vraven · class

VRavenReport

VRavenReport(trace: 'NeuralTrace', full_tensors: 'dict[str, Any]' = <factory>, lrp_result: 'LRPResult | None' = None, perturbation_result: 'PerturbationResult | None' = None, counterfactual_result: 'CounterfactualResult | None' = None) -> None

VRavenReport(trace: 'NeuralTrace', full_tensors: 'dict[str, Any]' = <factory>, lrp_result: 'LRPResult | None' = None, perturbation_result: 'PerturbationResult | None' = None, counterfactual_result: 'CounterfactualResult | None' = None)

Usage

from vraven import VRavenReport

value = VRavenReport(...)

Methods

activation_energy(self, *, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

attention_structure(self, *, layer: 'str | int | None' = None, head: 'int' = 0, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

capabilities(self) -> 'dict[str, bool]'

Public method.

causal_pathway(self, *, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

community_graph(self, *, layer: 'str | int | None' = None, n_clusters: 'int' = 6, similarity_threshold: 'float' = 0.2, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

conflict_field(self, *, sample_index: 'int' = 0, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

consistency_matrix(self, explanations: 'dict[str, list[float] | np.ndarray]', *, metric: 'AgreementMetric' = 'rank_correlation', top_k: 'int | None' = None, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

counterfactual_delta(self, result: 'CounterfactualResult | None' = None, *, feature_names: 'list[str] | tuple[str, ...] | None' = None, class_names: 'list[str] | tuple[str, ...] | None' = None, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

decision_flow(self, *, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

decision_sankey(self, *, feature_names: 'list[str] | tuple[str, ...] | None' = None, class_names: 'list[str] | tuple[str, ...] | None' = None, top_features: 'int' = 7, depth_stages: 'int' = 5, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

evidence_trajectory(self, *, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

export(self, visual: 'str', path: 'str | Path', *, mode: 'RenderMode' = 'publication_light', **kwargs: 'Any') -> 'Path'

Public method.

export_html(self, path: 'str | Path', *, visuals: 'list[str] | tuple[str, ...] | None' = None, mode: 'RenderMode' = 'interactive_dark', figure_kwargs: 'dict[str, dict[str, Any]] | None' = None, theme_toggle: 'bool' = True, include_scientific_audit: 'bool' = True) -> 'Path'

Public method.

feature_sequence_heatmap(self, matrix: 'Any' = None, *, feature_names: 'list[str] | tuple[str, ...] | None' = None, sample_labels: 'list[str] | tuple[str, ...] | None' = None, top_features: 'int' = 16, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

figure(self, name: 'str', *, mode: 'RenderMode' = 'publication_light', **kwargs: 'Any') -> 'go.Figure'

Public method.

full_tensor(self, layer_name: 'str') -> 'Any'

Public method.

input_attribution(self, *, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None, feature_names: 'list[str] | tuple[str, ...] | None' = None, top_k: 'int | None' = 25) -> 'go.Figure'

Public method.

latent_flight(self, *, labels: 'list[Any] | tuple[Any, ...] | None' = None, layer_names: 'list[str] | tuple[str, ...] | None' = None, max_layers: 'int' = 4, method: 'str' = 'pca', random_seed: 'int' = 17, perplexity: 'float' = 30.0, align: 'bool' = True, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

load_trace(path: 'str | Path') -> 'VRavenReport'

Public method.

mechanism_concordance(self, *, mode: 'RenderMode' = 'publication_light', random_seed: 'int' = 17, max_permutations: 'int' = 4096, show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

neuron_impact(self, *, layer: 'str | int | None' = None, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

neuron_importance(self, *, layer: 'str | int | None' = None, top_k: 'int' = 12, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

perturbation_stability(self, result: 'PerturbationResult | None' = None, *, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

plot(self, name: 'str', *, mode: 'RenderMode' = 'interactive_dark', **kwargs: 'Any') -> 'go.Figure'

VRAVEN plot dispatcher.

reasoning_dna(self, *, dimensions: 'int' = 128, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

reasoning_fingerprint(self, dimensions: 'int' = 128) -> 'np.ndarray'

Public method.

reasoning_graph(self, *, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

release_full_tensors(self) -> 'None'

Release explicitly retained tensors, including any GPU memory they reference.

relevance_flow(self, result: 'LRPResult | None' = None, *, max_layers: 'int' = 6, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

representation_evolution(self, *, labels: 'list[Any] | tuple[Any, ...] | None' = None, layer_names: 'list[str] | tuple[str, ...] | None' = None, max_layers: 'int' = 4, method: 'str' = 'pca', random_seed: 'int' = 17, perplexity: 'float' = 30.0, align: 'bool' = True, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

representation_landscape(self, *, labels: 'list[Any] | tuple[Any, ...] | None' = None, layer_names: 'list[str] | tuple[str, ...] | None' = None, max_layers: 'int' = 4, method: 'str' = 'pca', random_seed: 'int' = 17, perplexity: 'float' = 30.0, align: 'bool' = True, mode: 'RenderMode' = 'publication_light', show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

retained_tensor_names(self) -> 'tuple[str, ...]'

Names of full layer outputs retained by explicit opt-in capture.

save_trace(self, path: 'str | Path') -> 'Path'

Public method.

scientific_audit(self) -> 'ScientificEvidenceAudit'

Return a transparent concentration and causal-concordance audit.

show(self, *, visual: 'str | None' = None, mode: 'RenderMode' = 'interactive_dark', **kwargs: 'Any') -> 'None'

Public method.

spectral_trace(self, *, mode: 'RenderMode' = 'publication_light', max_layers: 'int' = 64, show_title: 'bool | None' = None) -> 'go.Figure'

Public method.

summary(self) -> 'dict[str, object]'

Public method.

write_report(self, path: 'str | Path', *, format: 'str | None' = None, visuals: 'list[str] | tuple[str, ...] | None' = None, mode: 'RenderMode' = 'interactive_dark', figure_kwargs: 'dict[str, dict[str, Any]] | None' = None, theme_toggle: 'bool' = True, include_scientific_audit: 'bool' = True) -> 'Path'

Write exactly one report file in HTML, PDF or JSON format.

Source of truthThis page is generated from the checked-out public symbol, signature, and docstring. Pin the package version for reproducible work.

← API index Output guide →