TE Curve Verification Pipeline Familywise ONNX Report

Build familywise ONNX TE Curve Verification Pipeline reports.

class scripts.reports.analysis.build_track2_familywise_onnx_report.ExportedModelEntry(dataset_id, input_mode, model_family, model_type, surface, run_name, run_instance_id, dataset_schema, onnx_model_path, python_model_path, training_config_path, source_output_directory, source_best_checkpoint_path, source_inventory_path)[source]

Store one exported ONNX model entry from a model inventory.

Parameters:

  • dataset_id (str)

  • input_mode (str)

  • model_family (str)

  • model_type (str)

  • surface (str)

  • run_name (str)

  • run_instance_id (str)

  • dataset_schema (str)

  • onnx_model_path (Path)

  • python_model_path (Path)

  • training_config_path (Path)

  • source_output_directory (str)

  • source_best_checkpoint_path (str)

  • source_inventory_path (Path)

dataset_id

Dataset identifier used by the source run.

Type:

str

input_mode

Input-mode contract used by the source run.

Type:

str

model_family

Base family requested by the report.

Type:

str

model_type

Training/export model type.

Type:

str

surface

Model surface, either forward, backward, or global.

Type:

str

run_name

Logical training run name.

Type:

str

run_instance_id

Immutable source run instance identifier.

Type:

str

dataset_schema

Dataset schema stored in the export inventory.

Type:

str

onnx_model_path

Project-resolved ONNX model path.

Type:

pathlib.Path

python_model_path

Project-resolved Python model path.

Type:

pathlib.Path

training_config_path

Project-resolved source-run training config path.

Type:

pathlib.Path

source_output_directory

Project-relative source output directory.

Type:

str

source_best_checkpoint_path

Project-relative source model path.

Type:

str

source_inventory_path

Project-resolved export inventory path.

Type:

pathlib.Path

class scripts.reports.analysis.build_track2_familywise_onnx_report.CurveEvaluationEntry(group_id, surface, dataset_index, source_file_path, direction_label, speed_rpm, torque_nm, oil_temperature_deg, angular_position_deg, target_curve_deg, prediction_curve_deg, plot_measured_angular_position_deg, plot_measured_curve_deg, plot_prediction_angular_position_deg, plot_prediction_curve_deg, metrics)[source]

Store one evaluated test curve and prediction payload.

Parameters:

  • group_id (str)

  • surface (str)

  • dataset_index (int)

  • source_file_path (str)

  • direction_label (str)

  • speed_rpm (float)

  • torque_nm (float)

  • oil_temperature_deg (float)

  • angular_position_deg (ndarray)

  • target_curve_deg (ndarray)

  • prediction_curve_deg (ndarray)

  • plot_measured_angular_position_deg (ndarray)

  • plot_measured_curve_deg (ndarray)

  • plot_prediction_angular_position_deg (ndarray)

  • plot_prediction_curve_deg (ndarray)

  • metrics (dict[str, float])

scripts.reports.analysis.build_track2_familywise_onnx_report.build_argument_parser()[source]

Build command-line arguments for the familywise report.

Return type:

ArgumentParser

scripts.reports.analysis.build_track2_familywise_onnx_report.parse_command_line_arguments()[source]

Parse command-line arguments.

Return type:

Namespace

scripts.reports.analysis.build_track2_familywise_onnx_report.load_yaml_dictionary(path_value)[source]

Load one YAML dictionary from disk.

Parameters:

path_value (Path)

Return type:

dict[str, Any]

scripts.reports.analysis.build_track2_familywise_onnx_report.save_yaml_dictionary(path_value, payload)[source]

Save one YAML dictionary to disk.

Parameters:

  • path_value (Path)

  • payload (dict[str, Any])

Return type:

None

scripts.reports.analysis.build_track2_familywise_onnx_report.format_project_path(path_value)[source]

Return a stable project-relative path string.

Parameters:

path_value (str | Path)

Return type:

str

scripts.reports.analysis.build_track2_familywise_onnx_report.resolve_project_path(path_value)[source]

Resolve a repository-relative or absolute path.

Parameters:

path_value (str | Path)

Return type:

Path

scripts.reports.analysis.build_track2_familywise_onnx_report.parse_group_specification(group_specification)[source]

Parse one dataset/input-mode group specification.

Parameters:

group_specification (str)

Return type:

tuple[str, str]

scripts.reports.analysis.build_track2_familywise_onnx_report.resolve_inventory_path(dataset_id, input_mode)[source]

Resolve the exported model-development inventory for one group.

Parameters:

  • dataset_id (str)

  • input_mode (str)

Return type:

Path

scripts.reports.analysis.build_track2_familywise_onnx_report.normalize_surface_name(surface_name)[source]

Normalize inventory and report surface labels.

Parameters:

surface_name (str)

Return type:

str

scripts.reports.analysis.build_track2_familywise_onnx_report.load_group_model_entries(dataset_id, input_mode, model_family)[source]

Load and validate exported model entries for one dataset/input-mode group.

Parameters:

  • dataset_id (str)

  • input_mode (str)

  • model_family (str)

Return type:

dict[str, ExportedModelEntry]

scripts.reports.analysis.build_track2_familywise_onnx_report.load_test_dataset(model_entry)[source]

Build the source-run test dataset for one exported model.

Parameters:

model_entry (ExportedModelEntry)

scripts.reports.analysis.build_track2_familywise_onnx_report.compute_curve_metrics(target_curve_deg, prediction_curve_deg)[source]

Compute TE curve metrics for one prediction.

Parameters:

  • target_curve_deg (ndarray)

  • prediction_curve_deg (ndarray)

Return type:

dict[str, float]

scripts.reports.analysis.build_track2_familywise_onnx_report.average_metric_dictionary(metric_dictionary_list)[source]

Average curve-level metrics for one model.

Parameters:

metric_dictionary_list (list[dict[str, float]])

Return type:

dict[str, float]

scripts.reports.analysis.build_track2_familywise_onnx_report.load_onnx_session(model_entry, provider_list)[source]

Load one ONNX Runtime session.

Parameters:
Return type:

onnxruntime.InferenceSession

scripts.reports.analysis.build_track2_familywise_onnx_report.resolve_static_onnx_dimension(dimension_value)[source]

Return an ONNX dimension when it is a concrete integer.

Parameters:

dimension_value (Any)

Return type:

int | None

scripts.reports.analysis.build_track2_familywise_onnx_report.build_temporal_sequence_window_view(padded_feature_matrix, sequence_length)[source]

Build a vectorized temporal sequence-window view.

Parameters:

  • padded_feature_matrix (ndarray)

  • sequence_length (int)

Return type:

ndarray

scripts.reports.analysis.build_track2_familywise_onnx_report.predict_rank2_curve(session, input_feature_matrix)[source]

Predict one pointwise TE curve with a rank-2 ONNX input.

Parameters:

  • session (onnxruntime.InferenceSession)

  • input_feature_matrix (ndarray)

Return type:

ndarray

scripts.reports.analysis.build_track2_familywise_onnx_report.predict_rank3_curve(session, input_feature_matrix, training_config)[source]

Predict one full TE curve for a temporal rank-3 ONNX input.

Parameters:

  • session (onnxruntime.InferenceSession)

  • input_feature_matrix (ndarray)

  • training_config (dict[str, Any])

Return type:

ndarray

scripts.reports.analysis.build_track2_familywise_onnx_report.predict_curve(session, input_feature_matrix, training_config)[source]

Predict one TE curve with an ONNX Runtime session.

Parameters:

  • session (onnxruntime.InferenceSession)

  • input_feature_matrix (ndarray)

  • training_config (dict[str, Any])

Return type:

ndarray

scripts.reports.analysis.build_track2_familywise_onnx_report.select_deterministic_prediction_curve(raw_prediction_curve_deg, target_curve_deg, training_config, mdn_playback_channel)[source]

Select the deterministic playback channel from multi-output predictions.

Parameters:

  • raw_prediction_curve_deg (ndarray)

  • target_curve_deg (ndarray)

  • training_config (dict[str, Any])

  • mdn_playback_channel (str)

Return type:

ndarray

scripts.reports.analysis.build_track2_familywise_onnx_report.build_model_input_payload(curve_sample, session, training_config)[source]

Build input, target, and angular-position arrays for one model contract.

Parameters:

  • curve_sample (dict[str, Any])

  • session (onnxruntime.InferenceSession)

  • training_config (dict[str, Any])

Return type:

tuple[ndarray, ndarray, ndarray]

scripts.reports.analysis.build_track2_familywise_onnx_report.build_plot_payload(curve_sample, prediction_curve_deg, prediction_angular_position_deg)[source]

Build full measured and model-aligned prediction arrays for plotting.

Parameters:

  • curve_sample (dict[str, Any])

  • prediction_curve_deg (ndarray)

  • prediction_angular_position_deg (ndarray)

Return type:

tuple[ndarray, ndarray, ndarray, ndarray]

scripts.reports.analysis.build_track2_familywise_onnx_report.surface_accepts_direction(surface_name, direction_label)[source]

Return whether one model surface should evaluate one direction.

Parameters:

  • surface_name (str)

  • direction_label (str)

Return type:

bool

scripts.reports.analysis.build_track2_familywise_onnx_report.evaluate_model_entry(group_id, model_entry, provider_list, mdn_playback_channel)[source]

Evaluate one exported ONNX model over its valid test curves.

Parameters:

  • group_id (str)

  • model_entry (ExportedModelEntry)

  • provider_list (list[str])

  • mdn_playback_channel (str)

Return type:

tuple[list[CurveEvaluationEntry], dict[str, float], dict[str, Any]]

scripts.reports.analysis.build_track2_familywise_onnx_report.select_representative_curve_entries(curve_entry_list, curves_per_page)[source]

Select deterministic representative curves for one collage page.

Parameters:
Return type:

list[CurveEvaluationEntry]

scripts.reports.analysis.build_track2_familywise_onnx_report.save_surface_collage(collage_path, title_text, selected_curve_entry_list)[source]

Save one measured-versus-predicted curve collage.

Parameters:
Return type:

None

scripts.reports.analysis.build_track2_familywise_onnx_report.save_model_inventory_csv(csv_path, model_entry_list)[source]

Save the model inventory used by the report.

Parameters:
Return type:

None

scripts.reports.analysis.build_track2_familywise_onnx_report.save_per_curve_metrics_csv(csv_path, curve_summary_list)[source]

Save one per-curve metric table.

Parameters:

  • csv_path (Path)

  • curve_summary_list (list[dict[str, Any]])

Return type:

None

scripts.reports.analysis.build_track2_familywise_onnx_report.build_relative_markdown_path(target_path, markdown_directory)[source]

Build a Markdown-safe relative path.

Parameters:

  • target_path (Path)

  • markdown_directory (Path)

Return type:

str

scripts.reports.analysis.build_track2_familywise_onnx_report.append_model_inventory_table(report_line_list, model_summary_list)[source]

Append a compact model inventory table to the report.

Parameters:

  • report_line_list (list[str])

  • model_summary_list (list[dict[str, Any]])

Return type:

None

scripts.reports.analysis.build_track2_familywise_onnx_report.append_metric_table(report_line_list, model_summary_list)[source]

Append the aggregate metrics table.

Parameters:

  • report_line_list (list[str])

  • model_summary_list (list[dict[str, Any]])

Return type:

None

scripts.reports.analysis.build_track2_familywise_onnx_report.build_report_markdown(report_path, model_family, group_summary_list, summary_path, model_inventory_csv_path, per_curve_metrics_csv_path, output_directory, curves_per_page, mdn_playback_channel)[source]

Build the Markdown report body.

Parameters:

  • report_path (Path)

  • model_family (str)

  • group_summary_list (list[dict[str, Any]])

  • summary_path (Path)

  • model_inventory_csv_path (Path)

  • per_curve_metrics_csv_path (Path)

  • output_directory (Path)

  • curves_per_page (int)

  • mdn_playback_channel (str)

Return type:

str

scripts.reports.analysis.build_track2_familywise_onnx_report.run_familywise_onnx_report(arguments)[source]

Run familywise ONNX evaluation and report generation.

Parameters:

arguments (Namespace)

Return type:

dict[str, Any]

scripts.reports.analysis.build_track2_familywise_onnx_report.main()[source]

Run the command-line entry point.

Return type:

None