def get_data_from_offline_batch(params,
env,
normalization_scope=None,
model='dynamics',
split_ratio=0.666667):
train_collection = DataCollection(batch_size=params[model]['batch_size'],
max_size=params['max_train_data'],
shuffle=True)
val_collection = DataCollection(batch_size=params[model]['batch_size'],
max_size=params['max_val_data'],
shuffle=False)
rollout_sampler = RolloutSampler(env)
rl_paths = rollout_sampler.generate_offline_data(
data_file=params['data_file'], n_train=params["n_train"])
path_collection = PathCollection()
obs_dim = env.observation_space.shape[0]
normalization = add_path_data_to_collection_and_update_normalization(
rl_paths,
path_collection,
train_collection,
val_collection,
normalization=None,
split_ratio=split_ratio,
obs_dim=obs_dim,
normalization_scope=normalization_scope)
return train_collection, val_collection, normalization, path_collection, rollout_sampler
def replace_path_data_to_collection_and_update_normalization(
paths,
train_collection,
val_collection,
normalization=None,
split_ratio=0.666667,
obs_dim=None,
normalization_scope=None):
# data
train_data, val_data = PathCollection.to_data_collections(
paths, split_ratio=split_ratio)
train_collection.replace_data(train_data)
val_collection.replace_data(val_data)
# normalization
if not normalization:
logger.log("Creating normalization for training data.")
normalization = Normalization(train_data,
obs_dim=obs_dim,
scope=normalization_scope)
logger.log("Done creating normalization for training data.")
return normalization
else:
logger.log("Updating normalization.")
normalization.update(train_collection.get_data())
logger.log("Done updating normalization.")
return normalization
def add_path_data_to_collection_and_update_normalization(
paths,
path_collection,
train_collection,
val_collection,
normalization=None,
split_ratio=0.666667,
train_discard_ratio=0.0,
obs_dim=None,
normalization_scope=None):
"""
Add new data from paths to collections. Update normalization stats.
:param path_collection: PathCollection object
:param train_collection: a "data_collection" object
:param val_collection: a "data_collection" object
:param normalization: a "Normalization" object
:param split_ratio: real number in [0, 1]. The split ratio between training and validation set
:param train_discard_ratio: real number in [0, 1). The ratio to discard training data
:param obs_dim: actual observation dimension that will be normalized
:return: an updated normalization
"""
# data
train_data, val_data = PathCollection.to_data_collections(
paths, split_ratio=split_ratio)
train_collection.add_data(train_data, discard_ratio=train_discard_ratio)
val_collection.add_data(val_data)
# normalization
if not normalization:
logger.log("Creating normalization for training data.")
normalization = Normalization(train_data,
obs_dim=obs_dim,
scope=normalization_scope)
logger.log("Done creating normalization for training data.")
return normalization
else:
logger.log("Updating normalization.")
normalization.update(train_collection.get_data())
logger.log("Done updating normalization.")
return normalization
def get_data_from_random_rollouts(params, env, normalization_scope=None):
train_collection = DataCollection(
batch_size=params['dynamics']['batch_size'],
max_size=params['max_train_data'],
shuffle=True)
val_collection = DataCollection(
batch_size=params['dynamics']['batch_size'],
max_size=params['max_val_data'],
shuffle=False)
rollout_sampler = RolloutSampler(env)
random_paths = rollout_sampler.generate_random_rollouts(
num_paths=params['num_path_random'], horizon=params['env_horizon'])
path_collection = PathCollection()
obs_dim = env.observation_space.shape[0]
normalization = add_path_data_to_collection_and_update_normalization(
random_paths,
path_collection,
train_collection,
val_collection,
normalization=None,
obs_dim=obs_dim,
normalization_scope=normalization_scope)
return train_collection, val_collection, normalization, path_collection, rollout_sampler
def get_data_from_random_rollouts(params,
env,
random_paths,
normalization_scope=None,
model='dynamics',
split_ratio=0.666667):
train_collection = DataCollection(batch_size=params[model]['batch_size'],
max_size=params['max_train_data'],
shuffle=True)
val_collection = DataCollection(batch_size=params[model]['batch_size'],
max_size=params['max_val_data'],
shuffle=False)
path_collection = PathCollection()
obs_dim = env.observation_space.shape[0]
normalization = add_path_data_to_collection_and_update_normalization(
random_paths,
path_collection,
train_collection,
val_collection,
normalization=None,
split_ratio=split_ratio,
obs_dim=obs_dim,
normalization_scope=normalization_scope)
return train_collection, val_collection, normalization, path_collection