Publish completed release report, anonymous preparation checks, and tested recovery tools
Browse files- COMPLETED_PUBLIC_PREPARATION.json +16 -0
- FINAL_REPORT.json +44 -0
- FINAL_REPORT.md +16 -0
COMPLETED_PUBLIC_PREPARATION.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"repo": "espnet/Bagpiper_SFT_Data_ESPnet",
|
| 4 |
+
"revision": "4a258e780a1697be5b6476a14edd7c257d8df350",
|
| 5 |
+
"downloaded_helper": true,
|
| 6 |
+
"anonymous": true,
|
| 7 |
+
"completed_release_preparation": true
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"repo": "espnet/Bagpiper_TTS_SFT_Data_ESPnet",
|
| 11 |
+
"revision": "821f04a8a4a6f4aaccc74ef7d7d39d54bc12d759",
|
| 12 |
+
"downloaded_helper": true,
|
| 13 |
+
"anonymous": true,
|
| 14 |
+
"completed_release_preparation": true
|
| 15 |
+
}
|
| 16 |
+
]
|
FINAL_REPORT.json
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"state": "complete",
|
| 3 |
+
"completed_at": "2026-10-01T01:31:49.992586+00:00",
|
| 4 |
+
"releases": [
|
| 5 |
+
{
|
| 6 |
+
"repository": "espnet/Bagpiper_SFT_Data_ESPnet",
|
| 7 |
+
"rows": 2667017,
|
| 8 |
+
"shards": 6804,
|
| 9 |
+
"archive_bytes": 719295381339,
|
| 10 |
+
"unique_audio_files": 1313767,
|
| 11 |
+
"unique_audio_scope": "partition_and_sha256",
|
| 12 |
+
"source_revision": "7dec4d68dc30e442368914619668fcd695a897f7",
|
| 13 |
+
"verified_archives_revision": "ddc7d626fecf295a0cc6c71d0515aa2e8bd2a0bf"
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"repository": "espnet/Bagpiper_TTS_SFT_Data_ESPnet",
|
| 17 |
+
"rows": 738123,
|
| 18 |
+
"shards": 1151,
|
| 19 |
+
"archive_bytes": 252559973183,
|
| 20 |
+
"unique_audio_files": 734969,
|
| 21 |
+
"unique_audio_scope": "partition_and_sha256",
|
| 22 |
+
"source_revision": "f0eb112765d33f30ff6136a6fa5feaa84ca79880",
|
| 23 |
+
"verified_archives_revision": "589d441d96e842087a64e6f3a120e4bc78130e16"
|
| 24 |
+
}
|
| 25 |
+
],
|
| 26 |
+
"validation": {
|
| 27 |
+
"source_and_remote_hashes": "all shards and archives verified",
|
| 28 |
+
"public_download_rows_compared": 5932,
|
| 29 |
+
"public_decoded_waveforms_compared": 12,
|
| 30 |
+
"all_partition_sample_rows_compared": 3814,
|
| 31 |
+
"all_partition_decoded_waveforms_compared": 24,
|
| 32 |
+
"converter_regression_tests_passed": 7,
|
| 33 |
+
"published_helpers_anonymously_downloaded_and_prepared": true,
|
| 34 |
+
"gpu_training_or_inference_validated": false
|
| 35 |
+
},
|
| 36 |
+
"preservation": {
|
| 37 |
+
"all_source_rows": true,
|
| 38 |
+
"original_audio_bytes": true,
|
| 39 |
+
"text_and_role_order": true,
|
| 40 |
+
"source_provenance": true,
|
| 41 |
+
"original_repositories_retained": true
|
| 42 |
+
},
|
| 43 |
+
"general_source_selection": "public recipe-v1; paper-v2 selection unchanged"
|
| 44 |
+
}
|
FINAL_REPORT.md
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Completed Bagpiper SFT format conversion
|
| 2 |
+
|
| 3 |
+
Both original SFT datasets are fully converted and publicly uploaded in portable lossless ESPnet dialogue format.
|
| 4 |
+
|
| 5 |
+
| Release | Preserved rows | Verified archives | Converted archive bytes |
|
| 6 |
+
| --- | ---: | ---: | ---: |
|
| 7 |
+
| [Bagpiper SFT](https://huggingface.co/datasets/espnet/Bagpiper_SFT_Data_ESPnet) | 2,667,017 | 6,804 | 719,295,381,339 |
|
| 8 |
+
| [Bagpiper-TTS SFT](https://huggingface.co/datasets/espnet/Bagpiper_TTS_SFT_Data_ESPnet) | 738,123 | 1,151 | 252,559,973,183 |
|
| 9 |
+
|
| 10 |
+
See FINAL_REPORT.json and each release's RELEASE_VALIDATION.json for precise statistics and pinned source revisions. Original encoded audio bytes, all rows, ordered text and roles, and provenance are preserved. Duplicate audio storage is removed within conversion streams; extraction verifies and merges identical files across streams. Original source repositories remain available. Their data cards now link directly to the converted training format.
|
| 11 |
+
|
| 12 |
+
The publicly downloaded preparation helper safely extracts audio, resolves local paths, checks global IDs and provenance linkage, and writes ESPnet all/train/valid manifests. The optional suggested validation split uses audio SHA256; all.json preserves the original complete pool. Source data terms and recipe-v1/paper-v2 selection are unchanged.
|
| 13 |
+
|
| 14 |
+
All source and remote archive hashes were reconciled. Actual ESPnet reading and exact original-versus-converted waveforms passed for the documented samples. Seven regression tests pass. No GPU training or inference was performed.
|
| 15 |
+
|
| 16 |
+
ESPnet recipe preparation documentation is ready as local commit 22666931398cecebf7c0646ca091b26fdacc559d and recipe-preparation.patch. GitHub publication was blocked by automatic approval review because the saved credential source was not specifically authorized; the permission question remains pending. This does not affect either completed Hugging Face release.
|