1 vin.modules.heads

vin.modules.heads

Reusable candidate scoring heads for VIN architectures.

This module owns small, task-level prediction heads that sit after scene, trajectory, and pose encoders. The heads are intentionally separate from aria_nbv.vin.models so myopic, multi-step, and legacy seminar scorer variants can share ordinal prediction logic without importing from experimental modules.

1.1 Classes

Name	Description
VinScorerHead	Predict ordinal RRI logits from per-candidate feature vectors.
VinScorerHeadConfig	Configure the shared VIN ordinal scoring head.

1.1.1 VinScorerHead

vin.modules.heads.VinScorerHead(config, *, in_dim=None)

Predict ordinal RRI logits from per-candidate feature vectors.

VinScorerHead is the shared MLP-plus-CORAL head used when a VIN scorer represents candidate quality as ordinal Relative Reconstruction Improvement bins. It consumes the final feature tensor produced by an architecture and returns threshold logits compatible with aria_nbv.vin.ordinal.

The MLP and CORAL layer act on the last dimension only and introduce no candidate-to-candidate interaction or graph structure. Row permutations are exactly equivariant in evaluation mode (or with zero dropout); sampled training dropout preserves that symmetry only in distribution.

1.1.1.1 Methods

Name	Description
forward	Return CORAL threshold logits for candidate features.

1.1.1.1.1 forward

vin.modules.heads.VinScorerHead.forward(x)

Return CORAL threshold logits for candidate features.

1.1.1.1.1.1 Parameters

Name	Type	Description	Default
x	Tensor	Tensor["... F", float32] feature vectors, where leading dimensions are preserved as candidate or batch axes.	required

1.1.1.1.1.2 Returns

Name	Type	Description
	Tensor	Tensor["... K-1", float32] CORAL logits for K ordinal RRI classes.

1.1.2 VinScorerHeadConfig

vin.modules.heads.VinScorerHeadConfig()

Configure the shared VIN ordinal scoring head.

1.1.2.1 Attributes

Name	Description
target_type	Factory target for BaseConfig.setup_target.
hidden_dim	Hidden dimension for MLP layers.
num_layers	Number of MLP layers before the CORAL layer.
dropout	Dropout probability in the MLP.
num_classes	Number of ordinal bins (VIN-NBV uses 15).
coral_preinit_bias	Whether to initialize CORAL threshold biases with ordinal defaults.
activation	Activation function.

1.1.2.2 Methods

Name	Description
setup_target	Instantiate the configured scoring head.

1.1.2.2.1 setup_target

vin.modules.heads.VinScorerHeadConfig.setup_target(in_dim=None)

Instantiate the configured scoring head.

1.1.2.2.1.1 Parameters

Name	Type	Description	Default
in_dim	int | None	Optional input feature width passed to VinScorerHead.	None

1.1.2.2.1.2 Returns

Name	Type	Description
	VinScorerHead	Runtime scoring module.