flow.residual_history

flow.residual_history(step_fn, theta, f0, options)

The same march, reporting the residual after every chunk.

Runs the full max_steps without early exit so the history has a fixed length and the whole thing stays one compiled computation. This is a diagnostic, not the path gradients are taken through.

Parameters

Name	Type	Description	Default
step_fn	Callable	(f, theta) -> f.	required
theta	Any	The differentiable parameters.	required
f0	jax.Array	Initial populations.	required
options	SteadyOptions	Convergence settings.	required

Returns

Name	Type	Description
	jax.Array	(f, history) – the final populations and the (chunks,)
	jax.Array	per-step relative change recorded after each chunk.