steps

Training steps that dispatch on the active Keras backend.

Machine-readable model

• NotSupportedclassThe active Keras backend is not supported by this trainer.
• require_backendfunctionReturn the active backend, or raise NotSupported naming feature and supported.
• no_gradfunctionDisable gradient tracking on Torch; a no-op on other backends.
• gradient_stepfunctionDifferentiate lossfn with the active backend and apply model.optimizer once.

class

NotSupported

Python

The active Keras backend is not supported by this trainer.

helia_edge/trainers/steps.py:12

NotSupported()

The active Keras backend is not supported by this trainer.

function

require_backend

Python

Return the active backend, or raise NotSupported naming feature and supported.

helia_edge/trainers/steps.py:16

require_backend(feature: str, supported: Sequence[str]) -> str

Return the active backend, or raise NotSupported naming feature and supported.

function

no_grad

Python

Disable gradient tracking on Torch; a no-op on other backends.

helia_edge/trainers/steps.py:24

no_grad() -> Iterator[None]

Disable gradient tracking on Torch; a no-op on other backends.

function

gradient_step

Python

Differentiate lossfn with the active backend and apply model.optimizer once.

helia_edge/trainers/steps.py:37

gradient_step(
    model: keras.Model,
    loss_fn: Callable[[], tuple[Any, ...]],
    variables: Sequence[Any] | None = None,
) -> tuple[Any, ...]

Differentiate loss_fn with the active backend and apply model.optimizer once.

Parameters

Parameters of gradient_step

Name	Type	Default	Description
model	keras.Model	Required	Compiled model whose optimizer applies the update.
loss_fn	Callable[[], tuple[Any, ...]]	Required	Computes ``(loss, *outputs)`` from the model's current weights.
variables	Sequence[Any] | None	None	Variables to update; when None, the model's trainable weights after ``loss_fn`` runs, so variables created by a first (building) call are included. Variables without a gradient are skipped.

Returns

Returns of gradient_step

Value	Type	Description
tuple	tuple[Any, ...]	What ``loss_fn`` returned.

Raises

Errors raised by gradient_step

Type	Description
NotSupported	On backends other than TensorFlow and Torch.