Skip to content

SaveAt

SaveAt sets what a solve saves and when, passed as saveat= on ds.pdesolve. See solutions and output.

diff_soup.SaveAt dataclass

Determines which times and quantities a solve saves.

Pass it as saveat= to pdesolve. With nothing requested the solve saves the final state only. Buffers keep a fixed size, so unfilled rows read NaN.

Parameters:

Name Type Description Default
t0 bool

If True, save at the start of the interval.

False
t1 bool

If True, save at the end of the interval.

False
ts Optional[Array]

Times inside the interval, sorted and deduplicated before the solve, and not combined with steps.

None
steps bool | int

True saves every accepted step and an integer k every k-th, into a buffer sized from max_steps.

False
dense bool

If True, record the interpolant for Solution.evaluate.

False
fn Optional[Callable[..., Any]]

Not supported, and must stay None.

None
subs Any

Not supported, and must stay None.

None
solver_state bool

Not supported, and must stay False.

False
controller_state bool

Not supported, and must stay False.

False
made_jump bool

Not supported, and must stay False.

False
output Optional[DenseOutputPolicy]

A DenseOutputPolicy for the algebraic rows of interpolated samples.

None
ordered bool

If True, use ts as given, strictly increasing and possibly traced, without t0, t1, or steps.

False
outputs Optional[Sequence[str]]

Names to store in place of the state, each a declared observation, a declared field, "state", or "lambdas".

None

diff_soup.timestepping.SubSaveAt dataclass

A Diffrax-style save request for one output, which no solver accepts.

SaveAt(subs=...) raises ConfigurationError. Name several stored quantities with SaveAt(outputs=...) instead.

diff_soup.DenseOutputPolicy dataclass

How the algebraic rows of interpolated samples are computed.

Pass it as SaveAt(output=...) or Event(state_policy=...).

Parameters:

Name Type Description Default
mode str

"fast" interpolates every row, and "projected" projects each sample's algebraic rows onto the constraints.

'fast'