dodox/fancy_gym
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

naming convention and running tests

Browse Source
This commit is contained in:
Fabian
2022-07-11 16:18:18 +02:00
parent 786da2290d
commit ade83b5ae6
13 changed files with 621 additions and 590 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/test_bb_envs.py
+13 -12
View File
@@ -40,9 +40,9 @@ class TestMPEnvironments(unittest.TestCase):
for i in range(iterations):
observations.append(obs)
ac = env.action_space.sample()
actions = env.action_space.sample()
# ac = np.random.uniform(env.action_space.low, env.action_space.high, env.action_space.shape)
obs, reward, done, info = env.step(ac)
obs, reward, done, info = env.step(actions)
self._verify_observations(obs, env.observation_space, "step()")
self._verify_reward(reward)
@@ -55,13 +55,13 @@ class TestMPEnvironments(unittest.TestCase):
env.render("human")
if done:
obs = env.reset()
break
assert done, "Done flag is not True after max episode length."
assert done, "Done flag is not True after end of episode."
observations.append(obs)
env.close()
del env
return np.array(observations), np.array(rewards), np.array(dones)
return np.array(observations), np.array(rewards), np.array(dones), np.array(actions)
def _run_env_determinism(self, ids):
seed = 0
@@ -70,8 +70,9 @@ class TestMPEnvironments(unittest.TestCase):
traj1 = self._run_env(env_id, seed=seed)
traj2 = self._run_env(env_id, seed=seed)
for i, time_step in enumerate(zip(*traj1, *traj2)):
obs1, rwd1, done1, obs2, rwd2, done2 = time_step
self.assertTrue(np.allclose(obs1, obs2), f"Observations [{i}] {obs1} and {obs2} do not match.")
obs1, rwd1, done1, ac1, obs2, rwd2, done2, ac2 = time_step
self.assertTrue(np.array_equal(ac1, ac2), f"Actions [{i}] delta {ac1 - ac2} is not zero.")
self.assertTrue(np.array_equal(obs1, obs2), f"Observations [{i}] delta {obs1 - obs2} is not zero.")
self.assertEqual(rwd1, rwd2, f"Rewards [{i}] {rwd1} and {rwd2} do not match.")
self.assertEqual(done1, done2, f"Dones [{i}] {done1} and {done2} do not match.")
@@ -81,7 +82,7 @@ class TestMPEnvironments(unittest.TestCase):
f"not contained in observation space {observation_space}.")
def _verify_reward(self, reward):
self.assertIsInstance(reward, float, f"Returned {reward} as reward, expected float.")
self.assertIsInstance(reward, (float, int), f"Returned type {type(reward)} as reward, expected float or int.")
def _verify_done(self, done):
self.assertIsInstance(done, bool, f"Returned {done} as done flag, expected bool.")
@@ -113,12 +114,12 @@ class TestMPEnvironments(unittest.TestCase):
def test_dmc_environment_functionality(self):
"""Tests that environments runs without errors using random actions for DMC MP envs."""
with self.subTest(msg="DMP"):
for env_id in alr_envs.ALL_DEEPMIND_MOTION_PRIMITIVE_ENVIRONMENTS['DMP']:
for env_id in alr_envs.ALL_DMC_MOVEMENT_PRIMITIVE_ENVIRONMENTS['DMP']:
with self.subTest(msg=env_id):
self._run_env(env_id)
with self.subTest(msg="ProMP"):
for env_id in alr_envs.ALL_DEEPMIND_MOTION_PRIMITIVE_ENVIRONMENTS['ProMP']:
for env_id in alr_envs.ALL_DMC_MOVEMENT_PRIMITIVE_ENVIRONMENTS['ProMP']:
with self.subTest(msg=env_id):
self._run_env(env_id)
@@ -151,9 +152,9 @@ class TestMPEnvironments(unittest.TestCase):
def test_dmc_environment_determinism(self):
"""Tests that identical seeds produce identical trajectories for DMC MP Envs."""
with self.subTest(msg="DMP"):
self._run_env_determinism(alr_envs.ALL_DEEPMIND_MOTION_PRIMITIVE_ENVIRONMENTS["DMP"])
self._run_env_determinism(alr_envs.ALL_DMC_MOVEMENT_PRIMITIVE_ENVIRONMENTS["DMP"])
with self.subTest(msg="ProMP"):
self._run_env_determinism(alr_envs.ALL_DEEPMIND_MOTION_PRIMITIVE_ENVIRONMENTS["ProMP"])
self._run_env_determinism(alr_envs.ALL_DMC_MOVEMENT_PRIMITIVE_ENVIRONMENTS["ProMP"])
def test_metaworld_environment_determinism(self):
"""Tests that identical seeds produce identical trajectories for Metaworld MP Envs."""
test/test_dmc_envs.py
+15 -11
View File
@@ -7,8 +7,8 @@ from dm_control import suite, manipulation
from alr_envs import make
DMC_ENVS = [f'{env}-{task}' for env, task in suite.ALL_TASKS if env != "lqr"]
MANIPULATION_SPECS = [f'manipulation-{task}' for task in manipulation.ALL if task.endswith('_features')]
DMC_ENVS = [f'dmc:{env}-{task}' for env, task in suite.ALL_TASKS if env != "lqr"]
MANIPULATION_SPECS = [f'dmc:manipulation-{task}' for task in manipulation.ALL if task.endswith('_features')]
SEED = 1
@@ -29,9 +29,11 @@ class TestStepDMCEnvironments(unittest.TestCase):
Returns:
"""
print(env_id)
env: gym.Env = make(env_id, seed=seed)
rewards = []
observations = []
actions = []
dones = []
obs = env.reset()
self._verify_observations(obs, env.observation_space, "reset()")
@@ -43,6 +45,7 @@ class TestStepDMCEnvironments(unittest.TestCase):
observations.append(obs)
ac = env.action_space.sample()
actions.append(ac)
# ac = np.random.uniform(env.action_space.low, env.action_space.high, env.action_space.shape)
obs, reward, done, info = env.step(ac)
@@ -57,13 +60,13 @@ class TestStepDMCEnvironments(unittest.TestCase):
env.render("human")
if done:
obs = env.reset()
break
assert done, "Done flag is not True after max episode length."
assert done, "Done flag is not True after end of episode."
observations.append(obs)
env.close()
del env
return np.array(observations), np.array(rewards), np.array(dones)
return np.array(observations), np.array(rewards), np.array(dones), np.array(actions)
def _verify_observations(self, obs, observation_space, obs_type="reset()"):
self.assertTrue(observation_space.contains(obs),
@@ -71,7 +74,7 @@ class TestStepDMCEnvironments(unittest.TestCase):
f"not contained in observation space {observation_space}.")
def _verify_reward(self, reward):
self.assertIsInstance(reward, float, f"Returned {reward} as reward, expected float.")
self.assertIsInstance(reward, (float, int), f"Returned type {type(reward)} as reward, expected float or int.")
def _verify_done(self, done):
self.assertIsInstance(done, bool, f"Returned {done} as done flag, expected bool.")
@@ -91,8 +94,9 @@ class TestStepDMCEnvironments(unittest.TestCase):
traj1 = self._run_env(env_id, seed=seed)
traj2 = self._run_env(env_id, seed=seed)
for i, time_step in enumerate(zip(*traj1, *traj2)):
obs1, rwd1, done1, obs2, rwd2, done2 = time_step
self.assertTrue(np.array_equal(obs1, obs2), f"Observations [{i}] {obs1} and {obs2} do not match.")
obs1, rwd1, done1, ac1, obs2, rwd2, done2, ac2 = time_step
self.assertTrue(np.array_equal(ac1, ac2), f"Actions [{i}] delta {ac1 - ac2} is not zero.")
self.assertTrue(np.array_equal(obs1, obs2), f"Observations [{i}] delta {obs1 - obs2} is not zero.")
self.assertEqual(rwd1, rwd2, f"Rewards [{i}] {rwd1} and {rwd2} do not match.")
self.assertEqual(done1, done2, f"Dones [{i}] {done1} and {done2} do not match.")
@@ -111,11 +115,11 @@ class TestStepDMCEnvironments(unittest.TestCase):
traj1 = self._run_env(env_id, seed=seed)
traj2 = self._run_env(env_id, seed=seed)
for i, time_step in enumerate(zip(*traj1, *traj2)):
obs1, rwd1, done1, obs2, rwd2, done2 = time_step
self.assertTrue(np.array_equal(obs1, obs2), f"Observations [{i}] {obs1} and {obs2} do not match.")
obs1, rwd1, done1, ac1, obs2, rwd2, done2, ac2 = time_step
self.assertTrue(np.array_equal(ac1, ac2), f"Actions [{i}] delta {ac1 - ac2} is not zero.")
self.assertTrue(np.array_equal(obs1, obs2), f"Observations [{i}] delta {obs1 - obs2} is not zero.")
self.assertEqual(rwd1, rwd2, f"Rewards [{i}] {rwd1} and {rwd2} do not match.")
self.assertEqual(done1, done2, f"Dones [{i}] {done1} and {done2} do not match.")
self.assertEqual(done1, done2, f"Dones [{i}] {done1} and {done2} do not match.")
if __name__ == '__main__':
test/test_metaworld_envs.py
+5 -5
View File
@@ -6,7 +6,7 @@ import numpy as np
from alr_envs import make
from metaworld.envs import ALL_V2_ENVIRONMENTS_GOAL_OBSERVABLE
ALL_ENVS = [env.split("-goal-observable")[0] for env, _ in ALL_V2_ENVIRONMENTS_GOAL_OBSERVABLE.items()]
ALL_ENVS = [f'metaworld:{env.split("-goal-observable")[0]}' for env, _ in ALL_V2_ENVIRONMENTS_GOAL_OBSERVABLE.items()]
SEED = 1
@@ -57,9 +57,9 @@ class TestStepMetaWorlEnvironments(unittest.TestCase):
env.render("human")
if done:
obs = env.reset()
break
assert done, "Done flag is not True after max episode length."
assert done, "Done flag is not True after end of episode."
observations.append(obs)
env.close()
del env
@@ -71,7 +71,7 @@ class TestStepMetaWorlEnvironments(unittest.TestCase):
f"not contained in observation space {observation_space}.")
def _verify_reward(self, reward):
self.assertIsInstance(reward, float, f"Returned {reward} as reward, expected float.")
self.assertIsInstance(reward, (float, int), f"Returned type {type(reward)} as reward, expected float or int.")
def _verify_done(self, done):
self.assertIsInstance(done, bool, f"Returned {done} as done flag, expected bool.")
@@ -94,7 +94,7 @@ class TestStepMetaWorlEnvironments(unittest.TestCase):
obs1, rwd1, done1, ac1, obs2, rwd2, done2, ac2 = time_step
self.assertTrue(np.array_equal(ac1, ac2), f"Actions [{i}] delta {ac1 - ac2} is not zero.")
self.assertTrue(np.array_equal(obs1, obs2), f"Observations [{i}] delta {obs1 - obs2} is not zero.")
self.assertAlmostEqual(rwd1, rwd2, f"Rewards [{i}] {rwd1} and {rwd2} do not match.")
self.assertEqual(rwd1, rwd2, f"Rewards [{i}] {rwd1} and {rwd2} do not match.")
self.assertEqual(done1, done2, f"Dones [{i}] {done1} and {done2} do not match.")