coco_pipe.report.foundation#

Reports for foundation-model extraction and decoding workflows.

Functions#

make_foundation_embedding_report(records[, title, ...])

Build a dataset-level extraction coverage and provenance report.

make_foundation_decoding_report(records, *[, ...])

Build the foundation decoding sweep report.

Module Contents#

coco_pipe.report.foundation.make_foundation_embedding_report(records, title='Foundation Embedding Extraction', config=None, output_path=None, asset_urls='inline')#

Build a dataset-level extraction coverage and provenance report.

Parameters:
Return type:

coco_pipe.report.core.Report

coco_pipe.report.foundation.make_foundation_decoding_report(records, *, capability_records=(), title='Foundation Model Decoding', dataset_name='dataset', strategy_note=None, group_by=('condition', 'target'), per_result_sections='compact', include_hp_tuning=True, frame=None, config=None, output_path=None, asset_urls='inline')#

Build the foundation decoding sweep report.

Renders the shared sweep skeleton for a foundation sweep: scientific overview, a preflight capability matrix, per-train-mode leaderboards, the model/train-mode comparison figures + per-result diagnostics, and a failures section. records are the per-model foundation decoding rows; capability_records are the preflight model/train-mode decisions. Pass a pre-built frame (from prepare_sweep_frame()) to avoid re-preparing it when the caller already has one.

Parameters:
Return type:

coco_pipe.report.core.Report