Options reference

This page lists every option of the typed configuration sections with its type and its default value. Each section links the guide that explains its options; start from Configuration guide. Unknown keys raise when a configuration loads (How configs are checked). Some effective defaults depend on other options, for example on the core type; the guides describe those cases.

The payload blocks model.core.population_network and model.core.unified_ei are checked by their model builders and documented with their families (Options reference, Unified E/I core). Sections that plugins provide are described in Configuration sections.

Top level

Explained in Configuration guide. Class dendritic_modeling.config.config.Config.

Option

Type

Default

config_version

int

4

experiment

ExperimentConfig

experiment

data

DataConfig

data

model

ModelConfig

model

training

TrainingConfig

training

analysis

AnalysisConfig

analysis

outputs

OutputsConfig

outputs

distributed

DistributedConfig

distributed

metadata

dict[str, Any]

{}

experiment

Explained in Random seeds and reproducibility flags and Training, validation and test splits. Class dendritic_modeling.config.config.ExperimentConfig.

Option

Type

Default

seed

int

42

dataset_seed

Optional[int]

null

split_seed

Optional[int]

null

model_seed

Optional[int]

null

topology_seed

Optional[int]

null

loader_seed

Optional[int]

null

evaluation_seed

Optional[int]

null

probe_seed

Optional[int]

null

strict_deterministic

bool

false

sealed_test

bool

false

allow_test_data

bool

true

max_test_samples

Optional[int]

null

record_dataset_fingerprints

bool

false

train_valid_split

float

0.8

fast_mode

bool

false

deterministic

Optional[bool]

null

cudnn_benchmark

Optional[bool]

null

allow_tf32

Optional[bool]

null

float32_matmul_precision

Optional[str]

null

checkpointing

CheckpointingConfig

experiment.checkpointing

data

Explained in Datasets. Class dendritic_modeling.config.config.DataConfig.

Option

Type

Default

dataset_name

str

cifar10

base_dir

str

''

processing

DataProcessingConfig

data.processing

dataset_params

dict[str, dict[str, Any]]

{}

model

Explained in Model architecture. Class dendritic_modeling.config.model.ModelConfig.

Option

Type

Default

task

str

classification

initial_checkpoint

Optional[str]

null

initial_checkpoint_sha256

Optional[str]

null

initial_checkpoint_strict

bool

true

fixed_output_scale

float

1.0

output_scale_mode

Optional[str]

null

encoder

EncoderConfig

model.encoder

core

CoreConfig

model.core

decoder

DecoderConfig

model.decoder

vision_replacement

plugin section

disabled; Configuration sections

pretrained_replacement

plugin section

disabled; Configuration sections

transformer_replacement

TransformerReplacementConfig or plugin section

disabled; Configuration sections

training

Explained in Training. Class dendritic_modeling.config.training.TrainingConfig.

Option

Type

Default

main

MainTrainingConfig

training.main

encoder_pretraining

EncoderPretrainingConfig

training.encoder_pretraining

transformer_replacement

plugin section

disabled; Configuration sections

vision_replacement

plugin section

disabled; Configuration sections

analysis

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.AnalysisConfig.

Option

Type

Default

runtime

AnalysisRuntimeConfig

analysis.runtime

require_completion

bool

false

performance

PerformanceAnalysisConfig

analysis.performance

synapse_turnover

SynapseTurnoverAnalysisConfig

analysis.synapse_turnover

synapse_pruning

SynapsePruningAnalysisConfig

analysis.synapse_pruning

noise_perturbation

NoisePerturbationAnalysisConfig

analysis.noise_perturbation

single_layer_contribution

SingleLayerContributionAnalysisConfig

analysis.single_layer_contribution

ablation

AblationAnalysisConfig

analysis.ablation

path_matched_intervention

PathMatchedInterventionAnalysisConfig

analysis.path_matched_intervention

branch_local_sensitivity

BranchLocalSensitivityAnalysisConfig

analysis.branch_local_sensitivity

node_perturbation_credit

NodePerturbationCreditAnalysisConfig

analysis.node_perturbation_credit

local_inhibitory_population_intervention

LocalInhibitoryPopulationInterventionAnalysisConfig

analysis.local_inhibitory_population_intervention

path_matched_information

PathMatchedInformationAnalysisConfig

analysis.path_matched_information

weight

WeightAnalysisConfig

analysis.weight

compartment_statistics

CompartmentStatisticsAnalysisConfig

analysis.compartment_statistics

compartment_snr

CompartmentSNRAnalysisConfig

analysis.compartment_snr

multiplicative_gain

MultiplicativeGainAnalysisConfig

analysis.multiplicative_gain

gain_load_perturbation

GainLoadPerturbationAnalysisConfig

analysis.gain_load_perturbation

synaptic_activation

SynapticActivationAnalysisConfig

analysis.synaptic_activation

branch_activation

BranchActivationAnalysisConfig

analysis.branch_activation

local_rule_components

LocalRuleComponentAnalysisConfig

analysis.local_rule_components

information

InformationAnalysisConfig

analysis.information

representation_accessibility

RepresentationAccessibilityAnalysisConfig

analysis.representation_accessibility

source_tuning_support

SourceTuningSupportAnalysisConfig

analysis.source_tuning_support

input_region_intervention

InputRegionInterventionAnalysisConfig

analysis.input_region_intervention

receptive_field

ReceptiveFieldAnalysisConfig

analysis.receptive_field

reactivation_dynamics

ReactivationDynamicsAnalysisConfig

analysis.reactivation_dynamics

correlation

CorrelationAnalysisConfig

analysis.correlation

routing

RoutingAnalysisConfig

analysis.routing

inhibitory_specialization

InhibitorySpecializationAnalysisConfig

analysis.inhibitory_specialization

jacobian_spectrum

JacobianSpectrumAnalysisConfig

analysis.jacobian_spectrum

dendritic_timetraces

DendriticTimetracesAnalysisConfig

analysis.dendritic_timetraces

spike_train

SpikeTrainAnalysisConfig

analysis.spike_train

adversarial_robustness

AdversarialRobustnessAnalysisConfig

analysis.adversarial_robustness

extensions

plugin sections

{}; Configuration sections

outputs

Explained in Run directory layout. Class dendritic_modeling.config.config.OutputsConfig.

Option

Type

Default

run_name

str

ei_sweep

results_dir

str

results

exact_run_dir

bool

false

distributed

Explained in Distributed training. Class dendritic_modeling.config.config.DistributedConfig.

Option

Type

Default

mode

str

none

fsdp

FSDPConfig

distributed.fsdp

timeout_seconds

Optional[int]

null

experiment.checkpointing

Explained in Run directory layout. Class dendritic_modeling.config.config.CheckpointingConfig.

Option

Type

Default

enabled

bool

false

save_every_n_epochs

int

10

data.processing

Explained in Datasets. Class dendritic_modeling.config.config.DataProcessingConfig.

Option

Type

Default

flatten

Optional[bool]

null

normalize

Optional[bool]

null

label_noise_rate

float

0.0

model.encoder

Explained in Encoder and decoder types. Class dendritic_modeling.config.model.EncoderConfig.

Option

Type

Default

type

str

identity

params

EncoderParamsConfig

model.encoder.params

model.core

Explained in Core types. Class dendritic_modeling.config.model.CoreConfig.

Option

Type

Default

type

str

einet

affine_mlp

Optional[AffineMLPConfig]

None; model.core.affine_mlp

shared_projection_shunting

Optional[SharedProjectionShuntingConfig]

None; model.core.shared_projection_shunting

routed_ei

Optional[RoutedEIConfig]

None; model.core.routed_ei

feature_readout

Optional[FeatureReadoutConfig]

None; model.core.feature_readout

target_active_parameters

Optional[int]

null

biological_neuron

Optional[bool]

null: resolves to true for standard dendritic E/I cores; false for signed/MLP controls and other families (Biological constraints)

initialization_seed

Optional[int]

null

architecture

ArchitectureConfig

model.core.architecture

connectivity

ConnectivityConfig

model.core.connectivity

transfer

TransferFunctionConfig

model.core.transfer

morphology

MorphologyConfig

model.core.morphology

sparsity

SparsityConfig

model.core.sparsity

reactivation

ReactivationConfig

model.core.reactivation

blocklinear

BlockLinearConfig

model.core.blocklinear

implementation

ImplementationConfig

model.core.implementation

spatial

SpatialConfig

model.core.spatial

synapse_types

dict[str, Any]

{}

dynamics

dict[str, Any]

{}

dendritic_spikes

dict[str, Any]

{}

soma_feedback

dict[str, Any]

{}

unified_ei

dict[str, Any]

{}

recurrent_ei

dict[str, Any]

{}

baseline_rnn

dict[str, Any]

{}

heterogeneous_leak_ctrnn

dict[str, Any]

{}

legendre_memory

dict[str, Any]

{}

population_network

dict[str, Any]

{}

vision

dict[str, Any]

{}

model.decoder

Explained in Encoder and decoder types. Class dendritic_modeling.config.model.DecoderConfig.

Option

Type

Default

type

str

MLP

params

DecoderParamsConfig

model.decoder.params

training.main

Explained in Training. Class dendritic_modeling.config.training.MainTrainingConfig.

Option

Type

Default

strategy

str

standard

common

CommonTrainerConfig

training.main.common

optimizer

OptimizerConfig

training.main.optimizer

regularization

RegularizationConfig

training.main.regularization

pruning

PruningConfig

training.main.pruning

strategies

TrainingStrategiesConfig

training.main.strategies

learning_strategy_config

Optional[LocalRuleConfig]

None; training.main.learning_strategy_config

training.encoder_pretraining

Explained in Training. Class dendritic_modeling.config.training.EncoderPretrainingConfig.

Option

Type

Default

epochs

int

0

batch_size

int

256

loss_function

str

mse

optimizer

str

adam

learning_rate

float

0.001

weight_decay

float

0.0

early_stopping

bool

true

patience

int

10

grad_clip_value

Optional[float]

null

freeze_encoder

bool

true

analysis.runtime

Explained in When analyzers run. Class dendritic_modeling.config.analysis.AnalysisRuntimeConfig.

Option

Type

Default

training

EvaluationRuntimeConfig

analysis.runtime.training

final

EvaluationRuntimeConfig

analysis.runtime.final

analysis.performance

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.PerformanceAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

PerformanceAnalysisParams

analysis.performance.params

training_splits

list[str]

[train, validation, test]

final_splits

list[str]

[train, validation, test]

analysis.synapse_turnover

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SynapseTurnoverAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

true

params

SynapseTurnoverAnalysisParams

analysis.synapse_turnover.params

analysis.synapse_pruning

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SynapsePruningAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

SynapsePruningAnalysisParams

analysis.synapse_pruning.params

analysis.noise_perturbation

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.NoisePerturbationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

NoisePerturbationAnalysisParams

analysis.noise_perturbation.params

analysis.single_layer_contribution

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SingleLayerContributionAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

SingleLayerContributionAnalysisParams

analysis.single_layer_contribution.params

analysis.ablation

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.AblationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

AblationAnalysisParams

analysis.ablation.params

analysis.path_matched_intervention

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.PathMatchedInterventionAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

PathMatchedInterventionAnalysisParams

analysis.path_matched_intervention.params

analysis.branch_local_sensitivity

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.BranchLocalSensitivityAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

BranchLocalSensitivityAnalysisParams

analysis.branch_local_sensitivity.params

analysis.node_perturbation_credit

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.NodePerturbationCreditAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

NodePerturbationCreditAnalysisParams

analysis.node_perturbation_credit.params

analysis.local_inhibitory_population_intervention

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.LocalInhibitoryPopulationInterventionAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

LocalInhibitoryPopulationInterventionAnalysisParams

analysis.local_inhibitory_population_intervention.params

analysis.path_matched_information

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.PathMatchedInformationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

PathMatchedInformationAnalysisParams

analysis.path_matched_information.params

analysis.weight

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.WeightAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

WeightAnalysisParams

analysis.weight.params

analysis.compartment_statistics

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CompartmentStatisticsAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

CompartmentStatisticsAnalysisParams

analysis.compartment_statistics.params

analysis.compartment_snr

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CompartmentSNRAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

CompartmentSNRAnalysisParams

analysis.compartment_snr.params

analysis.multiplicative_gain

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.MultiplicativeGainAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

MultiplicativeGainAnalysisParams

analysis.multiplicative_gain.params

analysis.gain_load_perturbation

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.GainLoadPerturbationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

GainLoadPerturbationAnalysisParams

analysis.gain_load_perturbation.params

analysis.synaptic_activation

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SynapticActivationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

SynapticActivationAnalysisParams

analysis.synaptic_activation.params

analysis.branch_activation

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.BranchActivationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

BranchActivationAnalysisParams

analysis.branch_activation.params

analysis.local_rule_components

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.LocalRuleComponentAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

LocalRuleComponentAnalysisParams

analysis.local_rule_components.params

analysis.information

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

InformationAnalysisParams

analysis.information.params

analysis.representation_accessibility

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationAccessibilityAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

RepresentationAccessibilityAnalysisParams

analysis.representation_accessibility.params

analysis.source_tuning_support

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SourceTuningSupportAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

SourceTuningSupportAnalysisParams

analysis.source_tuning_support.params

analysis.input_region_intervention

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InputRegionInterventionAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

InputRegionInterventionAnalysisParams

analysis.input_region_intervention.params

analysis.receptive_field

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.ReceptiveFieldAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

ReceptiveFieldAnalysisParams

analysis.receptive_field.params

analysis.reactivation_dynamics

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.ReactivationDynamicsAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

true

params

ReactivationDynamicsAnalysisParams

analysis.reactivation_dynamics.params

analysis.correlation

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CorrelationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

CorrelationAnalysisParams

analysis.correlation.params

analysis.routing

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RoutingAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

analysis.inhibitory_specialization

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InhibitorySpecializationAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

InhibitorySpecializationAnalysisParams

analysis.inhibitory_specialization.params

analysis.jacobian_spectrum

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.JacobianSpectrumAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

JacobianSpectrumAnalysisParams

analysis.jacobian_spectrum.params

analysis.dendritic_timetraces

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.DendriticTimetracesAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

DendriticTimetracesAnalysisParams

analysis.dendritic_timetraces.params

analysis.spike_train

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SpikeTrainAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

SpikeTrainAnalysisParams

analysis.spike_train.params

analysis.adversarial_robustness

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.AdversarialRobustnessAnalysisConfig.

Option

Type

Default

enabled

bool

false

training

bool

false

params

AdversarialRobustnessAnalysisParams

analysis.adversarial_robustness.params

distributed.fsdp

Explained in Distributed training. Class dendritic_modeling.config.config.FSDPConfig.

Option

Type

Default

sharding_strategy

str

FULL_SHARD

mixed_precision

bool

true

cpu_offload

bool

false

auto_wrap_policy

str

none

min_num_params

int

1000000

limit_all_gathers

bool

true

forward_prefetch

bool

true

sync_module_states

bool

true

use_orig_params

bool

true

backward_prefetch

str

BACKWARD_PRE

gradient_checkpointing

bool

false

freeze_core

bool

false

cache_frozen_core_outputs

bool

false

cached_core_batch_size

Optional[int]

null

mixed_precision_dtype

str

bf16

param_dtype

Optional[str]

null

reduce_dtype

Optional[str]

null

buffer_dtype

Optional[str]

null

model.encoder.params

Explained in Encoder and decoder types. Class dendritic_modeling.config.model.EncoderParamsConfig.

Option

Type

Default

input_dim

Optional[int]

null

input_shape

list[int]

[3, 32, 32]

encoder_hidden_dims

list[int]

[128, 64]

latent_dim

Optional[int]

null

latent_shape

Optional[list[int]]

null

decoder_hidden_dims

list[int]

[64, 128]

activation

str

relu

router_mode

str

fixed

pathway_groups

list[list[int]]

[]

shared_indices

list[int]

[]

n_pathways

int

2

pathway_dim

Optional[int]

null

pathway_dims

list[int]

[]

router_activation

str

none

learned_router_temperature

float

1.0

learned_router_init_scale

float

4.0

model.core.affine_mlp

Explained in Baselines and comparison models. Class dendritic_modeling.config.affine_mlp.AffineMLPConfig.

Option

Type

Default

dimensions

list[int]

[]

mode

str

dense

initialization

str

kaiming_normal

fanin

list[int]

[]

connectivity_seeds

list[int]

[]

activation

str

relu

negative_slope

float

0.1

activate_output

bool

true

dropout

float | None

null

output_weight_scale

float

1.0

projection_backend

str

eager

workspace_mb

float

128.0

index_dtype

str

int32

model.core.shared_projection_shunting

Explained in Baselines and comparison models. Class dendritic_modeling.config.shared_projection.SharedProjectionShuntingConfig.

Option

Type

Default

input_dim

int

1

output_dim

int

1

supports

list[list[int]]

[[0]]

nodes

list[float]

[1.0]

epsilon

float

1.0e-08

dtype

str

float32

train_projection

bool

true

train_conductance

bool

true

train_output

bool

true

model.core.routed_ei

Explained in Baselines and comparison models. Class dendritic_modeling.config.routed_ei.RoutedEIConfig.

Option

Type

Default

features

int

1

branches

int

1

output_dim

int

1

mechanism

str

shunting

routing

str

learned

fixed_assignment

list[int]

[]

inhibitory_support

str

matched

temperature

float

1.0

hard_routing

bool

true

leak_conductance

float

1.0

epsilon

float

1.0e-08

weight_normalization_floor

float | None

null

route_init_std

float

0.05

weight_init_std

float

0.04

branch_center

float

0.0

branch_scale

float

1.0

branch_divisor

float

1.0

branch_activation

str

none

trainable_activation

bool

false

activation_slope_raw

float

0.5413249

activation_offset

float

0.0

model.core.feature_readout

Explained in Baselines and comparison models. Class dendritic_modeling.config.feature_readout.FeatureReadoutConfig.

Option

Type

Default

input_shape

list[int]

[]

readout

AffineMLPConfig

model.core.feature_readout.readout

normalization

str

none

groups

int | None

null

divisor_offset

float | None

null

normalization_epsilon

float

1.0e-05

normalization_affine

bool

true

model.core.architecture

Explained in Structured E/I cores. Class dendritic_modeling.config.model.ArchitectureConfig.

Option

Type

Default

excitatory_layer_sizes

list[int]

[20]

inhibitory_layer_sizes

list[int]

[]

excitatory_branch_factors

list[int]

[3, 3, 3, 3]

inhibitory_branch_factors

list[int]

[1]

input_projection_dims

list[int]

[]

inhibitory_network_type

str

dendritic

mlp_inhibitory_network_params

dict

mapping; keys below

mlp_inhibitory_network_params.hidden_dims

list

[200]

mlp_inhibitory_network_params.activation

str

relu

model.core.connectivity

Explained in Structured E/I cores. Class dendritic_modeling.config.model.ConnectivityConfig.

Option

Type

Default

ee_synapses_per_branch_per_layer

list[int]

[40]

ei_synapses_per_branch_per_layer

list[int]

[0]

ie_synapses_per_branch_per_layer

list[int]

[10]

ii_synapses_per_branch_per_layer

list[int]

[0]

rec_ee_synapses_per_branch

list[int]

[0]

rec_ie_synapses_per_branch

list[int]

[0]

rec_ei_synapses_per_branch

list[int]

[0]

rec_ii_synapses_per_branch

list[int]

[0]

structured

StructuredConnectivityConfig

model.core.connectivity.structured

model.core.transfer

Explained in TransferLayer. Class dendritic_modeling.config.model.TransferFunctionConfig.

Option

Type

Default

input_transform

Optional[str]

null

independent_pathways

bool

false

input_source_identity

str

shared

excitatory_dim

Optional[int]

null

inhibitory_dim

Optional[int]

null

split_strategy

str

random

split_seed

Optional[int]

null

output_activation

Optional[str]

null

model.core.morphology

Explained in Dendritic neurons. Class dendritic_modeling.config.model.MorphologyConfig.

Option

Type

Default

somatic_synapses

bool

false

use_shunting

bool

true

use_additive_normalization

bool

false

leak_conductance

float

1.0

learnable_leak

bool

false

signal_rounding

str

none

additive_mode

str

raw

additive_tangent_n0

Optional[float]

null

additive_tangent_t0

Optional[float]

null

additive_options

dict[str, Any]

{}

additive_tangent_n0_by_depth

list[float]

[]

additive_tangent_t0_by_depth

list[float]

[]

allow_self_recurrence

bool

true

weight_transform

str

softplus

dbl_init_method

str

analytical_expectation

model.core.sparsity

Explained in Sparse synapse types. Class dendritic_modeling.config.model.SparsityConfig.

Option

Type

Default

init_method

str

xavier_normal

noise_level

float

0.0

type

str

standard

weight_norm_order

Optional[int]

null

gamma

float

1.0

temperature

float

0.5

ultrafast

bool

true

gradient_scaling

str

none

variance_score

str

absolute

options

dict

{}

deepst

dict

mapping; keys below

deepst.excitatory_target_density

float

0.1

deepst.inhibitory_target_density

float

0.1

deepst.use_noise

bool

true

deepst.sigma

float

0.05

deepst.rewiring_mode

str

global

deepst.rewire_frequency

int

1

deepst.synapses_per_branch

null

deepst.freeze_excitatory_connectivity

bool

false

deepst.freeze_inhibitory_connectivity

bool

false

deepst.init_method

str

xavier_normal

deepst.weight_threshold

float

1.0e-06

deepst.credit_trace_decay

float

0.95

deepst.credit_candidate_pool_size

int

8

deepst.credit_turnover_fraction

float

0.1

deepst.credit_weak_active_pool_fraction

float

0.5

deepst.credit_min_observations

int

5

deepst.credit_swap_margin

float

0.0

deepst.credit_force_turnover

bool

false

deepst.credit_selection

str

credit

deepst.credit_warmup_steps

int

5

indexed

dict

mapping; keys below

indexed.candidate_size

null

indexed.selection

str

standard

indexed.seed

null

indexed.output_chunk_size

int

2048

indexed.index_dtype

str

int64

indexed.workspace_mb

null

indexed.cache_transformed_weights

bool

false

indexed.recompute_backward

bool

false

indexed.projection_backend

str

eager

indexed.persistent_indices

bool

true

indexed.init_mode

str

per_rank

indexed.rewire_frequency

int

100

indexed.rewire_quantile

float

0.05

indexed.rewire_until_step

null

dense_to_sparse

dict

mapping; keys below

dense_to_sparse.initial_density

float

1.0

dense_to_sparse.initial_k

null

dense_to_sparse.start_step

int

0

dense_to_sparse.end_step

int

1000

dense_to_sparse.update_interval

int

1

dense_to_sparse.schedule

str

cubic

dense_to_sparse.freeze_on_end

bool

false

dense_to_sparse.advance_on_forward

bool

true

dense_to_sparse.prune_metric

str

weight

model.core.reactivation

Explained in Gate initialization and maintenance. Class dendritic_modeling.config.model.ReactivationConfig.

Option

Type

Default

enabled

bool

true

type

str

param_tanh

soma_type

Optional[str]

null

soma_init_m

float

1.0

soma_init_b

float

0.0

dendritic_activation

Optional[str]

null

init_policy

str

analytical

init_m

float

1.5

init_b

float

0.5

occupancy_quantile_low

Optional[float]

null

occupancy_quantile_high

Optional[float]

null

occupancy_target_low

Optional[float]

null

occupancy_target_high

Optional[float]

null

calibration_min_quantile_width

float

0.001

calibration_max_m

float

50.0

calibration_revert_on_invalid

bool

true

sigma_aware_k

float

0.25

gradient_scaling

str

none

memory_efficient

bool

false

model.core.blocklinear

Explained in Synapses and connectivity. Class dendritic_modeling.config.model.BlockLinearConfig.

Option

Type

Default

gradient_scaling

str

none

efficient

bool

false

model.core.implementation

Explained in Structured E/I cores. Class dendritic_modeling.config.model.ImplementationConfig.

Option

Type

Default

adaptive_initialization

bool

true

adaptive_initialization_policy

str

preserve_shunting_center

adaptive_target_conductance

float

5.0

initial_child_conductance

float

1.0

initial_conductance_scale

float

1.0

initial_child_conductance_log_sd

float

0.0

print_hooks

bool

false

store_routing

bool

false

compile_forward

bool

false

model.core.spatial

Explained in Embedded and spatial cores. Class dendritic_modeling.config.model.SpatialConfig.

Option

Type

Default

kernel_size

int

3

stride

int

1

padding

int

0

input_transform

str

identity

input_scale

float

1.0

output_adapter_mode

str

identity

output_initial_scale

float

1.0

output_initial_threshold

float

0.25

model.decoder.params

Explained in Encoder and decoder types. Class dendritic_modeling.config.model.DecoderParamsConfig.

Option

Type

Default

input_dim

int

20

hidden_dims

list[int]

[16]

activation

str

relu

output_dim

Optional[int]

10

input_normalization

str

none

input_norm_eps

float

1.0e-05

output_init

str

default

training.main.common

Explained in Training. Class dendritic_modeling.config.training.CommonTrainerConfig.

Option

Type

Default

epochs

int

1000

batch_size

int

512

shuffle

bool

true

grad_clip_value

Optional[float]

2.0

grad_clip_mode

Optional[str]

null

loss_function

str

ce

early_stopping

bool

true

patience

int

500

load_best_state_dict

bool

true

best_state_device

str

model

save_best_checkpoint

str

every_improvement

plot_losses

bool

true

suppress_prints

bool

false

print_every

int

10

num_workers

Optional[int]

null

pin_memory

Optional[bool]

null

persistent_workers

Optional[bool]

null

prefetch_factor

Optional[int]

null

in_order

Optional[bool]

null

use_amp

bool

false

lr_schedule

str

none

lr_warmup_epochs

int

0

enable_nan_checking

bool

false

profile_dataloader

bool

false

log_first_step_diagnostics

bool

false

reactivation_initialization_batch_size

int

256

reactivation_initialization_num_batches

int

3

reactivation_initialization_max_iterations

Optional[int]

null

synapse_decay_rate

float

0.01

weight_boosting

bool

false

recalibrate_reactivation_every

int

0

reactivation_update_mode

str

backprop

reactivation_recalibration_mode

str

from_init_policy

reactivation_recalibration_start_epoch

int

1

reactivation_recalibration_num_batches

int

3

reactivation_recalibration_ema_alpha

float

1.0

param_groups

ParamGroupsConfig

training.main.common.param_groups

impact_rewiring

dict[str, Any]

{}

training.main.optimizer

Explained in Training. Class dendritic_modeling.config.training.OptimizerConfig.

Option

Type

Default

name

str

adam

lr

float

0.001

weight_decay

float

0.0

momentum

float

0.9

betas

tuple

[0.9, 0.999]

eps

float

1.0e-08

foreach

bool | None

null

muon_lr_multiplier

float

10.0

muon_momentum

float

0.95

muon_nesterov

bool

true

muon_ns_steps

int

5

muon_hyperball

bool

true

muon_min_dim

int

2

muon_rank_consistency_check_steps

int

1

training.main.regularization

Explained in Regularization and pruning. Class dendritic_modeling.config.regularization.RegularizationConfig.

Option

Type

Default

l1_weight

float

0.0

l2_weight

float

0.0

routing_entropy_weight

float

0.0

routing_balance_weight

float

0.0

selective

SelectiveConfig

training.main.regularization.selective

split_params

bool

false

param_group_weights

dict[str, dict[str, float]]

mapping; keys below

param_group_weights.topk

dict

{l1_weight: 0.0, l2_weight: 0.0}

param_group_weights.blocklinear

dict

{l1_weight: 0.0, l2_weight: 0.0}

param_group_weights.reactivation

dict

{l1_weight: 0.0, l2_weight: 0.0}

param_group_weights.decoder

dict

{l1_weight: 0.0, l2_weight: 0.0}

enforce_ei_weight_ratio

bool

false

target_ei_weight_ratio

float

0.5

ei_ratio_loss_weight

float

0.5

ei_ratio_scope

str

per_branch

ei_ratio_metric

str

mean

training.main.pruning

Explained in Regularization and pruning. Class dendritic_modeling.config.regularization.PruningConfig.

Option

Type

Default

enabled

bool

false

threshold

float

2

selective

SelectiveConfig

training.main.pruning.selective

re_evaluate_after

bool

false

report_non_pruned

bool

true

save_pruning_stats

bool

false

detailed_branch_report

bool

true

redo_analysis_after_pruning

bool

false

training.main.strategies

Explained in Training strategies. Class dendritic_modeling.config.training.TrainingStrategiesConfig.

Option

Type

Default

standard

StandardStrategyConfig

training.main.strategies.standard

multi_stage

MultiStageStrategyConfig

training.main.strategies.multi_stage

recurrent

RecurrentStrategyConfig

training.main.strategies.recurrent

voltage_stabilization

VoltageStabilizationStrategyConfig

training.main.strategies.voltage_stabilization

homeostatic_control

HomeostaticControlStrategyConfig

training.main.strategies.homeostatic_control

ei_equilibrium

EIEquilibriumStrategyConfig

training.main.strategies.ei_equilibrium

freeze_layers

FreezeLayersStrategyConfig

training.main.strategies.freeze_layers

freeze_branches

FreezeBranchesStrategyConfig

training.main.strategies.freeze_branches

freeze_branch_kl

FreezeBranchKLStrategyConfig

training.main.strategies.freeze_branch_kl

presentation_stream

PresentationStreamStrategyConfig

training.main.strategies.presentation_stream

extensions

plugin sections

{}; Configuration sections

training.main.learning_strategy_config

Explained in Local credit assignment. Class dendritic_modeling.config.training.LocalRuleConfig.

Option

Type

Default

rule_variant

str

3f

allow_unmodeled_forward_features

bool

false

allow_scalar_fallback

bool

false

error_mode

str

auto

soma_error_source

str

decoder

dfa_feedback_seed

Optional[int]

null

dfa_feedback_scale

float

1.0

error_broadcast_mode

str

scalar

path_transport_projection

str

none

broadcast_seed

Optional[int]

null

error_noise_sigma

float

0.0

broadcast_rank

int

4

broadcast_init_scale

float

1.0

pathway_broadcast_residual

float

0.25

pathway_activity_gate_strength

float

1.0

broadcast_bandwidth

str

full

broadcast_bits

int

8

broadcast_topk_fraction

float

0.3

three_factor

ThreeFactorConfig

training.main.learning_strategy_config.three_factor

four_factor

FourFactorConfig

training.main.learning_strategy_config.four_factor

five_factor

FiveFactorConfig

training.main.learning_strategy_config.five_factor

morphology_aware

MorphologyAwareConfig

training.main.learning_strategy_config.morphology_aware

hsic

HSICConfig

training.main.learning_strategy_config.hsic

stdp

STDPConfig

training.main.learning_strategy_config.stdp

inhibitory_homeostasis

InhibitoryHomeostasisConfig

training.main.learning_strategy_config.inhibitory_homeostasis

voltage_homeostasis

LocalVoltageHomeostasisConfig

training.main.learning_strategy_config.voltage_homeostasis

gate_homeostasis

LocalGateHomeostasisConfig

training.main.learning_strategy_config.gate_homeostasis

update_inactive_weights

bool

false

update_reactivation

bool

true

explicit_inhibitory_update_mode

str

local_ca

update_explicit_inhibitory_cells

Optional[bool]

null

core_update_mode

str

local

encoder_update_mode

str

none

decoder_update_mode

str

backprop

freeze_encoder_epochs

int

0

freeze_reactivation_epochs

int

0

freeze_decoder_epochs

int

0

training_schedule

Optional[list[dict[str, Any]]]

null

clip_grad_value

float

5.0

normalize_by_batch

bool

true

analysis.runtime.training

Explained in When analyzers run. Class dendritic_modeling.config.analysis.EvaluationRuntimeConfig.

Option

Type

Default

mode

str

auto

batch_size

int

256

num_workers

int

4

pin_memory

bool

true

persistent_workers

bool

false

prefetch_factor

Optional[int]

2

in_order

Optional[bool]

null

max_batches

Optional[int]

null

max_samples

Optional[int]

null

materialize_threshold_mb

int

1024

seed

Optional[int]

null

analysis.runtime.final

Explained in When analyzers run. Class dendritic_modeling.config.analysis.EvaluationRuntimeConfig.

Option

Type

Default

mode

str

auto

batch_size

int

256

num_workers

int

4

pin_memory

bool

true

persistent_workers

bool

false

prefetch_factor

Optional[int]

2

in_order

Optional[bool]

null

max_batches

Optional[int]

null

max_samples

Optional[int]

null

materialize_threshold_mb

int

1024

seed

Optional[int]

null

analysis.performance.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.PerformanceAnalysisParams.

Option

Type

Default

accuracy

bool

true

auc

bool

true

categorical_loglikelihood

bool

true

mse

bool

false

cosine_similarity

bool

false

analysis.synapse_turnover.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SynapseTurnoverAnalysisParams.

Option

Type

Default

snapshot_interval

int

1

analysis.synapse_pruning.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SynapsePruningAnalysisParams.

Option

Type

Default

save_figures

bool

true

analysis.noise_perturbation.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.NoisePerturbationAnalysisParams.

Option

Type

Default

uniform_noise

bool

true

uniform_magnitudes

list[float]

[0.1, 0.2]

gaussian_noise

bool

true

gaussian_sdevs

list[float]

[0.1, 0.2]

n_samples

int

10

clamp_range

Optional[list[float]]

[0.0, 1.0]

accuracy

bool

true

auc

bool

true

categorical_loglikelihood

bool

true

mse

bool

true

cosine_similarity

bool

true

seed

Optional[int]

null

analysis.single_layer_contribution.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SingleLayerContributionAnalysisParams.

Option

Type

Default

excitation_contribution

bool

true

inhibition_contribution

bool

true

both_contribution

bool

true

accuracy

bool

true

auc

bool

true

categorical_loglikelihood

bool

true

mse

bool

true

cosine_similarity

bool

true

analysis.ablation.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.AblationAnalysisParams.

Option

Type

Default

selection

AblationSelectionParams

analysis.ablation.params.selection

methods

AblationMethodsParams

analysis.ablation.params.methods

analysis.path_matched_intervention.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.PathMatchedInterventionAnalysisParams.

Option

Type

Default

targets

list[str]

[excitation_current, inhibition_current, joint_EI_currents]

methods

list[str]

[zero, fixed_train_mean, global_test_shuffle]

depths

list[int]

[]

paths_per_soma

list[int]

[]

support_draws

int

5

support_seed

Optional[int]

null

support_namespace

str

path-matched-depth-intervention-v1

shuffle_seed

Optional[int]

null

module_name_prefix

Optional[str]

null

expected_n_somas

Optional[int]

null

expected_branch_factors

list[int]

[]

chance_accuracy

Optional[float]

null

max_samples

Optional[int]

null

analysis.branch_local_sensitivity.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.BranchLocalSensitivityAnalysisParams.

Option

Type

Default

depths

list[int]

[]

module_name_prefix

Optional[str]

null

expected_n_somas

Optional[int]

null

expected_branch_factors

list[int]

[]

max_samples

Optional[int]

null

include_soma

bool

false

analysis.node_perturbation_credit.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.NodePerturbationCreditAnalysisParams.

Option

Type

Default

depths

list[int]

[1]

module_name_prefix

Optional[str]

null

coordinate

str

output

dictionaries

list[str]

[independent]

relative_scales

list[float]

[0.01]

draws

list[int]

[1, 4, 16, 64]

match_perturbation_energy

bool

false

loss

str

cross_entropy

max_samples

Optional[int]

256

chunk

int

16

seed

Optional[int]

null

analysis.local_inhibitory_population_intervention.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.LocalInhibitoryPopulationInterventionAnalysisParams.

Option

Type

Default

methods

list[str]

[global_test_shuffle, within_class_test_shuffle, zero]

permutation_draws

int

3

shuffle_seed

Optional[int]

null

expected_network_layers

Optional[int]

null

expected_inhibitory_width

Optional[int]

null

max_samples

Optional[int]

null

analysis.path_matched_information.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.PathMatchedInformationAnalysisParams.

Option

Type

Default

signals

list[str]

[excitation_current, inhibition_current, joint_EI_current...

depths

list[int]

[]

support_draws

int

5

support_seed

Optional[int]

null

support_namespace

str

path-matched-depth-intervention-v1

information_seed

Optional[int]

null

n_neighbors

int

10

n_label_shuffles

int

20

module_name_prefix

Optional[str]

null

expected_n_somas

Optional[int]

null

expected_branch_factors

list[int]

[]

max_samples

Optional[int]

null

analysis.weight.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.WeightAnalysisParams.

Option

Type

Default

synapse_weights

bool

true

branch_weights

bool

true

computation_level

str

single_branch

branch_aggregation

str

mean

per_layer_analysis

bool

true

per_einet_analysis

bool

true

per_neuron_analysis

bool

false

analyze_excitatory

bool

true

analyze_inhibitory

bool

true

analyze_branch_output

bool

true

compute_mean

bool

true

compute_variance

bool

true

compute_percentiles

bool

true

compute_min_max

bool

true

compute_dendritic_strength

bool

true

weight_threshold

float

1.0e-06

compute_source_support

bool

false

source_support_target_polarities

list[str]

[excitatory]

source_support_pathways

list[str]

[ff_excitatory, ff_inhibitory]

source_support_image_shape

Optional[tuple[int, int]]

null

source_support_plot_soma_indices

list[int]

[]

source_support_fig_save_format

str

pdf

compute_source_region_alignment

bool

false

source_region_soma_to_class

list[int]

[]

source_region_reference_split

str

validation

source_region_image_network_layers

list[int]

[0]

source_region_foreground_thresholds

list[float]

[0.0, 0.5]

source_region_center_bounds

Optional[tuple[int, int, int, int]]

null

source_region_max_reference_samples

Optional[int]

null

analysis.compartment_statistics.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CompartmentStatisticsAnalysisParams.

Option

Type

Default

synapse_weights

bool

true

branch_weights

bool

true

inputs

bool

true

activations

bool

true

global_analysis

bool

true

layer_analysis

bool

true

branch_analysis

bool

true

compute_mean

bool

true

compute_variance

bool

true

compute_percentiles

bool

true

compute_min_max

bool

true

compute_entropy

bool

true

per_class_analysis

bool

false

n_samples

Optional[int]

null

seed

Optional[int]

null

analysis.compartment_snr.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CompartmentSNRAnalysisParams.

Option

Type

Default

input_analysis

bool

true

input_capture_mode

str

branch

layer_analysis

bool

true

branch_analysis

bool

true

global_aggregation

bool

true

layer_aggregation

bool

true

branch_aggregation

bool

true

epsilon

float

1.0e-06

null_permutations

int

100

null_quantile

float

0.95

null_seed

Optional[int]

null

analysis.multiplicative_gain.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.MultiplicativeGainAnalysisParams.

Option

Type

Default

uniform_gain

bool

true

gain_min

float

0.0

gain_max

float

10.0

n_steps

int

10

logspace

bool

false

input_preprocess

str

none

noise_model

str

none

poisson_eps

float

1.0e-06

noise_seed

Optional[int]

null

accuracy

bool

false

auc

bool

false

categorical_loglikelihood

bool

false

mse

bool

false

cosine_similarity

bool

false

analysis.gain_load_perturbation.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.GainLoadPerturbationAnalysisParams.

Option

Type

Default

gain_log_sds

list[float]

[0.0, 0.2, 0.5, 0.8]

load_levels

list[float]

[0.0, 0.1, 0.3, 0.5]

load_mode

str

gaussian_input

load_log_sd

float

0.2

inhibitory_start

Optional[int]

null

draws_per_condition

int

5

base_draw_seed

Optional[int]

null

max_samples

Optional[int]

null

clamp_min

Optional[float]

0.0

clamp_max

Optional[float]

null

accuracy

bool

true

categorical_loglikelihood

bool

true

record_input_statistics

bool

false

record_branch_local_sensitivity

bool

false

branch_sensitivity_module_name_prefix

Optional[str]

null

branch_sensitivity_expected_n_somas

Optional[int]

null

branch_sensitivity_include_soma

bool

false

branch_sensitivity_max_samples

Optional[int]

null

branch_sensitivity_draws

list[int]

[0]

analysis.synaptic_activation.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SynapticActivationAnalysisParams.

Option

Type

Default

network_summary

bool

true

layer_summary

bool

true

branch_summary

bool

true

raw_summary

bool

true

logspace

bool

false

downsample

bool

true

samples

int

5

analysis.branch_activation.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.BranchActivationAnalysisParams.

Option

Type

Default

synapse_activations

bool

true

branch_activations

bool

true

network_summary

bool

true

layer_summary

bool

true

branch_summary

bool

true

raw_summary

bool

true

logspace

bool

false

downsample

bool

true

samples

int

5

analysis.local_rule_components.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.LocalRuleComponentAnalysisParams.

Option

Type

Default

max_samples

Optional[int]

1024

include_dendritic

bool

true

include_broadcast

bool

true

broadcast_mode

str

scalar

error_mode

str

auto

e_rev_exc

float

1.0

e_rev_inh

float

0.0

save_json

bool

true

save_csv

bool

true

append_training_csv

bool

true

analysis.information.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationAnalysisParams.

Option

Type

Default

selection

InformationSelectionParams

analysis.information.params.selection

compute

InformationComputeParams

analysis.information.params.compute

estimator

InformationEstimatorParams

analysis.information.params.estimator

baselines

InformationBaselinesParams

analysis.information.params.baselines

pid

InformationPIDParams

analysis.information.params.pid

method

str

kraskov

max_samples

Optional[int]

null

n_neighbors

int

10

n_bins

int

20

copula_type

str

gaussian

decoder_cv_folds

int

5

decoder_C

float

1.0

decoder_standardize

bool

true

decoder_seed

Optional[int]

null

decoder_continuous_strategy

str

none

decoder_continuous_max_total_dim

int

2

decoder_gaussian_ridge

float

1.0e-06

binning_strategy

str

quantile

binning_bias_correction

str

none

mi_null_shuffles

int

0

mi_null_seed

Optional[int]

null

compute_basic_mi

bool

true

compute_pairwise_mi

bool

true

compute_conditional_mi

bool

true

components

list[str]

[]

compute_ablation_aligned_mi

bool

true

compute_upstream_unique_cmi

bool

false

compute_layer_total_proxies

bool

true

layer_total_topk

int

10

scopes

list[str]

[]

views

list[str]

[]

pid_binarize

bool

false

pid_binarize_method

str

median

pid_binarize_threshold

float

0.5

include_vinf

bool

true

normalize_mi

bool

false

verbose

bool

false

computation_level

str

single_branch

branch_aggregation

str

mean

per_neuron_analysis

bool

false

per_einet_analysis

bool

false

per_layer_analysis

bool

false

compute_lda_weights

bool

false

signal_variants

list[str]

[]

lda_n_shuffles

int

0

analysis.representation_accessibility.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationAccessibilityAnalysisParams.

Option

Type

Default

representations

list[str]

[excitatory, inhibitory, joint]

projections

list[RepresentationProjectionConfig]

analysis.representation_accessibility.params.projections

probes

list[RepresentationProbeConfig]

analysis.representation_accessibility.params.probes

information

RepresentationInformationConfig

analysis.representation_accessibility.params.information

subset_decoding

RepresentationSubsetProbeConfig

analysis.representation_accessibility.params.subset_decoding

geometry

RepresentationGeometryConfig

analysis.representation_accessibility.params.geometry

max_samples

Optional[int]

null

seed

Optional[int]

null

analysis.source_tuning_support.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SourceTuningSupportAnalysisParams.

Option

Type

Default

pathways

list[str]

[ff_excitatory, ff_inhibitory]

target_polarities

list[str]

[excitatory, inhibitory]

image_shape

Optional[tuple[int, ...]]

null

image_source_layers

list[int]

[]

image_source_names

list[str]

[input, input_e, input_i]

image_binarization_threshold

float

0.5

continuous_activity_threshold

float

0.0

input_batch_index

int

0

label_batch_index

int

1

max_reference_samples

Optional[int]

null

max_evaluation_samples

Optional[int]

null

require_independent_splits

bool

true

retain_coordinate_profiles

bool

false

epsilon

float

1.0e-12

reference_split

str

train

evaluation_split

str

validation

analysis.input_region_intervention.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InputRegionInterventionAnalysisParams.

Option

Type

Default

image_shape

Optional[tuple[int, ...]]

null

regions

list[InputImageRegionConfig]

analysis.input_region_intervention.params.regions

operations

list[str]

[remove_only, retain_only]

replacement_methods

list[str]

[zero, reference_mean]

reference_split

str

train

evaluation_split

str

test

max_reference_samples

Optional[int]

null

max_evaluation_samples

Optional[int]

null

target_population_polarities

list[str]

[excitatory]

identity_soma_class_mapping

Optional[list[int]]

null

identity_soma_population_name

Optional[str]

null

identity_soma_network_layer_index

Optional[int]

null

analysis.receptive_field.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.ReceptiveFieldAnalysisParams.

Option

Type

Default

n_rows

int

2

n_cols

int

2

rf_shape

Optional[tuple[int, int]]

null

exc_rf_shape

Optional[tuple[int, int]]

null

inh_rf_shape

Optional[tuple[int, int]]

null

fig_save_format

str

png

compute_activation_tuning

bool

false

activation_sample_signals

list[str]

[]

activation_sample_depths

list[int]

[]

activation_sample_branch_indices

list[int]

[]

activation_sample_branch_indices_by_depth

dict[int, list[int]]

{}

activation_samples_per_class

int

0

activation_sample_seed

Optional[int]

null

compute_component_rfs

bool

false

plot_component_rfs

bool

true

compute_rf_tuning

bool

false

plot_rf_tuning

bool

true

rf_tuning_layer_index

int

0

n_class_examples

int

5

save_raw_data

bool

true

load_raw_data

bool

false

raw_filename

Optional[str]

null

epsilon

float

1.0e-08

analysis.reactivation_dynamics.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.ReactivationDynamicsAnalysisParams.

Option

Type

Default

max_samples

int

1024

analysis.correlation.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CorrelationAnalysisParams.

Option

Type

Default

selection

CorrelationSelectionParams

analysis.correlation.params.selection

compute

CorrelationComputeParams

analysis.correlation.params.compute

sampling

CorrelationSamplingParams

analysis.correlation.params.sampling

computation_level

str

layer_branch

components

list[str]

[]

pairs

list[str]

[]

scopes

list[str]

[]

compute_noise_correlations

bool

true

compute_signal_correlations

bool

true

compute_total_correlations

bool

true

compute_tuning_curves

bool

true

analyze_EE

bool

true

analyze_II

bool

true

analyze_EI

bool

true

analyze_E_output

bool

true

analyze_I_output

bool

true

analyze_output_output

bool

true

per_layer_analysis

bool

true

per_einet_analysis

bool

true

per_neuron_analysis

bool

false

n_synapse_pairs

int

10000

sampling_strategy

str

stratified

stratified_categories

list[str]

[within_branch, within_layer]

max_pairs_per_category

int

1000

max_samples

Optional[int]

5000

analysis.inhibitory_specialization.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InhibitorySpecializationAnalysisParams.

Option

Type

Default

gate_threshold

float

0.1

std_threshold

float

0.15

save_figures

bool

true

analysis.jacobian_spectrum.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.JacobianSpectrumAnalysisParams.

Option

Type

Default

top_k

int

16

save_figures

bool

true

analysis.dendritic_timetraces.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.DendriticTimetracesAnalysisParams.

Option

Type

Default

n_timesteps

int

90

stim1_window

Optional[tuple[int, int]]

[5, 25]

stim2_window

Optional[tuple[int, int]]

[55, 75]

response_window

Optional[tuple[int, int]]

[75, 90]

stim1_amp

float

0.8

stim2_amp

float

0.6

noise_std

float

0.05

seed

Optional[int]

null

save_figures

bool

true

analysis.spike_train.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.SpikeTrainAnalysisParams.

Option

Type

Default

dt

float

1.0

threshold

float

0.5

max_samples

Optional[int]

512

include_raster

bool

true

max_raster_samples

int

8

include_per_timestep_rate

bool

true

analysis.adversarial_robustness.params

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.AdversarialRobustnessAnalysisParams.

Option

Type

Default

epsilons

list[float]

[0.0, 0.05, 0.1, 0.15, 0.2, 0.3]

batch_size

int

256

model.core.feature_readout.readout

Explained in Baselines and comparison models. Class dendritic_modeling.config.affine_mlp.AffineMLPConfig.

Option

Type

Default

dimensions

list[int]

[]

mode

str

dense

initialization

str

kaiming_normal

fanin

list[int]

[]

connectivity_seeds

list[int]

[]

activation

str

relu

negative_slope

float

0.1

activate_output

bool

true

dropout

float | None

null

output_weight_scale

float

1.0

projection_backend

str

eager

workspace_mb

float

128.0

index_dtype

str

int32

model.core.connectivity.structured

Explained in Structured connectivity. Class dendritic_modeling.config.model.StructuredConnectivityConfig.

Option

Type

Default

enabled

bool

false

seed

Optional[int]

null

method

str

bernoulli

probability

Optional[float]

null

indegree

Optional[int]

null

distance_kernel

str

gaussian

distance_sigma

float

1.0

spatial

dict[str, Any]

{}

pathways

dict[str, Any]

{}

training.main.common.param_groups

Explained in Parameter groups and learning rates. Class dendritic_modeling.config.training.ParamGroupsConfig.

Option

Type

Default

lr

float

0.01

split_params

bool

true

topk_lr

float

0.001

blocklinear_lr

float

0.0001

reactivation_lr

float

0.0001

decoder_lr

float

0.001

decoder_input_lr

Optional[float]

null

encoder_lr

Optional[float]

null

training.main.regularization.selective

Explained in Regularization and pruning. Class dendritic_modeling.config.regularization.SelectiveConfig.

Option

Type

Default

layers

list[int]

[]

branches

list[str]

[]

weight_types

list[str]

[all]

training.main.pruning.selective

Explained in Regularization and pruning. Class dendritic_modeling.config.regularization.SelectiveConfig.

Option

Type

Default

layers

list[int]

[]

branches

list[str]

[]

weight_types

list[str]

[all]

training.main.strategies.standard

Explained in Training strategies. Class dendritic_modeling.config.training.StandardStrategyConfig.

This section takes no options.

training.main.strategies.multi_stage

Explained in Multi-stage training. Class dendritic_modeling.config.multi_stage_training.MultiStageStrategyConfig.

Option

Type

Default

analyze_between_stages

bool

true

final_state

str

last_stage

continue_on_failure

bool

false

stages

list[TrainingStageConfig]

training.main.strategies.multi_stage.stages

training.main.strategies.recurrent

Explained in Training strategies. Class dendritic_modeling.config.training.RecurrentStrategyConfig.

Option

Type

Default

grad_clip_value

Optional[float]

null

use_amp

Optional[bool]

null

lr_schedule

Optional[str]

null

lr_warmup_epochs

Optional[int]

null

training.main.strategies.voltage_stabilization

Explained in Training strategies. Class dendritic_modeling.config.training.VoltageStabilizationStrategyConfig.

Option

Type

Default

pretrain_epochs

int

1

stabilize_mode

str

vinf

freeze_reactivation

bool

false

target_voltage

float

0.5

target_saturation

float

0.2

loss_metric

str

mse

training.main.strategies.homeostatic_control

Explained in Training strategies. Class dendritic_modeling.config.training.HomeostaticControlStrategyConfig.

Option

Type

Default

pretrain_epochs

int

1

stabilize_mode

str

vinf

freeze_reactivation

bool

false

target_voltage

float

0.5

target_saturation

float

0.2

loss_metric

str

mse

training.main.strategies.ei_equilibrium

Explained in Training strategies. Class dendritic_modeling.config.training.EIEquilibriumStrategyConfig.

Option

Type

Default

pretrain_epochs

int

1

freeze_reactivation

bool

false

loss_metric

str

mse

training.main.strategies.freeze_layers

Explained in Training strategies. Class dendritic_modeling.config.training.FreezeLayersStrategyConfig.

Option

Type

Default

pretrain_epochs

int

10

reverse_training

bool

false

epochs_per_layer

int

5

final_tune_epochs

int

10

training.main.strategies.freeze_branches

Explained in Training strategies. Class dendritic_modeling.config.training.FreezeBranchesStrategyConfig.

Option

Type

Default

pretrain_epochs

int

5

reverse_training

bool

false

epochs_per_branch

int

50

final_tune_epochs

int

10

training.main.strategies.freeze_branch_kl

Explained in Training strategies. Class dendritic_modeling.config.training.FreezeBranchKLStrategyConfig.

Option

Type

Default

pretrain_epochs

int

5

reverse_training

bool

false

epochs_per_branch

int

50

final_tune_epochs

int

10

kl_weight

float

0.1

training.main.strategies.presentation_stream

Explained in Training strategies. Class dendritic_modeling.config.training.PresentationStreamStrategyConfig.

Option

Type

Default

presentations_per_epoch

int

65536

lr_terminal_multiplier

float

1.0

sample_seed_offset

int

4100

noise_seed_offset

int

9100

selection_metric

str

accuracy

selection_variant

Optional[str]

null

evaluation_batch_size

int

512

require_finite

bool

true

metrics_file

str

metrics.jsonl

stop_after_epochs

int

0

training.main.learning_strategy_config.three_factor

Explained in Local credit assignment. Class dendritic_modeling.config.training.ThreeFactorConfig.

Option

Type

Default

dynamics_mode

str

auto

allow_mismatched_dynamics_mode

bool

false

use_conductance_scaling

bool

true

use_driving_force

bool

true

theta

float

0.0

e_rev_exc

float

1.0

e_rev_inh

float

0.0

additive_gain_mode

str

none

additive_stats_ema_alpha

float

0.01

training.main.learning_strategy_config.four_factor

Explained in Local credit assignment. Class dendritic_modeling.config.training.FourFactorConfig.

Option

Type

Default

rho_mode

str

pearson

rho_estimator

str

ema

ema_alpha

float

0.05

layer_wise_rho_scale

float

1.0

augment_k

int

8

augment_noise_sigma

float

0.01

training.main.learning_strategy_config.five_factor

Explained in Local credit assignment. Class dendritic_modeling.config.training.FiveFactorConfig.

Option

Type

Default

phi_mode

str

conditional

phi_estimator

str

conditional_ema

phi_ridge_lambda

float

0.001

layer_wise_phi_scale

float

1.0

phi_clamp_min

float

0.25

phi_clamp_max

float

4.0

training.main.learning_strategy_config.morphology_aware

Explained in Local credit assignment. Class dendritic_modeling.config.training.MorphologyAwareConfig.

Option

Type

Default

use_path_propagation

bool

false

path_factor_mode

str

per_branch

morphology_modulator_mode

str

none

morphology_depth_offset

float

1.0

use_dendritic_normalization

bool

false

use_branch_type_rules

bool

false

apical_branch_scale

float

1.0

basal_branch_scale

float

1.0

use_branch_length_modulation

bool

false

use_branch_role_rules

bool

false

branch_role_source

str

router_pathways

specialized_branch_scale

float

1.0

mixed_branch_scale

float

1.0

branch_role_power

float

1.0

branch_role_alignment_weight

float

0.0

training.main.learning_strategy_config.hsic

Explained in Local credit assignment. Class dendritic_modeling.config.training.HSICConfig.

Option

Type

Default

enabled

bool

false

weight

float

0.0

self_weight

float

1.0

target_weight

float

1.0

target_source

str

labels

kernel

str

linear

sigma

float

1.0

degree

int

2

coef0

float

1.0

grad_clip_value

float

0.0

warmup_epochs

int

0

apply_last_layer_only

bool

false

training.main.learning_strategy_config.stdp

Explained in Local credit assignment. Class dendritic_modeling.config.training.STDPConfig.

Option

Type

Default

enabled

bool

false

apply_to

list[str]

[exc, rec_exc]

activity_mode

str

relu

pre_threshold

float

0.0

post_threshold

float

0.0

tau_pre

float

20.0

tau_post

float

20.0

a_plus

float

1.0

a_minus

float

0.5

learning_rate_scale

float

1.0

inhibitory_update_sign

float

-1.0

use_error_modulation

bool

false

error_modulation_mode

str

scalar_abs

clamp_update

float

0.0

detach_traces

bool

true

training.main.learning_strategy_config.inhibitory_homeostasis

Explained in Local credit assignment. Class dendritic_modeling.config.training.InhibitoryHomeostasisConfig.

Option

Type

Default

enabled

bool

false

mode

str

r_tot

weight

float

0.0

target_r_tot

float

0.35

target_voltage

float

0.15

training.main.learning_strategy_config.voltage_homeostasis

Explained in Local credit assignment. Class dendritic_modeling.config.training.LocalVoltageHomeostasisConfig.

Option

Type

Default

enabled

bool

false

weight

float

0.0

target_voltage

float

0.5

training.main.learning_strategy_config.gate_homeostasis

Explained in Local credit assignment. Class dendritic_modeling.config.training.LocalGateHomeostasisConfig.

Option

Type

Default

enabled

bool

false

center_weight

float

0.0

saturation_weight

float

0.0

target_mean

float

0.5

target_saturation

float

0.2

low_threshold

float

0.1

high_threshold

float

0.9

saturation_slope

float

20.0

analysis.ablation.params.selection

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.AblationSelectionParams.

Option

Type

Default

levels

list[str]

[]

targets

list[str]

[]

metrics

list[str]

[]

analysis.ablation.params.methods

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.AblationMethodsParams.

Option

Type

Default

ablation_methods

list[str]

[lesion]

shuffle_seed

Optional[int]

null

clamp_statistic

str

mean

analysis.information.params.selection

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationSelectionParams.

Option

Type

Default

components

list[str]

[]

scopes

list[str]

[]

signal_variants

list[str]

[]

views

list[str]

[]

analysis.information.params.compute

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationComputeParams.

Option

Type

Default

max_samples

Optional[int]

null

include_vinf

bool

true

normalize_mi

bool

false

network_population_information

bool

false

output_units

str

bits

analysis_split

str

test

verbose

bool

false

computation_level

str

single_branch

branch_aggregation

str

mean

branch_sample_count_per_layer

Optional[int]

null

branch_sample_seed

Optional[int]

null

branch_sample_strategy

str

uniform

layer_total_topk

int

10

gaussian_fisher

InformationGaussianFisherParams

analysis.information.params.compute.gaussian_fisher

per_neuron_analysis

bool

false

per_einet_analysis

bool

false

per_layer_analysis

bool

false

compute_lda_weights

bool

false

analysis.information.params.estimator

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationEstimatorParams.

Option

Type

Default

method

str

kraskov

kraskov

InformationEstimatorKraskovParams

analysis.information.params.estimator.kraskov

binned

InformationEstimatorBinnedParams

analysis.information.params.estimator.binned

decoder

InformationEstimatorDecoderParams

analysis.information.params.estimator.decoder

copula

InformationEstimatorCopulaParams

analysis.information.params.estimator.copula

sklearn

InformationEstimatorSklearnParams

analysis.information.params.estimator.sklearn

analysis.information.params.baselines

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationBaselinesParams.

Option

Type

Default

random_weights

InformationBaselinesRandomWeightsParams

analysis.information.params.baselines.random_weights

label_shuffle_null

InformationBaselinesLabelShuffleNullParams

analysis.information.params.baselines.label_shuffle_null

analysis.information.params.pid

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationPIDParams.

Option

Type

Default

binarize

bool

false

binarize_method

str

median

binarize_threshold

float

0.5

analysis.representation_accessibility.params.projections

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationProjectionConfig.

Option

Type

Default

name

str

native

type

str

identity

output_dim

Optional[int]

null

draws

int

1

analysis.representation_accessibility.params.probes

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationProbeConfig.

Option

Type

Default

name

str

linear

type

str

ridge

draws

int

1

max_projection_draws

Optional[int]

null

n_label_shuffles

int

20

ridge_alpha

float

1.0

hidden_dims

list[int]

[32, 16]

activation

str

relu

alpha

float

0.0001

learning_rate_init

float

0.001

batch_size

int

256

max_iter

int

300

early_stopping

bool

true

validation_fraction

float

0.1

n_iter_no_change

int

20

analysis.representation_accessibility.params.information

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationInformationConfig.

Option

Type

Default

enabled

bool

false

compute_joint

bool

true

compute_per_soma

bool

false

projection_names

list[str]

[matched16]

per_soma_representations

list[str]

[excitatory, inhibitory]

n_neighbors

int

5

n_label_shuffles

int

20

standardize

bool

true

analysis.representation_accessibility.params.subset_decoding

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationSubsetProbeConfig.

Option

Type

Default

enabled

bool

false

representations

list[str]

[excitatory]

subset_sizes

list[int]

[1, 2, 4, 8, 16, 32, 64]

draws

int

10

probe

RepresentationProbeConfig

analysis.representation_accessibility.params.subset_decoding.probe

analysis.representation_accessibility.params.geometry

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationGeometryConfig.

Option

Type

Default

enabled

bool

false

representations

list[str]

[excitatory]

variance_epsilon

float

1.0e-12

analysis.input_region_intervention.params.regions

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InputImageRegionConfig.

Option

Type

Default

name

str

required

kind

str

required

center_fraction

float

0.5

foreground_threshold

float

0.5

analysis.correlation.params.selection

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CorrelationSelectionParams.

Option

Type

Default

components

list[str]

[]

pairs

list[str]

[]

scopes

list[str]

[]

analysis.correlation.params.compute

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CorrelationComputeParams.

Option

Type

Default

computation_level

str

layer_branch

max_samples

Optional[int]

5000

per_layer_analysis

bool

true

per_einet_analysis

bool

true

per_neuron_analysis

bool

false

analysis.correlation.params.sampling

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.CorrelationSamplingParams.

Option

Type

Default

n_synapse_pairs

int

10000

sampling_strategy

str

stratified

stratified_categories

list[str]

[within_branch, within_layer]

max_pairs_per_category

int

1000

training.main.strategies.multi_stage.stages

Explained in Multi-stage training. Class dendritic_modeling.config.multi_stage_training.TrainingStageConfig.

Option

Type

Default

learning_strategy

str

standard

stage_name

Optional[str]

null

epochs

Optional[int]

null

trainer_config

dict[str, Any]

{}

batch_size

Optional[int]

null

param_groups

Optional[dict[str, Any]]

null

reset_optimizer

bool

false

reset_model

bool

false

load_checkpoint

Optional[str]

null

save_checkpoint

bool

true

local_rule_config

Optional[dict[str, Any]]

null

analysis.information.params.compute.gaussian_fisher

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationGaussianFisherParams.

Option

Type

Default

aggregation

str

sum

mi_transform

str

half_log1p

eps

float

1.0e-08

ridge

Optional[float]

null

analysis.information.params.estimator.kraskov

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationEstimatorKraskovParams.

Option

Type

Default

n_neighbors

int

10

analysis.information.params.estimator.binned

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationEstimatorBinnedParams.

Option

Type

Default

n_bins

int

20

binning_strategy

str

quantile

bias_correction

str

none

analysis.information.params.estimator.decoder

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationEstimatorDecoderParams.

Option

Type

Default

cv_folds

int

5

C

float

1.0

standardize

bool

true

seed

Optional[int]

null

continuous

InformationEstimatorDecoderContinuousParams

analysis.information.params.estimator.decoder.continuous

analysis.information.params.estimator.copula

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationEstimatorCopulaParams.

Option

Type

Default

type

str

gaussian

analysis.information.params.estimator.sklearn

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationEstimatorSklearnParams.

Option

Type

Default

n_bins

int

20

bandwidth

Optional[float]

null

analysis.information.params.baselines.random_weights

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationBaselinesRandomWeightsParams.

Option

Type

Default

n_shuffles

int

0

seed

Optional[int]

null

analysis.information.params.baselines.label_shuffle_null

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationBaselinesLabelShuffleNullParams.

Option

Type

Default

n_shuffles

int

0

seed

Optional[int]

null

components

list[str]

[basic]

analysis.representation_accessibility.params.subset_decoding.probe

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.RepresentationProbeConfig.

Option

Type

Default

name

str

linear

type

str

ridge

draws

int

1

max_projection_draws

Optional[int]

null

n_label_shuffles

int

20

ridge_alpha

float

1.0

hidden_dims

list[int]

[32, 16]

activation

str

relu

alpha

float

0.0001

learning_rate_init

float

0.001

batch_size

int

256

max_iter

int

300

early_stopping

bool

true

validation_fraction

float

0.1

n_iter_no_change

int

20

analysis.information.params.estimator.decoder.continuous

Explained in Analyzer catalogue. Class dendritic_modeling.config.analysis.InformationEstimatorDecoderContinuousParams.

Option

Type

Default

strategy

str

none

max_total_dim

int

2

gaussian_ridge

float

1.0e-06