starling.inference.constraints.MultiConstraint

class MultiConstraint

Bases: Constraint

Combines multiple constraints into a single optimization step.

Methods

__init__
apply	Apply the constraint to the given latents.
bell_shaped_schedule	Bell-shaped schedule for time-dependent guidance strength.
compute_loss	Compute weighted combination of all constraint losses.
cosine_weight	Cosine schedule for time-dependent guidance strength.
get_adaptive_clip_threshold	Get an adaptive clipping threshold that follows a cosine schedule.
get_time_scale	Get the time-dependent scaling factor.
initialize	Initialize all constraints with the model parameters.
should_apply_guidance	Check if guidance should be applied at the current timestep.

__init__(constraints, schedule='cosine', verbose=True)

Parameters:

• constraints (list) – List of constraint objects to combine

• constraint_weights (list, optional) – Relative weights for each constraint (defaults to equal weights)

• guidance_scale (float) – Overall guidance scale for the combined constraint

• schedule (str) – Time schedule for constraint application (“cosine” or “linear”)

• verbose (bool) – Whether to print debug info

initialize(encoder_model, latent_space_scaling_factor, n_steps, sequence_length)

Initialize all constraints with the model parameters.

compute_loss(distance_maps: Tensor) → Tuple[Tensor, Tensor]

Compute weighted combination of all constraint losses.