def _protected_ProcessTransition(
self,
brain: base.Brain,
transition: base.Transition,
step_idx: int,
) -> None:
brain.UpdateFromTransitions([transition])
def _protected_ProcessTransition(
self,
brain: base.Brain,
transition: base.Transition,
step_idx: int,
) -> None:
"""Processes a new transition; e.g. to train the QFunction."""
brain.UpdateFromTransitions([transition])
def _protected_ProcessTransition(
self,
brain: base.Brain,
transition: base.Transition,
step_idx: int,
) -> None:
self._experience.AddTransition(transition)
if step_idx % self._train_every_n_steps == 0:
brain.UpdateFromTransitions(
self._experience.Sample(self._experience_sample_batch_size))
def _protected_ProcessTransition(
self,
brain: base.Brain,
transition: base.Transition,
step_idx: int,
) -> None:
"""Processes a new transition; e.g. to train the QFunction."""
self._memory.append(transition)
if len(self._memory) == self._batch_size:
brain.UpdateFromTransitions(self._memory)
self._memory = []
def _protected_ProcessTransition(
self,
brain: base.Brain,
transition: base.Transition,
step_idx: int,
) -> None:
train_transitions = []
self._memory.append(transition)
train_transitions.append(self._GetNStepTransition())
if len(self._memory) >= self._n_step_return:
self._memory.pop(0)
if transition.sp is None:
while self._memory:
train_transitions.append(self._GetNStepTransition())
self._memory.pop(0)
brain.UpdateFromTransitions(train_transitions)