LION.utils.parameter

Classes

LIONParameter(**kwargs)	LION's parameters: a plain namespace of settings (p.n_iters = 10), used for every model's, solver's, experiment's and reconstructor's settings.
OperatorParameters(**kwargs)	The parameters of an operator, which make it again: one class per kind of operator.

Functions

resolve_reconstruction(value, op[, name])	value, a reconstruction f(y, op); for "default", the operator's (op.default_reconstruction_fn(), e.g. fdk for CT).
with_current_names(value)	A record read from a file, with the old names of the operator's parameters (OLD_OPERATOR_NAMES, at any depth) as operator_params.

---

class LION.utils.parameter.LIONParameter(**kwargs)

Bases: object

LION’s parameters: a plain namespace of settings (p.n_iters = 10), used for every model’s, solver’s, experiment’s and reconstructor’s settings. Values can be other parameters (nested); save/load write and read them as JSON, and print shows them all.

The parameters given as keywords, LIONParameter(n_iters=10). A dict value becomes a LIONParameter, unless it is a spec (a dict with "_target_", see LION.utils.spec), which describes an object and stays a dict.

is_filled()

Check if the values are not None

serialize()

This only exists to allow LIONParameter() inside LIONParameter() otherwise, its equivalent to vars(self)

deserialize()

Produces a dict of LIONParameter()

save(fname)

Save LIONParameter into a JSON file

load(fname)

Load LIONParameter from JSON file, and initialize instance

unpack()

If the LIONParameter is made of other Parameters, unpack those.

class LION.utils.parameter.OperatorParameters(**kwargs)

Bases: LIONParameter

The parameters of an operator, which make it again: one class per kind of operator. For CT, it’s the Geometry (LION.operators.ct.ct_geometry): wherever LION asks for operator_params, a CT user gives their geometry.

Components that need an operator take its parameters (operator_params) and call make_operator(), so they work with any kind of operator.

The parameters given as keywords, LIONParameter(n_iters=10). A dict value becomes a LIONParameter, unless it is a spec (a dict with "_target_", see LION.utils.spec), which describes an object and stays a dict.

make_operator()

The operator these parameters describe.

property split_count

How many measurement units the operator can be split along (for CT, the angles).

split(indices)

New parameters for the measurement units indices only (for CT, those angles). These parameters are not changed.

split_measurements(y, indices)

The measurement units indices of y (batched or not): the measurements of the operator split(indices) makes.

LION.utils.parameter.resolve_reconstruction(value, op, name='reconstruction')

value, a reconstruction f(y, op); for "default", the operator’s (op.default_reconstruction_fn(), e.g. fdk for CT).

LION.utils.parameter.with_current_names(value)

A record read from a file, with the old names of the operator’s parameters (OLD_OPERATOR_NAMES, at any depth) as operator_params.