Wave 5.1 Grouped Harmonic Heads Network

Embryonic Wave 5.1 grouped harmonic-heads TE model.

class scripts.models.wave3_grouped_harmonic_heads_network.Wave3GroupedHarmonicHeadsNetwork(input_size, output_size=1, harmonic_order=240, coefficient_mode='linear_conditioned', low_order_harmonic_index_list=None, stable_middle_harmonic_index_list=None, high_order_harmonic_index_list=None, residual_hidden_size=None, residual_activation_name='GELU', residual_dropout_probability=0.05, residual_use_layer_norm=True, low_order_scale=1.0, stable_middle_scale=1.0, high_order_scale=1.0, residual_scale=1.0, readout_position='center', freeze_harmonic_heads=False)

Bases: Module

Grouped harmonic heads plus residual correction skeleton.

Parameters:

• input_size (int)

• output_size (int)

• harmonic_order (int)

• coefficient_mode (str)

• low_order_harmonic_index_list (Sequence[int] | None)

• stable_middle_harmonic_index_list (Sequence[int] | None)

• high_order_harmonic_index_list (Sequence[int] | None)

• residual_hidden_size (list[int] | None)

• residual_activation_name (str)

• residual_dropout_probability (float)

• residual_use_layer_norm (bool)

• low_order_scale (float)

• stable_middle_scale (float)

• high_order_scale (float)

• residual_scale (float)

• readout_position (str)

• freeze_harmonic_heads (bool)

__init__(input_size, output_size=1, harmonic_order=240, coefficient_mode='linear_conditioned', low_order_harmonic_index_list=None, stable_middle_harmonic_index_list=None, high_order_harmonic_index_list=None, residual_hidden_size=None, residual_activation_name='GELU', residual_dropout_probability=0.05, residual_use_layer_norm=True, low_order_scale=1.0, stable_middle_scale=1.0, high_order_scale=1.0, residual_scale=1.0, readout_position='center', freeze_harmonic_heads=False)

Initialize the embryonic grouped-head Wave 5.1 model.

Parameters:

• input_size (int) – Input feature count, with angular position in the first feature column.

• output_size (int) – Regression target count. The skeleton supports the repository’s scalar TE target.

• harmonic_order (int) – Contiguous fallback harmonic order.

• coefficient_mode (str) – Harmonic branch coefficient mode.

• low_order_harmonic_index_list (Sequence[int] | None) – Low-order / offset harmonic group.

• stable_middle_harmonic_index_list (Sequence[int] | None) – Stable middle harmonic group.

• high_order_harmonic_index_list (Sequence[int] | None) – High-order fragile harmonic group.

• residual_hidden_size (list[int] | None) – Residual branch hidden widths.

• residual_activation_name (str) – Residual branch activation name.

• residual_dropout_probability (float) – Residual branch dropout.

• residual_use_layer_norm (bool) – Whether the residual branch uses layer normalization.

• low_order_scale (float) – Scale applied to the low-order harmonic head.

• stable_middle_scale (float) – Scale applied to the stable middle head.

• high_order_scale (float) – Scale applied to the high-order harmonic head.

• residual_scale (float) – Scale applied to the residual correction head.

• readout_position (str) – Readout position for sequence batches.

• freeze_harmonic_heads (bool) – Whether all harmonic heads are frozen.

Return type:

None

resolve_readout_tensor(input_tensor)

Return point-level features from point or sequence input.

Parameters:

input_tensor (Tensor)

Return type:

Tensor

compute_auxiliary_output_dictionary(input_tensor, normalized_input_tensor)

Expose grouped harmonic, residual, and combined prediction tensors.

Parameters:

• input_tensor (Tensor)

• normalized_input_tensor (Tensor)

Return type:

dict[str, Tensor]

forward_with_input_context(input_tensor, normalized_input_tensor)

Predict normalized TE using raw angle context and normalized features.

Parameters:

• input_tensor (Tensor)

• normalized_input_tensor (Tensor)

Return type:

Tensor

forward(normalized_input_tensor)

Run inference from normalized inputs when raw context is unavailable.

Parameters:

normalized_input_tensor (Tensor)

Return type:

Tensor