def get_from_config(cfg):
env_id = cfg["env_id"]
if "bsuite" in env_id:
from Environments import BSuiteWrapper
env_name = env_id.split(" ")[-1]
env = BSuiteWrapper(env_name)
return GymEnvironment(env_id=None, env=env)
return GymEnvironment(cfg["env_id"])
def test_policy():
cfg = {"rng": np.random}
env = GymEnvironment("CartPole-v1")
input_shape = env.get_policy_input_shape()
output_shape = env.get_policy_output_shape()
policy_instructions = \
{
"layers": [64, 64],
"layer_functions": ['relu', 'relu'],
"layer_extras": ['bn', 'bn'],
"output_function": 'softmax',
"output_extras": 'bn',
}
policy = FeedForward(input_shape, output_shape,
PolicyActionParsers.random_sample, cfg)
policy.build_model(policy_instructions)
env.reset()
obs = env.get_random_obs()
while not env.needs_reset:
action = policy.get_action(obs)
obs, rew = env.step(action)
print("TOOK ACTION", action, "GOT REWARD", rew)
print("POLICY TEST COMPLETE")
def test_random_action():
env = GymEnvironment("CartPole-v1")
env.reset()
action = env.get_random_action()
print("GOT RANDOM ACTION", action)
env.step(action)
print("SUCCESSFULLY STEPPED ENV WITH ACTION", action)
def test_build():
env = GymEnvironment("CartPole-v1")
print("BUILT CartPole-v1", "OBSERAVTION SHAPE:", env.observation_shape,
"ACTION SHAPE:", env.action_shape)
def test_seed():
env = GymEnvironment("CartPole-v1")
env.test_seeding()
print("SEEDING TEST COMPLETE")
def test_render():
env = GymEnvironment("CartPole-v1")
env.reset()
while not env.needs_reset:
env.step(env.get_random_action())
env.render()
env.close()
print("RENDER TEST SUCCESSFUL!")
def test_get_batch():
env = GymEnvironment("CartPole-v1")
env.reset()
batch = env.get_random_batch(500)
print("GOT BATCH OF SHAPE", batch.shape,
"EXPECTED SHAPE WITH 500 ENTRIES OF", env.observation_shape)