def pred_step(
self,
x: tp.Any,
states: types.States,
initializing: bool,
training: bool,
) -> PredStep:
raise types.MissingMethod()
def init_step(
self,
x: tp.Any,
y_true: tp.Any,
sample_weight: tp.Optional[np.ndarray],
class_weight: tp.Optional[np.ndarray],
states: types.States,
) -> types.States:
raise types.MissingMethod()
def train_step(
self,
x: tp.Any,
y_true: tp.Any,
sample_weight: tp.Optional[np.ndarray],
class_weight: tp.Optional[np.ndarray],
states: types.States,
initializing: bool,
training: bool,
) -> TrainStep:
raise types.MissingMethod()
def summary_step(
self,
x: tp.Any,
states: types.States,
) -> tp.List[types.SummaryTableEntry]:
raise types.MissingMethod()
def train_step(self, *args, **kwargs):
raise types.MissingMethod()
def reset_metrics(self) -> None:
raise types.MissingMethod()
def train_step(
self: M,
inputs: tp.Any,
labels: tp.Mapping[str, tp.Any],
) -> TrainStepOutput[M]:
raise types.MissingMethod()
def grad_step(
self: M,
inputs: tp.Any,
labels: tp.Mapping[str, tp.Any],
) -> GradStepOutput[M]:
raise types.MissingMethod()
def pred_step(
self: M,
inputs: tp.Any,
) -> PredStepOutput[M]:
raise types.MissingMethod()
def init_step(
self: M,
key: jnp.ndarray,
inputs: tp.Any,
) -> M:
raise types.MissingMethod()
