def add_sample(self, observation, action, reward, terminal, initial):
self.check_replacement()
self._observations[self._top] = observation
if self._is_action_discrete and not isinstance(action,
(int, np.int64)):
action = special.from_onehot(action)
self._actions[self._top] = action
self._rewards[self._top] = reward
self._terminals[self._top] = terminal
self._initials[self._top] = initial
self.advance()
def add_sample(self, observation, action, reward, terminal, initial):
"""
Add a sample to current replay buffer.
Parameters
----------
observation (np.array):
# TODO (ewei)
"""
self.check_replacement()
self._observations[self._top] = observation
if self._is_action_discrete and not isinstance(action,
(int, np.int64)):
action = special.from_onehot(action)
self._actions[self._top] = action
self._rewards[self._top] = reward
self._terminals[self._top] = terminal
self._initials[self._top] = initial
self.advance()
def unflatten(self, x):
return special.from_onehot(x)