**2
rewards = np.linspace(-1, 2, 25)
terms = np.zeros(25)
self.memory = ReplayBuffer(self.memory_params)
for s, a, r, s2, d in zip(states, actions, rewards, next_states,
terms):
self.memory.append(s, a, r, s2, d)
self.memory.append_recent(s, d)
def test_combined_with_last(self):
s = np.array([500, 501])
s2 = np.array([600, 601])
a = 2
r = 2
d = False
self.memory.append(s, a, r, s2, d)
sample = self.memory.sample(10)
assert (sample[0][-1].cpu().detach().numpy() == s).all()
def test_len_recents_observation(self):
assert len(self.memory.recent_observations) == 1
def test_only_zeroed_observation(self):
s = np.array([500, 501])
d = True
self.memory.append_recent(s, d)
assert (self.memory.get_recent_states(np.array([77, 88])) == np.array(
[[0, 0], [77, 88]])).all()
class TestReplayBufferAtStart(unittest.TestCase):
def setUp(self):
self.memory_params = MemoryParams({"verbose": 0})
self.memory_params.window_length = 5
self.memory_params.memory_size = 1000
self.memory_params.seed = 123
self.memory_params.combined_with_last = False
self.memory = ReplayBuffer(self.memory_params)
def test_only_zeroed_observation(self):
assert (self.memory.get_recent_states(np.array([77, 88])) == np.array(
[[0, 0], [0, 0], [0, 0], [0, 0], [0, 0], [77, 88]])).all()
class TestReplayBufferMethods(unittest.TestCase):
def setUp(self):
self.memory_params = MemoryParams({"verbose": 0})
self.memory_params.window_length = 2
self.memory_params.memory_size = 1000
self.memory_params.seed = 123
self.memory_params.combined_with_last = False
states = np.arange(50).reshape(-1, 2)
next_states = np.arange(2, 52).reshape(-1, 2)
actions = np.arange(0, 25, 1)**2
rewards = np.linspace(-1, 2, 25)
terms = np.zeros(25)
self.memory = ReplayBuffer(self.memory_params)
for s, a, r, s2, d in zip(states, actions, rewards, next_states,
terms):
self.memory.append(s, a, r, s2, d)
self.memory.append_recent(s, d)
def test_length_memory(self):
self.assertEqual(25, len(self.memory))
def test_memory(self):
self.assertTrue((np.array([6,
7]) == self.memory.memory[3].state).all())
def test_get_historic_state(self):
pass
def test_same_sample(self):
sample = self.memory.sample(4)[1].cpu().detach().numpy().flatten()
self.assertEqual([1, 64, 4, 576], list(sample))
def test_sample_size(self):
self.assertEqual(4, len(self.memory.sample(4)[0]))
def test_sample_return_sarsd(self):
self.assertEqual(
5,
len(self.memory.sample(1)),
msg="Must return (state, action, reward, next_state, done) tuple",
)
def test_missing_params_memory(self):
with pytest.raises(TypeError):
ReplayBuffer()
def test_get_recent_states(self):
assert (self.memory.get_recent_states(np.array([77, 88])) == np.array(
[[46, 47], [48, 49], [77, 88]])).all()
def test_zeroed_observation(self):
s = np.array([500, 501])
d = True
self.memory.append_recent(s, d)
assert (self.memory.get_recent_states(np.array([77, 88])) == np.array(
[[0, 0], [0, 0], [77, 88]])).all()
def test_zeroed_observation_2(self):
s = np.array([500, 501])
self.memory.append_recent(s, True)
self.memory.append_recent(s, False)
assert (self.memory.get_recent_states(np.array([77, 88])) == np.array(
[[0, 0], [500, 501], [77, 88]])).all()
@mock.patch("core.memories.Memory.__init__")
def test_super_init(self, mock_super):
ReplayBuffer(self.memory_params)
self.assertTrue(mock_super.called)
def test_super_init(self, mock_super):
ReplayBuffer(self.memory_params)
self.assertTrue(mock_super.called)
def test_missing_params_memory(self):
with pytest.raises(TypeError):
ReplayBuffer()