Block-forecast models

Module contract

CausalBlockForecastModel predicts native outputs through causal block forecasting. The top-level AxoSimGRU alias points to this implementation. Its backbone and BlockForecastConfig constructor contract differs from the named GRU-profile factory.

Source revision: 306a51ed950b. Public export index.

BlockForecastConfig

Source

Configuration for causal next-block trajectory prediction.

BlockForecastConfig(core_kind: BlockForecastCore, patch_size: int, max_patch_rank: int = 8, max_trajectory_rank: int = 8, keep_local_tcn: bool = False, gru_hidden_units: int | None = None, residual_scale_init: float = 0.1, voltage_anchor_count: int | None = None, event_template: tuple[float, ...] | None = None, event_template_center_init: float = 0.9208316802978516, event_template_temperature: float = 0.25, event_template_hard: bool = False, behavior_adaptation: bool = False) -> None

Fields

Parameter	Type	Default
core_kind	BlockForecastCore	required
patch_size	int	required
max_patch_rank	int	8
max_trajectory_rank	int	8
keep_local_tcn	bool	False
gru_hidden_units	int | None	None
residual_scale_init	float	0.1
voltage_anchor_count	int | None	None
event_template	tuple[float, ...] | None	None
event_template_center_init	float	0.9208316802978516
event_template_temperature	float	0.25
event_template_hard	bool	False
behavior_adaptation	bool	False

patch_size: Native timesteps represented by one block. behavior_adaptation: Enable the learned behavior adaptation bank.

CausalBlockForecastModel

Source

Predict each native-resolution output block from prior input blocks.

Bases: nn.Module.

CausalBlockForecastModel.init

Source

__init__(self, source: nn.Module, forecast_config: BlockForecastConfig) -> None

Parameter	Type	Default
source	nn.Module	required
forecast_config	BlockForecastConfig	required

Returns None.

CausalBlockForecastModel.config

Source

CausalBlockForecastModel.config

Read-only property. Access as instance.config; do not call it as a function.

CausalBlockForecastModel.num_input

Source

CausalBlockForecastModel.num_input: int

Read-only property. Access as instance.num_input; do not call it as a function.

Returns int.

CausalBlockForecastModel.num_output

Source

CausalBlockForecastModel.num_output: int

Read-only property. Access as instance.num_output; do not call it as a function.

Returns int.

CausalBlockForecastModel.num_branch

Source

CausalBlockForecastModel.num_branch: int

Read-only property. Access as instance.num_branch; do not call it as a function.

Returns int.

CausalBlockForecastModel.patch_size

Source

CausalBlockForecastModel.patch_size: int

Read-only property. Access as instance.patch_size; do not call it as a function.

Returns int.

CausalBlockForecastModel.sparse_voltage_output

Source

CausalBlockForecastModel.sparse_voltage_output: bool

Read-only property. Access as instance.sparse_voltage_output; do not call it as a function.

Returns bool.

CausalBlockForecastModel.behavior_parameter_count

Source

CausalBlockForecastModel.behavior_parameter_count: int

Read-only property. Access as instance.behavior_parameter_count; do not call it as a function.

Returns int.

CausalBlockForecastModel.block_state_sequence

Source

block_state_sequence(self, x: torch.Tensor, *, morphology_indices: torch.Tensor | None=None, behavior_parameters: torch.Tensor | None=None) -> torch.Tensor

Encode a native sequence and return each causal macro state.

Parameter	Type	Default
x	torch.Tensor	required
morphology_indices	torch.Tensor | None	None
behavior_parameters	torch.Tensor | None	None

Returns torch.Tensor.

CausalBlockForecastModel.decode_block_state_sequence

Source

decode_block_state_sequence(self, block_state: torch.Tensor, *, behavior_parameters: torch.Tensor | None=None) -> torch.Tensor

Decode dense native-rate trajectories from causal macro states.

Parameter	Type	Default
block_state	torch.Tensor	required
behavior_parameters	torch.Tensor | None	None

Returns torch.Tensor.

CausalBlockForecastModel.forward

Source

forward(self, x: torch.Tensor, *, morphology_indices: torch.Tensor | None=None, behavior_parameters: torch.Tensor | None=None) -> torch.Tensor

Parameter	Type	Default
x	torch.Tensor	required
morphology_indices	torch.Tensor | None	None
behavior_parameters	torch.Tensor | None	None

Returns torch.Tensor.

CausalBlockForecastModel.component_manifest

Source

component_manifest(self) -> dict[str, object]

Returns dict[str, object].