def __init__(self, env, model, nsteps=5, gamma=0.99):
super().__init__(env=env, model=model, nsteps=nsteps)
self.gamma = gamma
self.batch_action_shape = [
x if x is not None else -1
for x in model.train_model.action.shape.as_list()
]
self.ob_dtype = model.train_model.X.dtype.as_numpy_dtype
if USE_IMMITATION_ENV:
print("a2c runner")
self.featurizer = TDCFeaturizer(92,
92,
84,
84,
feature_vector_size=1024,
learning_rate=0,
experiment_name='PE')
self.featurizer.load("zelda")
video_dataset = generate_dataset('zelda',
framerate=60,
width=84,
height=84)[0]
self.featurized_dataset = self.featurizer.featurize(video_dataset)
self.checkpoint_indexes = None #[0] * nenvs
self.rewards = 0
self.counter = 0
def __init__(self, *args, **kwargs):
print("MontezumaImmitationEnv")
print(args)
print(kwargs)
featurizer = TDCFeaturizer(92,
92,
84,
84,
feature_vector_size=1024,
learning_rate=0,
experiment_name='default')
featurizer.load()
video_dataset = generate_dataset('default',
framerate=30 / 15,
width=84,
height=84)[0]
featurized_dataset = featurizer.featurize(video_dataset)
self._env = ImmitationWrapper(
gym.make('MontezumaRevengeNoFrameskip-v4'),
featurizer=featurizer,
featurized_dataset=featurized_dataset)
self.observation_space = self._env.unwrapped.observation_space
self.action_space = self._env.unwrapped.action_space
self.np_random = self._env.unwrapped.np_random
self.ale = self._env.unwrapped.ale
def __init__(self, env_fns, spaces=None):
"""
envs: list of gym environments to run in subprocesses
"""
self.waiting = False
self.closed = False
nenvs = len(env_fns)
self.nenvs = nenvs
self.remotes, self.work_remotes = zip(*[Pipe() for _ in range(nenvs)])
self.ps = [Process(target=worker, args=(work_remote, remote, CloudpickleWrapper(env_fn)))
for (work_remote, remote, env_fn) in zip(self.work_remotes, self.remotes, env_fns)]
for p in self.ps:
p.daemon = True # if the main process crashes, we should not cause things to hang
p.start()
for remote in self.work_remotes:
remote.close()
self.remotes[0].send(('get_spaces', None))
observation_space, action_space = self.remotes[0].recv()
self.viewer = None
VecEnv.__init__(self, len(env_fns), observation_space, action_space)
if USE_IMMITATION_ENV:
self.featurizer = TDCFeaturizer(92, 92, 84, 84, feature_vector_size=1024, learning_rate=0, experiment_name='default')
self.featurizer.load()
video_dataset = generate_dataset('default', framerate=30/15, width=84, height=84)[0]
self.featurized_dataset = self.featurizer.featurize(video_dataset)
self.checkpoint_indexes = [0] * nenvs
self.rewards = 0
self.counter = 0
def visualize_features(videos_path, featurizer_type, initial_width,
initial_height, desired_width, desired_height, costar):
if featurizer_type == 'vae':
featurizer_class = VAEFeaturizer
elif featurizer_type == 'tdc':
featurizer_class = TDCFeaturizer
elif featurizer_type == 'forward_model':
featurizer_class = ForwardModelFeaturizer
else:
raise TypeError
featurizer = featurizer_class(initial_width,
initial_height,
desired_width,
desired_height,
costar,
feature_vector_size=1024,
learning_rate=0.0001,
experiment_name='default')
d = generate_dataset(videos_path, 10, initial_width, initial_height,
costar)
features1 = featurizer.featurize(d[0])
features2 = featurizer.featurize(d[1])
#features3 = featurizer.featurize(d[2])
features_all = [features1, features2] #, features3]
visualize_embeddings(features_all)
def __init__(self,
env_fns,
spaces=None,
level_selector=None,
experiment=None,
dataset=None):
"""
Arguments:
env_fns: iterable of callables functions that build environments
"""
self.envs = [fn() for fn in env_fns]
self.num_envs = len(env_fns)
env = self.envs[0]
VecEnv.__init__(self, len(env_fns), env.observation_space,
env.action_space)
obs_space = env.observation_space
self.keys, shapes, dtypes = obs_space_info(obs_space)
self.buf_obs = {
k: np.zeros((self.num_envs, ) + tuple(shapes[k]), dtype=dtypes[k])
for k in self.keys
}
self.buf_dones = np.zeros((self.num_envs, ), dtype=np.bool)
self.buf_rews = np.zeros((self.num_envs, ), dtype=np.float32)
self.buf_infos = [{} for _ in range(self.num_envs)]
self.actions = None
self.spec = self.envs[0].spec
self.finsihed = [False for _ in range(self.num_envs)]
self.last_mes = [None for _ in range(self.num_envs)]
self.level_selector = level_selector
if USE_IMMITATION_ENV:
self.featurizer = TDCFeaturizer(92,
92,
84,
84,
feature_vector_size=1024,
learning_rate=0,
experiment_name=experiment)
self.featurizer.load(dataset)
video_dataset = generate_dataset(experiment,
framerate=60,
width=84,
height=84)[0]
self.featurized_dataset = self.featurizer.featurize(video_dataset)
self.checkpoint_indexes = [0] * nenvs
self.rewards = 0
self.counter = 0
def __init__(self, *args, **kwargs):
print("ZeldaImmitationEnv")
featurizer = TDCFeaturizer(84,
84,
84,
84,
feature_vector_size=1024,
learning_rate=0,
experiment_name='PE')
featurizer.load("zelda")
video_dataset = generate_dataset('PE',
framerate=60,
width=84,
height=84)[0]
featurized_dataset = featurizer.featurize(video_dataset)
self._env = ImmitationWrapper(gym.make('gvgai-zelda-lvl0-v0'),
featurizer=featurizer,
featurized_dataset=featurized_dataset)
# self._env = gym.make('gvgai-zelda-lvl0-v0')
self.observation_space = self._env.unwrapped.observation_space
self.action_space = self._env.unwrapped.action_space
def __init__(self, *, env, model, nsteps, gamma, lam):
super().__init__(env=env, model=model, nsteps=nsteps)
self.lam = lam
self.gamma = gamma
if USE_IMMITATION_ENV:
self.featurizer = TDCFeaturizer(92,
92,
84,
84,
feature_vector_size=1024,
learning_rate=0,
experiment_name='default',
is_variational=False)
self.featurizer.load()
video_dataset = generate_dataset('default',
framerate=30 / 15,
width=84,
height=84)[0]
self.featurized_dataset = self.featurizer.featurize(video_dataset)
self.checkpoint_indexes = None #[0] * nenvs
self.rewards = 0
self.counter = 0
from train_featurizer import generate_dataset
from embedding_visualization import visualize_embeddings
from TDCFeaturizer import TDCFeaturizer
featurizer = TDCFeaturizer(92, 92, 84, 84)
featurizer.load('montezuma')
d = generate_dataset('montezuma.txt', .25, 84, 84)
features1 = featurizer.featurize(d[0])
features2 = featurizer.featurize(d[1])
features3 = featurizer.featurize(d[2])
features_all = [features1, features2, features3]
visualize_embeddings(features_all, experiment_name='montezuma')