cminst/imslp-pdf-index

IMSLP PDF Index

This dataset is the canonical PDF-level index for the ReScore IMSLP PDF collection. It contains one row per unique IMSLP PDF and points to the PDF payload stored in cminst/imslp-raw-pdf-collection.

The PDF payload repository is append-only and may contain duplicate rows from retry launches. This index is deduplicated by imslp_id; duplicate content was validated to have identical SHA256, byte size, and page count before publishing.

Summary

• —Dataset repo: cminst/imslp-pdf-index
• —Source PDF archive repo: cminst/imslp-raw-pdf-collection
• —Collection id: imslp_pdf_collection
• —Rows: 44,592 PDFs
• —Pages: 650,464 PDF pages
• —Rows with page_count == 0: 34
• —PDF bytes: 223,682,096,447
• —Source manifest rows before dedupe: 45,172
• —Duplicate extra rows removed: 580

Columns

• —imslp_id: IMSLP file id.
• —work_title, work_url, special_url, description, file_title, source: IMSLP/discovery metadata.
• —page_count, page_count_is_positive, byte_size, sha256: validated PDF metadata.
• —hf_pdf_path: hf:// pointer to the PDF tar member in the source archive.
• —pdf_archive_member: member name inside the PDF tar shard.
• —hf_pdf_shard_path, hf_manifest_path, artifact_manifest_path: source archive provenance.
• —collection_id, profile_name, launch_id, worker_id: collection provenance.
• —completed_at, completed_record_path, modal_pdf_path: original worker provenance.
• —metadata_json, artifacts_json: extra publisher metadata as JSON strings.