Params¶

build_param_labels ¶

build_param_labels(spec)

Generate human-friendly labels for flattened parameters.

flat_vector_to_python(vector, keys, shapes)

Convert a flat parameter vector back to a python dict.

Parameters:

Name	Type	Description	Default
`vector`	`ndarray`	Flattened parameter vector.	required
`keys`	`Sequence[str]`	Ordered parameter names.	required
`shapes`	`Mapping[str, Sequence[int]]`	Target shapes keyed by parameter name.	required

Returns:

Type	Description
`Dict[str, object]`	Mapping from parameter name to Python scalars/lists.

flatten_bounds(keys, bounds, shapes)

Flatten per-parameter bounds into two aligned vectors.

Parameters:

Name	Type	Description	Default
`keys`	`Sequence[str]`	Parameter names.	required
`bounds`	`Sequence[Sequence[object]]`	Lower/upper bounds for each parameter.	required
`shapes`	`Mapping[str, Sequence[int]]`	Target shapes keyed by parameter name.	required

Returns:

Type	Description
`Tuple[ndarray, ndarray]`	Concatenated lower and upper bounds.

infer_hidden_sizes_from_params(params_tree)

Infer dense layer widths from a flax parameter tree.

params_to_vector(param_obj, spec)

Flatten a parameter dataclass into a single vector.

Parameters:

Name	Type	Description	Default
`param_obj`	`object`	Dataclass containing parameter fields.	required
`spec`	`ModelSpec - like`	Object exposing `param_keys` and optional shapes.	required

Returns:

Type	Description
`ndarray`	Concatenated parameter vector.

shape_size(shape)

Return the total element count for a shape (0-dim -> 1).

shape_tuple(shape)

Normalize a possibly-None shape into a tuple of ints.