Release [run release-20261112T130049Z]

#37
by Battam - opened
This view is limited to 50 files because it contains too many changes. See the raw diff here.
Files changed (50) hide show
  1. CHANGELOG.md +21 -0
  2. README.md +22 -21
  3. README.zh.md +22 -21
  4. campaign_files/2026-10-a-spreadsheets-with-answers-2/unimi-dataverse-workbooks-cc0/README.md +0 -2
  5. campaign_files/2026-10-a-spreadsheets-with-answers-2/us-cbo-github-workbooks/README.md +0 -2
  6. campaign_files/2026-10-a-spreadsheets-with-answers-2/us-federal-agency-workbooks/README.md +0 -2
  7. campaign_files/2026-10-a-spreadsheets-with-answers-2/utk-arec-extension-budgets/README.md +0 -2
  8. campaign_files/2026-10-b-disk-images-3/uralcup-2026-forensics2/README.md +0 -3
  9. campaign_files/2026-10-b-disk-images-3/virtdisk-vhdx-process-kill/README.md +0 -4
  10. campaign_files/2026-10-c-repros-2/voqc-repro/README.md +0 -2
  11. campaign_files/2026-10-c-repros-2/wangsw-privacyamplification-repro/README.md +0 -2
  12. campaign_files/2026-10-g-solver-instances-2/weise-jssp-instances-and-results/README.md +0 -4
  13. campaign_files/2026-10-hw-eda-tests-2/universal-g-code-sender-ugs-core-fixture-tests-master-at-commit-5e43f9d3/README.md +0 -16
  14. campaign_files/2026-10-hw-eda-tests-2/uvvm/README.md +0 -6
  15. campaign_files/2026-10-hw-eda-tests-2/verilog-axi/README.md +0 -16
  16. campaign_files/2026-10-hw-eda-tests-2/verilog-axis/README.md +0 -8
  17. campaign_files/2026-10-hw-eda-tests-2/verilog-ethernet/README.md +0 -6
  18. campaign_files/2026-10-hw-eda-tests-2/verilog-lfsr/README.md +0 -4
  19. campaign_files/2026-10-hw-eda-tests-2/verilog-uart/README.md +0 -4
  20. campaign_files/2026-10-hw-eda-tests-2/vermaete-ipxact2systemverilog-example-directory/README.md +0 -6
  21. campaign_files/2026-10-hw-eda-tests-2/vunit/README.md +0 -8
  22. campaign_files/2026-10-m-issue-trackers-2/unilink-passenger-gtfs/README.md +0 -4
  23. campaign_files/2026-10-m-issue-trackers-2/victoria-clipper-gtfs/README.md +0 -4
  24. campaign_files/2026-10-m-issue-trackers-2/vinyl-misc/README.md +0 -4
  25. campaign_files/2026-10-m-issue-trackers-2/vscode-github-issues/README.md +0 -4
  26. campaign_files/2026-10-m-issue-trackers-2/waikato-busit-gtfs/README.md +0 -4
  27. campaign_files/2026-10-m-issue-trackers-2/warrington-passenger-gtfs/README.md +0 -4
  28. campaign_files/2026-10-md-media-and-documents/unstructured/README.md +0 -4
  29. campaign_files/2026-10-md-media-and-documents/voa-learning-english-3160704/README.md +0 -4
  30. campaign_files/2026-10-md-media-and-documents/voa-learning-english-7932530/README.md +0 -4
  31. campaign_files/2026-10-md-media-and-documents/voa-learning-english-7966519/README.md +0 -4
  32. campaign_files/2026-10-md-media-and-documents/voa-learning-english-7979814/README.md +0 -4
  33. campaign_files/2026-10-md-media-and-documents/voa-learning-english-7980848/README.md +0 -4
  34. campaign_files/2026-10-md-media-and-documents/voa-learning-english-a-short-history-of-spacecraft-landings-on-the-mo/README.md +0 -4
  35. campaign_files/2026-10-md-media-and-documents/voa-learning-english-american-inventors-rocket-scientist-robert-godda/README.md +0 -4
  36. campaign_files/2026-10-md-media-and-documents/voa-learning-english-american-places/README.md +0 -4
  37. campaign_files/2026-10-md-media-and-documents/voa-learning-english-scientists-discover-stonehenge-like-circle-in-de/README.md +0 -4
  38. campaign_files/2026-10-md-media-and-documents/voa-learning-english-watching-the-grass-grow-is-not-fun/README.md +0 -4
  39. campaign_files/2026-10-md-media-and-documents/voa-learning-english-wilbur-and-orville-wright-the-first-airplane/README.md +0 -4
  40. campaign_files/2026-10-md-media-and-documents/voa-learning-english-yellowstone/README.md +0 -4
  41. campaign_files/2026-10-md-media-and-documents/voxforge-16khz-16bit-main/README.md +0 -6
  42. campaign_files/2026-10-ml-serving-configs-2/vllm-gsm8k-server-gate/README.md +0 -4
  43. campaign_files/2026-10-ml-serving-configs-2/vllm-lm-eval-gate/README.md +0 -12
  44. campaign_files/2026-10-ml-serving-configs-2/vllm-tool-and-reasoning-parser-test-expectations/README.md +0 -22
  45. campaign_files/2026-10-ml-serving-configs-2/vlosing-driftdatasets/README.md +0 -8
  46. campaign_files/2026-10-op-rule-calculations-2/uscis-aao-non-precedent-decisions/README.md +0 -10
  47. campaign_files/2026-10-s-security-detection-2/wazuh-ruleset-tests/README.md +0 -16
  48. campaign_files/2026-10-s-security-detection/velociraptor-sigma-golden-test/README.md +0 -4
  49. campaign_files/2026-10-s-security-detection/virustotal-yara-x-tests/README.md +0 -4
  50. campaign_files/2026-10-s-security-detection/wazuh-ruleset-logtest/README.md +0 -4
CHANGELOG.md CHANGED
@@ -3480,3 +3480,24 @@ Changes to the content of this index, oldest first. There is one entry for each
3480
  - scotland-government-workbooks: changed pointer group file data/scotland-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 2 key(s) of the old version are in no data file of source scotland-government-workbooks after the commit (row 3 from batch 2026-10-a-spreadsheets-with-answers-2, row 6 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3481
  - stackexchange-api-posts: refused
3482
  - uk-government-workbooks: changed pointer group file data/uk-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 1 key(s) of the old version are in no data file of source uk-government-workbooks after the commit (row 22 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3480
  - scotland-government-workbooks: changed pointer group file data/scotland-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 2 key(s) of the old version are in no data file of source scotland-government-workbooks after the commit (row 3 from batch 2026-10-a-spreadsheets-with-answers-2, row 6 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3481
  - stackexchange-api-posts: refused
3482
  - uk-government-workbooks: changed pointer group file data/uk-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 1 key(s) of the old version are in no data file of source uk-government-workbooks after the commit (row 22 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3483
+
3484
+ ## 2026-11-12: sources held back
3485
+
3486
+ - Each source below keeps its newest version that passed every check; its newer batch waits until it passes them:
3487
+ - borealis-dataverse-workbooks-cc-by: changed pointer group file data/borealis-dataverse-workbooks-cc-by/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source borealis-dataverse-workbooks-cc-by after the commit (row 4 from batch 2026-10-a-spreadsheets-with-answers, row 5 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3488
+ - borealis-dataverse-workbooks-cc0: changed pointer group file data/borealis-dataverse-workbooks-cc0/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source borealis-dataverse-workbooks-cc0 after the commit (row 7 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3489
+ - cfpb-complaints: refused
3490
+ - data-gov-uk-workbooks: changed pointer group file data/data-gov-uk-workbooks/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 4 key(s) of the old version are in no data file of source data-gov-uk-workbooks after the commit (row 5 from batch 2026-10-a-spreadsheets-with-answers, row 6 from batch 2026-10-a-spreadsheets-with-answers, row 12 from batch 2026-10-a-spreadsheets-with-answers, 1 more); the removal record holds none of them, and none fails a publish check now
3491
+ - data-gov-uk-workbooks-noassertion: changed pointer group file data/data-gov-uk-workbooks-noassertion/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source data-gov-uk-workbooks-noassertion after the commit (row 6 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3492
+ - dataversenl-workbooks-cc-by: changed pointer group file data/dataversenl-workbooks-cc-by/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source dataversenl-workbooks-cc-by after the commit (row 17 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3493
+ - dataversenl-workbooks-cc0: changed pointer group file data/dataversenl-workbooks-cc0/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 3 key(s) of the old version are in no data file of source dataversenl-workbooks-cc0 after the commit (row 7 from batch 2026-10-a-spreadsheets-with-answers, row 13 from batch 2026-10-a-spreadsheets-with-answers, row 16 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3494
+ - dvn-fvhlgg-repro: changed pointer group file data/dvn-fvhlgg-repro/2026-10-c-repros/part-00000.parquet: 3 key(s) of the old version are in no data file of source dvn-fvhlgg-repro after the commit (row 5 from batch 2026-10-c-repros, row 7 from batch 2026-10-c-repros, row 16 from batch 2026-10-c-repros); the removal record holds none of them, and none fails a publish check now
3495
+ - eu-institution-workbooks-cc-by: changed pointer group file data/eu-institution-workbooks-cc-by/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 3 key(s) of the old version are in no data file of source eu-institution-workbooks-cc-by after the commit (row 2 from batch 2026-10-a-spreadsheets-with-answers, row 4 from batch 2026-10-a-spreadsheets-with-answers, row 5 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3496
+ - hdx-workbooks-cc0: changed pointer group file data/hdx-workbooks-cc0/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source hdx-workbooks-cc0 after the commit (row 4 from batch 2026-10-a-spreadsheets-with-answers, row 5 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3497
+ - hdx-workbooks-odbl: changed pointer group file data/hdx-workbooks-odbl/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source hdx-workbooks-odbl after the commit (row 4 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3498
+ - hf-training-records: refused
3499
+ - npm: refused
3500
+ - scotland-government-workbooks: changed pointer group file data/scotland-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 2 key(s) of the old version are in no data file of source scotland-government-workbooks after the commit (row 3 from batch 2026-10-a-spreadsheets-with-answers-2, row 6 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3501
+ - stackexchange-api-posts: refused
3502
+ - uk-government-workbooks: changed pointer group file data/uk-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 1 key(s) of the old version are in no data file of source uk-government-workbooks after the commit (row 22 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3503
+ - victoria-government-workbooks: changed pointer group file data/victoria-government-workbooks/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source victoria-government-workbooks after the commit (row 6 from batch 2026-10-a-spreadsheets-with-answers, row 7 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
README.md CHANGED
@@ -35,18 +35,18 @@ this dataset is the foundation's public index.
35
  English | [简体中文](README.zh.md) · [Data page](https://huggingface.co/spaces/Battam/training-data-foundation) · [Usage video](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
36
 
37
  <picture class="not-prose">
38
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-narrow-light.svg">
39
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-wide-light.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,763 material sets, from 1,820 upstream sources, 1,411 allowed for training.">
40
  </picture>
41
  <picture class="not-prose">
42
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-narrow-dark.svg">
43
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-wide-dark.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,763 material sets, from 1,820 upstream sources, 1,411 allowed for training.">
44
  </picture>
45
 
46
  ## Get data in one command
47
 
48
  ```sh
49
- uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/tools/fetch.py --form spreadsheets --sample 3 --out ./data
50
  ```
51
 
52
  This fetches 3 units of every spreadsheet material set into `./data`, one folder per unit, and checks each file against the size and hashes its row records.
@@ -56,7 +56,7 @@ the same command takes the same units of this release.
56
  <details>
57
  <summary>What the command takes, and in which order</summary>
58
 
59
- The command and the queries below are pinned to the release tag `release-2026-11-12.8`.
60
  [uv](https://docs.astral.sh/uv/) reads the script's dependencies, at the versions published before the day of this release.
61
 
62
  The command returns units: 3 units from each material set whose form is `spreadsheets`, each in a folder of its own named after its `unit_id`.
@@ -84,7 +84,7 @@ Choose material sets from the catalog with [DuckDB](https://duckdb.org/):
84
 
85
  ```sql
86
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
87
- FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.8/meta/catalog.parquet'
88
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
89
  ORDER BY collection, units DESC;
90
  ```
@@ -95,7 +95,7 @@ with `kind = 'corpus'` it returns corpora instead, counted in rows (`files`).
95
  Choose files in one corpus:
96
 
97
  ```sql
98
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.8/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.8/meta/catalog.parquet' WHERE source = 'if-archive'));
99
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
100
  FROM read_parquet(getvariable('files'))
101
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
@@ -111,12 +111,12 @@ A data file may hold the rows of several sources, so the query keeps this one's
111
  ## What's inside
112
 
113
  <picture class="not-prose">
114
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-narrow-light.svg">
115
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-wide-light.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 33.">
116
  </picture>
117
  <picture class="not-prose">
118
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-narrow-dark.svg">
119
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-wide-dark.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 33.">
120
  </picture>
121
 
122
  <details>
@@ -138,7 +138,7 @@ A data file may hold the rows of several sources, so the query keeps this one's
138
  | Unclassified | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
139
 
140
  Each source is counted under its main domain, the top-level domain of its first domain code.
141
- Every size is deduplicated, and every count is what a stage keeps after the rules every stage shares ([How it is made in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual)).
142
 
143
  </details>
144
 
@@ -148,21 +148,22 @@ Every material set and every corpus, with its counts, size, license tier and log
148
  ## How it is made
149
 
150
  <picture class="not-prose">
151
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-narrow-light.svg">
152
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-wide-light.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.8.">
153
  </picture>
154
  <picture class="not-prose">
155
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-narrow-dark.svg">
156
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-wide-dark.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.8.">
157
  </picture>
158
 
159
  Collectors read whole public upstreams into corpora, the material of pre-training and mid-training, and admission turns works found upstream and chosen corpus rows into material sets, the material of post-training.
160
  Before each release every source is checked for its license, personal data, overlap with benchmarks and sizes, and each release is one tagged commit whose card and figures are computed from its own files.
161
- The glossary, each step, how sizes and the stage counts are measured, and the checks of this release: [How it is made in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual).
 
162
 
163
  ## Reference
164
 
165
- Every page and file of this dataset, the fields used most, Croissant metadata, the Dataset Viewer and the SQL Console: [Pages and files of the dataset in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual).
166
 
167
  ## License, privacy and limits
168
 
@@ -172,7 +173,7 @@ On the data page, the Training filter chooses rows by their permit, and each sou
172
  Rows that copy a public benchmark task, and rows not compared with the benchmark list yet, are left out of every training stage and of the fetch tool's default;
173
  only `--allow-benchmark` and `--allow-not-compared` take them.
174
  The catalog itself is released under CC0 1.0, and each file keeps the license of its source.
175
- The full text, with the license details, personal information read from sample files, known limits, uses, releases and removal requests: [License, privacy and limits in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual).
176
 
177
  ## Citation
178
 
@@ -180,7 +181,7 @@ The full text, with the license details, personal information read from sample f
180
  @misc{training_data_foundation,
181
  title = {training-data-foundation: an index of public materials for building agent environments},
182
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
183
- note = {Release release-2026-11-12.8},
184
  year = {2026}
185
  }
186
  ```
 
35
  English | [简体中文](README.zh.md) · [Data page](https://huggingface.co/spaces/Battam/training-data-foundation) · [Usage video](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
36
 
37
  <picture class="not-prose">
38
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-light.svg">
39
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-light.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,763 material sets, from 1,820 upstream sources, 1,411 allowed for training.">
40
  </picture>
41
  <picture class="not-prose">
42
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-dark.svg">
43
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-dark.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,763 material sets, from 1,820 upstream sources, 1,411 allowed for training.">
44
  </picture>
45
 
46
  ## Get data in one command
47
 
48
  ```sh
49
+ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/tools/fetch.py --form spreadsheets --sample 3 --out ./data
50
  ```
51
 
52
  This fetches 3 units of every spreadsheet material set into `./data`, one folder per unit, and checks each file against the size and hashes its row records.
 
56
  <details>
57
  <summary>What the command takes, and in which order</summary>
58
 
59
+ The command and the queries below are pinned to the release tag `release-2026-11-12.9`.
60
  [uv](https://docs.astral.sh/uv/) reads the script's dependencies, at the versions published before the day of this release.
61
 
62
  The command returns units: 3 units from each material set whose form is `spreadsheets`, each in a folder of its own named after its `unit_id`.
 
84
 
85
  ```sql
86
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
87
+ FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet'
88
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
89
  ORDER BY collection, units DESC;
90
  ```
 
95
  Choose files in one corpus:
96
 
97
  ```sql
98
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'if-archive'));
99
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
100
  FROM read_parquet(getvariable('files'))
101
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
 
111
  ## What's inside
112
 
113
  <picture class="not-prose">
114
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-light.svg">
115
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-light.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 33.">
116
  </picture>
117
  <picture class="not-prose">
118
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-dark.svg">
119
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-dark.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 33.">
120
  </picture>
121
 
122
  <details>
 
138
  | Unclassified | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
139
 
140
  Each source is counted under its main domain, the top-level domain of its first domain code.
141
+ Every size is deduplicated, and every count is what a stage keeps after the rules every stage shares ([How it is made in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual/stage-counting)).
142
 
143
  </details>
144
 
 
148
  ## How it is made
149
 
150
  <picture class="not-prose">
151
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-light.svg">
152
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-light.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.9.">
153
  </picture>
154
  <picture class="not-prose">
155
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-dark.svg">
156
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-dark.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.9.">
157
  </picture>
158
 
159
  Collectors read whole public upstreams into corpora, the material of pre-training and mid-training, and admission turns works found upstream and chosen corpus rows into material sets, the material of post-training.
160
  Before each release every source is checked for its license, personal data, overlap with benchmarks and sizes, and each release is one tagged commit whose card and figures are computed from its own files.
161
+ Each step, and how sizes and the stage counts are measured: [How it is made in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual/made);
162
+ the glossary and the checks of this release: [Words and numbers in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual/words).
163
 
164
  ## Reference
165
 
166
+ Every page and file of this dataset, the fields used most, Croissant metadata, the Dataset Viewer and the SQL Console: [Pages and files of the dataset in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual/dataset).
167
 
168
  ## License, privacy and limits
169
 
 
173
  Rows that copy a public benchmark task, and rows not compared with the benchmark list yet, are left out of every training stage and of the fetch tool's default;
174
  only `--allow-benchmark` and `--allow-not-compared` take them.
175
  The catalog itself is released under CC0 1.0, and each file keeps the license of its source.
176
+ The full text, with the license details, personal information read from sample files, known limits, uses, releases and removal requests: [License, privacy and limits in the data page's manual](https://battam-training-data-foundation.static.hf.space/index.html?lng=en#manual/license).
177
 
178
  ## Citation
179
 
 
181
  @misc{training_data_foundation,
182
  title = {training-data-foundation: an index of public materials for building agent environments},
183
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
184
+ note = {Release release-2026-11-12.9},
185
  year = {2026}
186
  }
187
  ```
README.zh.md CHANGED
@@ -22,18 +22,18 @@ tags:
22
  [English](README.md) | 简体中文 · [数据页](https://huggingface.co/spaces/Battam/training-data-foundation) · [使用示例视频](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
23
 
24
  <picture class="not-prose">
25
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-narrow-light.zh.svg">
26
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-wide-light.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,763,来自 1,820 个上游来源,其中 1,411 允许用于训练。">
27
  </picture>
28
  <picture class="not-prose">
29
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-narrow-dark.zh.svg">
30
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/stages-wide-dark.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,763,来自 1,820 个上游来源,其中 1,411 允许用于训练。">
31
  </picture>
32
 
33
  ## 一条命令取数据
34
 
35
  ```sh
36
- uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/tools/fetch.py --form spreadsheets --sample 3 --out ./data
37
  ```
38
 
39
  这条命令从每个电子表格材料集各取 3 个单元放进 `./data`,每个单元一个文件夹,并按每个文件所在行记录的大小和哈希值校验它。
@@ -43,7 +43,7 @@ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/r
43
  <details>
44
  <summary>命令取什么、按什么顺序取</summary>
45
 
46
- 这条命令和下面的查询都固定在发布标签 `release-2026-11-12.8` 上。
47
  [uv](https://docs.astral.sh/uv/) 会读取脚本声明的依赖,取这个发布当天之前发布的版本。
48
 
49
  命令返回单元:形式为 `spreadsheets` 的每个材料集各 3 个单元,每个单元放在以它的 `unit_id` 命名的单独文件夹里。
@@ -71,7 +71,7 @@ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/r
71
 
72
  ```sql
73
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
74
- FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.8/meta/catalog.parquet'
75
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
76
  ORDER BY collection, units DESC;
77
  ```
@@ -82,7 +82,7 @@ ORDER BY collection, units DESC;
82
  在一个语料库里挑选文件:
83
 
84
  ```sql
85
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.8/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.8/meta/catalog.parquet' WHERE source = 'if-archive'));
86
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
87
  FROM read_parquet(getvariable('files'))
88
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
@@ -98,12 +98,12 @@ ORDER BY size_bytes LIMIT 20;
98
  ## 内容概览
99
 
100
  <picture class="not-prose">
101
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-narrow-light.zh.svg">
102
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-wide-light.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 33。">
103
  </picture>
104
  <picture class="not-prose">
105
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-narrow-dark.zh.svg">
106
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/inside-wide-dark.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 33。">
107
  </picture>
108
 
109
  <details>
@@ -125,7 +125,7 @@ ORDER BY size_bytes LIMIT 20;
125
  | 未分类 | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
126
 
127
  每个来源计入它的主要领域,即它的第一个领域代码所属的顶层领域。
128
- 每个大小都是去重后的大小,每个计数都是一个阶段在各阶段共用的规则之后保留下来的部分([数据页使用手册的“怎样做成”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual))。
129
 
130
  </details>
131
 
@@ -135,21 +135,22 @@ ORDER BY size_bytes LIMIT 20;
135
  ## 怎样做成
136
 
137
  <picture class="not-prose">
138
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-narrow-light.zh.svg">
139
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-wide-light.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.8。">
140
  </picture>
141
  <picture class="not-prose">
142
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-narrow-dark.zh.svg">
143
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.8/docs/figures/steps-wide-dark.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.8。">
144
  </picture>
145
 
146
  采集器整批读取完整的公开上游,记成语料库,供预训练和中训练使用,收录则把检索在上游找到的作品和选出的语料库行做成材料集,供后训练使用。
147
  每次发布前,每个来源都要核对许可证、个人数据、与基准测试的重叠和大小,每次发布是一个带标签的提交,它的卡片和图都由它自己的文件算出。
148
- 术语、每一步、大小和阶段计数怎样测得,以及本次发布的核对结果:[数据页使用手册的“怎样做成”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual)。
 
149
 
150
  ## 查阅
151
 
152
- 本数据集的每个页面和文件、最常用的字段、Croissant 元数据、数据集查看器和 SQL 控制台:[数据页使用手册的“数据集的页面和文件”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual)。
153
 
154
  ## 许可、隐私与局限
155
 
@@ -159,7 +160,7 @@ ORDER BY size_bytes LIMIT 20;
159
  复制了公开基准测试任务的行,以及还没有与基准测试清单比对的行,不计入任何训练阶段,下载工具默认也不取;
160
  只有 `--allow-benchmark` 和 `--allow-not-compared` 会取它们。
161
  目录本身以 CC0 1.0 发布,每个文件保留其来源的许可证。
162
- 完整说明,包括许可详情、从样本文件读出的个人信息、已知局限、用途、发布和移除请求:[数据页使用手册的“许可、隐私与局限”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual)。
163
 
164
  ## 引用
165
 
@@ -167,7 +168,7 @@ ORDER BY size_bytes LIMIT 20;
167
  @misc{training_data_foundation,
168
  title = {training-data-foundation: an index of public materials for building agent environments},
169
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
170
- note = {Release release-2026-11-12.8},
171
  year = {2026}
172
  }
173
  ```
 
22
  [English](README.md) | 简体中文 · [数据页](https://huggingface.co/spaces/Battam/training-data-foundation) · [使用示例视频](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
23
 
24
  <picture class="not-prose">
25
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-light.zh.svg">
26
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-light.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,763,来自 1,820 个上游来源,其中 1,411 允许用于训练。">
27
  </picture>
28
  <picture class="not-prose">
29
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-dark.zh.svg">
30
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-dark.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,763,来自 1,820 个上游来源,其中 1,411 允许用于训练。">
31
  </picture>
32
 
33
  ## 一条命令取数据
34
 
35
  ```sh
36
+ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/tools/fetch.py --form spreadsheets --sample 3 --out ./data
37
  ```
38
 
39
  这条命令从每个电子表格材料集各取 3 个单元放进 `./data`,每个单元一个文件夹,并按每个文件所在行记录的大小和哈希值校验它。
 
43
  <details>
44
  <summary>命令取什么、按什么顺序取</summary>
45
 
46
+ 这条命令和下面的查询都固定在发布标签 `release-2026-11-12.9` 上。
47
  [uv](https://docs.astral.sh/uv/) 会读取脚本声明的依赖,取这个发布当天之前发布的版本。
48
 
49
  命令返回单元:形式为 `spreadsheets` 的每个材料集各 3 个单元,每个单元放在以它的 `unit_id` 命名的单独文件夹里。
 
71
 
72
  ```sql
73
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
74
+ FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet'
75
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
76
  ORDER BY collection, units DESC;
77
  ```
 
82
  在一个语料库里挑选文件:
83
 
84
  ```sql
85
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'if-archive'));
86
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
87
  FROM read_parquet(getvariable('files'))
88
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
 
98
  ## 内容概览
99
 
100
  <picture class="not-prose">
101
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-light.zh.svg">
102
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-light.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 33。">
103
  </picture>
104
  <picture class="not-prose">
105
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-dark.zh.svg">
106
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-dark.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 33。">
107
  </picture>
108
 
109
  <details>
 
125
  | 未分类 | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
126
 
127
  每个来源计入它的主要领域,即它的第一个领域代码所属的顶层领域。
128
+ 每个大小都是去重后的大小,每个计数都是一个阶段在各阶段共用的规则之后保留下来的部分([数据页使用手册的“怎样做成”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual/stage-counting))。
129
 
130
  </details>
131
 
 
135
  ## 怎样做成
136
 
137
  <picture class="not-prose">
138
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-light.zh.svg">
139
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-light.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.9。">
140
  </picture>
141
  <picture class="not-prose">
142
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-dark.zh.svg">
143
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-dark.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.9。">
144
  </picture>
145
 
146
  采集器整批读取完整的公开上游,记成语料库,供预训练和中训练使用,收录则把检索在上游找到的作品和选出的语料库行做成材料集,供后训练使用。
147
  每次发布前,每个来源都要核对许可证、个人数据、与基准测试的重叠和大小,每次发布是一个带标签的提交,它的卡片和图都由它自己的文件算出。
148
+ 每一步,以及大小和��段计数怎样测得:[数据页使用手册的“怎样做成”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual/made);
149
+ 术语和本次发布的核对结果:[数据页使用手册的“术语与数字”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual/words)。
150
 
151
  ## 查阅
152
 
153
+ 本数据集的每个页面和文件、最常用的字段、Croissant 元数据、数据集查看器和 SQL 控制台:[数据页使用手册的“数据集的页面和文件”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual/dataset)。
154
 
155
  ## 许可、隐私与局限
156
 
 
160
  复制了公开基准测试任务的行,以及还没有与基准测试清单比对的行,不计入任何训练阶段,下载工具默认也不取;
161
  只有 `--allow-benchmark` 和 `--allow-not-compared` 会取它们。
162
  目录本身以 CC0 1.0 发布,每个文件保留其来源的许可证。
163
+ 完整说明,包括许可详情、从样本文件读出的个人信息、已知局限、用途、发布和移除请求:[数据页使用手册的“许可、隐私与局限”](https://battam-training-data-foundation.static.hf.space/index.html?lng=zh#manual/license)。
164
 
165
  ## 引用
166
 
 
168
  @misc{training_data_foundation,
169
  title = {training-data-foundation: an index of public materials for building agent environments},
170
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
171
+ note = {Release release-2026-11-12.9},
172
  year = {2026}
173
  }
174
  ```
campaign_files/2026-10-a-spreadsheets-with-answers-2/unimi-dataverse-workbooks-cc0/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained.
10
 
11
- **Answers:** Each file with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: key_formula_values.
12
-
13
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
14
 
15
  **Licence (SPDX):** CC0-1.0
 
8
 
9
  **Self-contained:** The file is self-contained.
10
 
 
 
11
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
12
 
13
  **Licence (SPDX):** CC0-1.0
campaign_files/2026-10-a-spreadsheets-with-answers-2/us-cbo-github-workbooks/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is self-contained. 6 of 6 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** Each file with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: key_formula_values.
12
-
13
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx); openpyxl (formulas and cached values, keep_vba not needed); LibreOffice headless recalculation (soffice --headless --convert-to xlsm).
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
8
 
9
  **Self-contained:** Each file is self-contained. 6 of 6 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx); openpyxl (formulas and cached values, keep_vba not needed); LibreOffice headless recalculation (soffice --headless --convert-to xlsm).
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
campaign_files/2026-10-a-spreadsheets-with-answers-2/us-federal-agency-workbooks/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
11
- **Answers:** Each file with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: key_formula_values.
12
-
13
  **Tools to open and check:** openpyxl (keep_vba=True) to read; openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
 
 
11
  **Tools to open and check:** openpyxl (keep_vba=True) to read; openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
campaign_files/2026-10-a-spreadsheets-with-answers-2/utk-arec-extension-budgets/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained.
10
 
11
- **Answers:** Each file with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: key_formula_values.
12
-
13
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
14
 
15
  **Licence (SPDX):** NOASSERTION
 
8
 
9
  **Self-contained:** The file is self-contained.
10
 
 
 
11
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
12
 
13
  **Licence (SPDX):** NOASSERTION
campaign_files/2026-10-b-disk-images-3/uralcup-2026-forensics2/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether the file needs other upstream files is not stated. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** Each file with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: deleted_files.
12
-
13
  **Tools to open and check:** Open with tools for these file types: .img.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -20,5 +18,4 @@ Files:
20
 
21
  - `tasks/baby/forensics2/public/flash_dump.img`: FAT12 floppy image with one deleted SECRET1.TXT and a slack-space note (UralCUP 2026 baby/forensics2).
22
  Origin: CTF image built by src/gen.py with mformat, mcopy and mdel, then a slack patch.
23
- Reference result: deleted_files
24
  Dependencies: only needed to rerun the exercise, not to grade the answer
 
8
 
9
  **Self-contained:** Whether the file needs other upstream files is not stated. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Open with tools for these file types: .img.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
18
 
19
  - `tasks/baby/forensics2/public/flash_dump.img`: FAT12 floppy image with one deleted SECRET1.TXT and a slack-space note (UralCUP 2026 baby/forensics2).
20
  Origin: CTF image built by src/gen.py with mformat, mcopy and mdel, then a slack patch.
 
21
  Dependencies: only needed to rerun the exercise, not to grade the answer
campaign_files/2026-10-b-disk-images-3/virtdisk-vhdx-process-kill/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each unit needs other upstream files is not stated. 2 of 2 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** Each unit with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: corruption_desc.
12
-
13
  **Tools to open and check:** Open with tools for these file types: .vhdx.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,12 +19,10 @@ Units:
21
  - `tests/fixtures/vhdx/windows-process-kill/512-child.vhdx`: Dirty-log VHDX differencing disk after the writer was killed at flush 14 (512-byte sectors), with its parent.
22
  Decompressed from: https://raw.githubusercontent.com/CaddyGlow/virtdisk/52878d44518126fbab593a4d99aee291aa282003/tests/fixtures/vhdx/windows-process-kill/512-child.vhdx.gz
23
  Origin: provenance.json records the capture
24
- Reference result: corruption_desc
25
  Companions (not included): tests/fixtures/vhdx/windows-process-kill/512-base.vhdx.gz
26
  Dependencies: gunzip or Python gzip to unpack; only needed to rerun, not to grade (grading compares the logical sha256 and byte evidence)
27
  - `tests/fixtures/vhdx/windows-process-kill/4096-child.vhdx`: Dirty-log VHDX differencing disk after the writer was killed at flush 14 (4096-byte sectors), with its parent.
28
  Decompressed from: https://raw.githubusercontent.com/CaddyGlow/virtdisk/52878d44518126fbab593a4d99aee291aa282003/tests/fixtures/vhdx/windows-process-kill/4096-child.vhdx.gz
29
  Origin: provenance.json records the capture
30
- Reference result: corruption_desc
31
  Companions (not included): tests/fixtures/vhdx/windows-process-kill/4096-base.vhdx.gz
32
  Dependencies: gunzip or Python gzip to unpack; only needed to rerun, not to grade (grading compares the logical sha256 and byte evidence)
 
8
 
9
  **Self-contained:** Whether each unit needs other upstream files is not stated. 2 of 2 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Open with tools for these file types: .vhdx.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `tests/fixtures/vhdx/windows-process-kill/512-child.vhdx`: Dirty-log VHDX differencing disk after the writer was killed at flush 14 (512-byte sectors), with its parent.
20
  Decompressed from: https://raw.githubusercontent.com/CaddyGlow/virtdisk/52878d44518126fbab593a4d99aee291aa282003/tests/fixtures/vhdx/windows-process-kill/512-child.vhdx.gz
21
  Origin: provenance.json records the capture
 
22
  Companions (not included): tests/fixtures/vhdx/windows-process-kill/512-base.vhdx.gz
23
  Dependencies: gunzip or Python gzip to unpack; only needed to rerun, not to grade (grading compares the logical sha256 and byte evidence)
24
  - `tests/fixtures/vhdx/windows-process-kill/4096-child.vhdx`: Dirty-log VHDX differencing disk after the writer was killed at flush 14 (4096-byte sectors), with its parent.
25
  Decompressed from: https://raw.githubusercontent.com/CaddyGlow/virtdisk/52878d44518126fbab593a4d99aee291aa282003/tests/fixtures/vhdx/windows-process-kill/4096-child.vhdx.gz
26
  Origin: provenance.json records the capture
 
27
  Companions (not included): tests/fixtures/vhdx/windows-process-kill/4096-base.vhdx.gz
28
  Dependencies: gunzip or Python gzip to unpack; only needed to rerun, not to grade (grading compares the logical sha256 and byte evidence)
campaign_files/2026-10-c-repros-2/voqc-repro/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each unit needs other upstream files is not stated. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** Each unit with an answer has one answer file, at answer/<package>/reported.json.
12
-
13
  **Tools to open and check:** Open with tools for these file types: .sh; .py; .md.
14
 
15
  **Licence (SPDX):** MIT
 
8
 
9
  **Self-contained:** Whether each unit needs other upstream files is not stated. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Open with tools for these file types: .sh; .py; .md.
12
 
13
  **Licence (SPDX):** MIT
campaign_files/2026-10-c-repros-2/wangsw-privacyamplification-repro/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether the unit needs other upstream files is not stated. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** Each unit with an answer has one answer file, at answer/<package>/reported.json.
12
-
13
  **Tools to open and check:** Open with tools for these file types: .py.
14
 
15
  **Licence (SPDX):** NOASSERTION
 
8
 
9
  **Self-contained:** Whether the unit needs other upstream files is not stated. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Open with tools for these file types: .py.
12
 
13
  **Licence (SPDX):** NOASSERTION
campaign_files/2026-10-g-solver-instances-2/weise-jssp-instances-and-results/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Best known makespan per instance (optimal where lower bound equals BKS). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Program: parse the agent's schedule (start time per operation), check precedence within each job and no overlap on each machine, compute the makespan and require it at most the inst.bks value of that instance.
14
 
15
  **Licence (SPDX):** NOASSERTION AND GPL-3.0-only
@@ -21,7 +19,5 @@ Files:
21
  - `data-raw/instance-data/instance_data.txt`: 242 job shop scheduling instances (abz, dmu, ft, la, orb, swv, ta, yn) with lower bounds and best known makespans.
22
  Format: plain text, OR-Library job shop format: per instance a line ' instance NAME', a description line, 'jobs machines', then one line per job of (machine, time) pairs; 242 cases in one file (line 1: "This file contains a set of 242 JSP test instances."); companion CSV gives per instance lower bound and BKS
23
  Origin: compiled from published JSSP benchmarks for Weise's optimisation teaching material
24
- Reference result: best known makespan per instance (optimal where lower bound equals BKS)
25
- Reference source: instances_with_bks.txt (CSV, header line 1: inst.id, inst.opt.bound.lower, inst.bks, references), e.g. line 2 abz5 lower 1234 bks 1234, line 160 ta01 lower 1231 bks 1231
26
  Check: Program: parse the agent's schedule (start time per operation), check precedence within each job and no overlap on each machine, compute the makespan and require it at most the inst.bks value of that instance
27
  Companions (not included): data-raw/instances/instances_with_bks.txt
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Program: parse the agent's schedule (start time per operation), check precedence within each job and no overlap on each machine, compute the makespan and require it at most the inst.bks value of that instance.
12
 
13
  **Licence (SPDX):** NOASSERTION AND GPL-3.0-only
 
19
  - `data-raw/instance-data/instance_data.txt`: 242 job shop scheduling instances (abz, dmu, ft, la, orb, swv, ta, yn) with lower bounds and best known makespans.
20
  Format: plain text, OR-Library job shop format: per instance a line ' instance NAME', a description line, 'jobs machines', then one line per job of (machine, time) pairs; 242 cases in one file (line 1: "This file contains a set of 242 JSP test instances."); companion CSV gives per instance lower bound and BKS
21
  Origin: compiled from published JSSP benchmarks for Weise's optimisation teaching material
 
 
22
  Check: Program: parse the agent's schedule (start time per operation), check precedence within each job and no overlap on each machine, compute the makespan and require it at most the inst.bks value of that instance
23
  Companions (not included): data-raw/instances/instances_with_bks.txt
campaign_files/2026-10-hw-eda-tests-2/universal-g-code-sender-ugs-core-fixture-tests-master-at-commit-5e43f9d3/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 7 of 7 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: processed_gcode. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
14
 
15
  **Licence (SPDX):** GPL-3.0-only
@@ -21,8 +19,6 @@ Files:
21
  - `ugs-core/test/resources/gcode/fixtures/arc/G17_G2_one_arc.input.nc`: One clockwise half circle in the XY plane (G17 G2, radius 1.5): expand the arc into straight G1 segments with UGS ArcExpander at 1 unit segment length.
22
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
23
  Origin: a small upstream fixture program for the sender preprocessor
24
- Reference result: processed_gcode
25
- Reference source: ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.parsed_output.nc (8 non-empty lines) and ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.stream_output.nc: the outputs recorded upstream for configuration ArcCoarse and checked line by line by FixturesTest.checkFixture (lines 193 to 210).
26
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 1) after CommentProcessor: arcs become G1 segments of at most 1 unit of planar arc length), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
27
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
28
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
@@ -30,8 +26,6 @@ Files:
30
  - `ugs-core/test/resources/gcode/fixtures/arc/G19_G3_one_arc.input.nc`: One counterclockwise half circle in the YZ plane: expand into G1 segments of at most 0.1 unit and split long moves.
31
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
32
  Origin: a small upstream fixture program for the sender preprocessor
33
- Reference result: processed_gcode
34
- Reference source: ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.parsed_output.nc (51 non-empty lines) and ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.stream_output.nc: the outputs recorded upstream for configuration Arc and checked line by line by FixturesTest.checkFixture (lines 193 to 210); the last line differs only in how the endpoint is rounded (upstream keeps the programmed endpoint `G1X0Y1.5Z0`)
35
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 0.1) after CommentProcessor, then LineSplitter(1): arcs become G1 segments of at most 0.1 units of planar arc length, and moves longer than 1 unit are split), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
36
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
37
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
@@ -39,8 +33,6 @@ Files:
39
  - `ugs-core/test/resources/gcode/fixtures/arc/Helix.input.nc`: Helical moves (G2 then G3 full circles with Z climbing 0 to 5 to 10, changing feed): expand into G1 segments with UGS ArcExpander at 1 unit segment length.
40
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
41
  Origin: a small upstream fixture program for the sender preprocessor
42
- Reference result: processed_gcode
43
- Reference source: ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.parsed_output.nc (22 non-empty lines) and ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.stream_output.nc: the outputs recorded upstream for configuration ArcCoarse and checked line by line by FixturesTest.checkFixture (lines 193 to 210).
44
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 1) after CommentProcessor: arcs become G1 segments of at most 1 unit of planar arc length), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
45
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
46
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
@@ -48,8 +40,6 @@ Files:
48
  - `ugs-core/test/resources/gcode/fixtures/arc/arc_feed.input.nc`: Helical full circles in G17 with feeds 200 and 300: expand into G1 segments of at most 0.1 unit with Z interpolated, and split long moves.
49
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
50
  Origin: a small upstream fixture program for the sender preprocessor
51
- Reference result: processed_gcode
52
- Reference source: ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.parsed_output.nc (195 non-empty lines) and ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.stream_output.nc: the outputs recorded upstream for configuration Arc and checked line by line by FixturesTest.checkFixture (lines 193 to 210).
53
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 0.1) after CommentProcessor, then LineSplitter(1): arcs become G1 segments of at most 0.1 units of planar arc length, and moves longer than 1 unit are split), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
54
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
55
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
@@ -57,8 +47,6 @@ Files:
57
  - `ugs-core/test/resources/gcode/fixtures/arc/circle_test.input.nc`: A CAM post output (2D contour with a 1.4 mm flat end mill, arcs on several depths): expand every arc into G1 segments with UGS ArcExpander at 1 unit segment length.
58
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
59
  Origin: a real CAM post-processor output used upstream as a parser fixture
60
- Reference result: processed_gcode
61
- Reference source: ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.parsed_output.nc (125 non-empty lines) and ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.stream_output.nc: the outputs recorded upstream for configuration ArcCoarse and checked line by line by FixturesTest.checkFixture (lines 193 to 210); comments are removed and each arc is replaced by G1 segments; the parsed output has 125 non-empty lines; the first expanded arc was checked by hand for segment count and endpoints
62
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 1) after CommentProcessor: arcs become G1 segments of at most 1 unit of planar arc length), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
63
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
64
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
@@ -66,8 +54,6 @@ Files:
66
  - `ugs-core/test/resources/gcode/fixtures/run-from/simple.g1.input.nc`: Resume a 57 line inch program (circular depth passes) from line 17 with UGS RunFromProcessor: produce the safe preamble that restores modal state and position, then the rest of the program.
67
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
68
  Origin: a small upstream fixture program for the sender preprocessor
69
- Reference result: processed_gcode
70
- Reference source: ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.parsed_output.nc (44 non-empty lines) and ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.stream_output.nc: the outputs recorded upstream for configuration RunFrom and checked line by line by FixturesTest.checkFixture (lines 193 to 210).
71
  Check: Task: given the input program and the processor configuration (RunFromProcessor(17): start the program from line 17 as UGS numbers lines, after a generated preamble that restores units, distance mode, plane, the last position, feed and spindle state), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
72
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
73
  Companions (not included): ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
@@ -75,8 +61,6 @@ Files:
75
  - `ugs-core/test/resources/gcode/fixtures/run-from/simple.g2.input.nc`: Resume a 43 line inch program (stacked G2 depth passes) from line 17 with UGS RunFromProcessor: produce the safe preamble that restores modal state and position, then the rest of the program.
76
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
77
  Origin: a small upstream fixture program for the sender preprocessor
78
- Reference result: processed_gcode
79
- Reference source: ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.parsed_output.nc (31 non-empty lines) and ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.stream_output.nc: the outputs recorded upstream for configuration RunFrom and checked line by line by FixturesTest.checkFixture (lines 193 to 210); the output then continues with `G1 Z-.5` (input line 19) and the rest of the program unchanged
80
  Check: Task: given the input program and the processor configuration (RunFromProcessor(17): start the program from line 17 as UGS numbers lines, after a generated preamble that restores units, distance mode, plane, the last position, feed and spindle state), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
81
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
82
  Companions (not included): ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
 
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 7 of 7 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
12
 
13
  **Licence (SPDX):** GPL-3.0-only
 
19
  - `ugs-core/test/resources/gcode/fixtures/arc/G17_G2_one_arc.input.nc`: One clockwise half circle in the XY plane (G17 G2, radius 1.5): expand the arc into straight G1 segments with UGS ArcExpander at 1 unit segment length.
20
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
21
  Origin: a small upstream fixture program for the sender preprocessor
 
 
22
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 1) after CommentProcessor: arcs become G1 segments of at most 1 unit of planar arc length), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
23
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
24
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/G17_G2_one_arc.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
 
26
  - `ugs-core/test/resources/gcode/fixtures/arc/G19_G3_one_arc.input.nc`: One counterclockwise half circle in the YZ plane: expand into G1 segments of at most 0.1 unit and split long moves.
27
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
28
  Origin: a small upstream fixture program for the sender preprocessor
 
 
29
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 0.1) after CommentProcessor, then LineSplitter(1): arcs become G1 segments of at most 0.1 units of planar arc length, and moves longer than 1 unit are split), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
30
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
31
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArc/G19_G3_one_arc.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
 
33
  - `ugs-core/test/resources/gcode/fixtures/arc/Helix.input.nc`: Helical moves (G2 then G3 full circles with Z climbing 0 to 5 to 10, changing feed): expand into G1 segments with UGS ArcExpander at 1 unit segment length.
34
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
35
  Origin: a small upstream fixture program for the sender preprocessor
 
 
36
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 1) after CommentProcessor: arcs become G1 segments of at most 1 unit of planar arc length), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
37
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
38
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/Helix.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
 
40
  - `ugs-core/test/resources/gcode/fixtures/arc/arc_feed.input.nc`: Helical full circles in G17 with feeds 200 and 300: expand into G1 segments of at most 0.1 unit with Z interpolated, and split long moves.
41
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
42
  Origin: a small upstream fixture program for the sender preprocessor
 
 
43
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 0.1) after CommentProcessor, then LineSplitter(1): arcs become G1 segments of at most 0.1 units of planar arc length, and moves longer than 1 unit are split), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
44
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
45
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArc/arc_feed.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
 
47
  - `ugs-core/test/resources/gcode/fixtures/arc/circle_test.input.nc`: A CAM post output (2D contour with a 1.4 mm flat end mill, arcs on several depths): expand every arc into G1 segments with UGS ArcExpander at 1 unit segment length.
48
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
49
  Origin: a real CAM post-processor output used upstream as a parser fixture
 
 
50
  Check: Task: given the input program and the processor configuration (ArcExpander(true, 1) after CommentProcessor: arcs become G1 segments of at most 1 unit of planar arc length), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
51
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
52
  Companions (not included): ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/arc/outArcCoarse/circle_test.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
 
54
  - `ugs-core/test/resources/gcode/fixtures/run-from/simple.g1.input.nc`: Resume a 57 line inch program (circular depth passes) from line 17 with UGS RunFromProcessor: produce the safe preamble that restores modal state and position, then the rest of the program.
55
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
56
  Origin: a small upstream fixture program for the sender preprocessor
 
 
57
  Check: Task: given the input program and the processor configuration (RunFromProcessor(17): start the program from line 17 as UGS numbers lines, after a generated preamble that restores units, distance mode, plane, the last position, feed and spindle state), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
58
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
59
  Companions (not included): ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g1.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
 
61
  - `ugs-core/test/resources/gcode/fixtures/run-from/simple.g2.input.nc`: Resume a 43 line inch program (stacked G2 depth passes) from line 17 with UGS RunFromProcessor: produce the safe preamble that restores modal state and position, then the rest of the program.
62
  Format: G-code text (.nc, Grbl dialect); reference outputs are processed G-code text (parsed_output.nc) and the UGS stream format (stream_output.nc, header `gsw_meta:N` then `original¶¶processed¶¶line¶¶` rows)
63
  Origin: a small upstream fixture program for the sender preprocessor
 
 
64
  Check: Task: given the input program and the processor configuration (RunFromProcessor(17): start the program from line 17 as UGS numbers lines, after a generated preamble that restores units, distance mode, plane, the last position, feed and spindle state), produce the processed G-code the UGS parser emits. Judge: compare the agent output with ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.parsed_output.nc line by line after trimming each line, and require the same number of lines, as FixturesTest.checkFixture does; the stream_output file is checked the same way when the stream format is asked for. outputs traced by hand.
65
  Tools: UGS core Java library used headless (no GUI): build a GcodeParser, add the processors of the test configuration and call GcodeParserUtils.processAndExport(parser, inputFile, new GcodeStreamWriter(out)) as FixturesTest.java line 156 does; run with Maven (`mvn -pl ugs-core test -Dtest=FixturesTest`); the program and both expected files are plain text
66
  Companions (not included): ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.parsed_output.nc; ugs-core/test/resources/gcode/fixtures/run-from/outRunFrom/simple.g2.stream_output.nc; ugs-core/test/com/willwinder/universalgcodesender/gcode/fixtures/FixturesTest.java; COPYING
campaign_files/2026-10-hw-eda-tests-2/uvvm/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 2 of 2 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: self_checking_testbench_pass_fail. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_irqc/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles uvvm_util, bitvis_vip_sbi and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_irqc irqc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts (UVVM prints `>> Simulation SUCCESS: No mismatch between counted and expected serious alerts`) and the log reaches `SIMULATION COMPLETED`. Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_uart/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles the UVVM libraries, the VIPs and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_uart uart_vvc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts and the log reaches `SIMULATION COMPLETED`.
14
 
15
  **Licence (SPDX):** Apache-2.0
@@ -21,16 +19,12 @@ Files:
21
  - `bitvis_irqc/tb/irqc_demo_tb.vhd`
22
  Format: VHDL-2008 DUT with a UVVM self-checking testbench and a GHDL shell script
23
  Origin: self-checking testbench written by the UVVM authors for their own demo IP (tool_fixture): expected register values and interrupt timing are written in the test
24
- Reference result: self_checking_testbench_pass_fail
25
- Reference source: report_alert_counters(FINAL) prints the pass or fail conclusion
26
  Check: Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_irqc/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles uvvm_util, bitvis_vip_sbi and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_irqc irqc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts (UVVM prints `>> Simulation SUCCESS: No mismatch between counted and expected serious alerts`) and the log reaches `SIMULATION COMPLETED`.
27
  Companions (not included): bitvis_irqc/LICENSE.TXT; bitvis_irqc/script/ghdl_compile_all_and_simulate.sh; bitvis_irqc/src/irqc.vhd; bitvis_irqc/src/irqc_core.vhd; bitvis_irqc/src/irqc_pif.vhd; bitvis_irqc/src/irqc_pif_pkg.vhd
28
  Dependencies: GHDL v6.0.0 (bitvis_irqc/script/ghdl_compile_all_and_simulate.sh calls ghdl with --std=08 -frelaxed-rules; version not pinned upstream; latest release on or before the unit commit date 2026-04-22, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; UVVM uvvm_util and bitvis_vip_sbi at the unit commit 90d56e93c542bf0d5e2ab9f791cc7395bd1aa896 (the script compiles ../../uvvm_util and ../../bitvis_vip_sbi; not taken, needs the UVVM source tree); open source, not packed
29
  - `bitvis_uart/tb/uart_vvc_demo_tb.vhd`: UVVM bitvis_uart: UART with a self-checking UVVM VVC testbench.
30
  Format: VHDL-2008 DUT with a UVVM self-checking testbench and a GHDL shell script
31
  Origin: self-checking testbench written by the UVVM authors for their own demo IP (tool_fixture): the expected bytes are written in the test
32
- Reference result: self_checking_testbench_pass_fail
33
- Reference source: the testbench sends bytes over the register bus and over the serial line through the SBI and UART verification components and checks each byte received with sbi_check or uart_expect (fixed bytes 0x55, 0xAA, 0xB4, 0x87, 0xA1 to 0xA3 and two runs of 102 bytes 0x80+i); await_uvvm_completion with REPORT_ALERT_COUNTERS_FINAL prints the conclusion
34
  Check: Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_uart/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles the UVVM libraries, the VIPs and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_uart uart_vvc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts and the log reaches `SIMULATION COMPLETED`.
35
  Companions (not included): bitvis_uart/LICENSE.TXT; bitvis_uart/script/ghdl_compile_all_and_simulate.sh; bitvis_uart/src/uart.vhd; bitvis_uart/src/uart_core.vhd; bitvis_uart/src/uart_pif.vhd; bitvis_uart/src/uart_pif_pkg.vhd; bitvis_uart/src/uart_pkg.vhd; bitvis_uart/tb/uart_vvc_demo_th.vhd
36
  Dependencies: GHDL v6.0.0 (bitvis_uart/script/ghdl_compile_all_and_simulate.sh calls ghdl with --std=08 -frelaxed-rules; version not pinned upstream; latest release on or before the unit commit date 2026-04-22, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; UVVM uvvm_util, uvvm_vvc_framework, bitvis_vip_scoreboard, bitvis_vip_sbi, bitvis_vip_uart and bitvis_vip_clock_generator at the unit commit 90d56e93c542bf0d5e2ab9f791cc7395bd1aa896 (compiled by the script; not taken, needs the UVVM source tree); open source, not packed
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 2 of 2 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_irqc/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles uvvm_util, bitvis_vip_sbi and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_irqc irqc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts (UVVM prints `>> Simulation SUCCESS: No mismatch between counted and expected serious alerts`) and the log reaches `SIMULATION COMPLETED`. Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_uart/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles the UVVM libraries, the VIPs and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_uart uart_vvc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts and the log reaches `SIMULATION COMPLETED`.
12
 
13
  **Licence (SPDX):** Apache-2.0
 
19
  - `bitvis_irqc/tb/irqc_demo_tb.vhd`
20
  Format: VHDL-2008 DUT with a UVVM self-checking testbench and a GHDL shell script
21
  Origin: self-checking testbench written by the UVVM authors for their own demo IP (tool_fixture): expected register values and interrupt timing are written in the test
 
 
22
  Check: Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_irqc/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles uvvm_util, bitvis_vip_sbi and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_irqc irqc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts (UVVM prints `>> Simulation SUCCESS: No mismatch between counted and expected serious alerts`) and the log reaches `SIMULATION COMPLETED`.
23
  Companions (not included): bitvis_irqc/LICENSE.TXT; bitvis_irqc/script/ghdl_compile_all_and_simulate.sh; bitvis_irqc/src/irqc.vhd; bitvis_irqc/src/irqc_core.vhd; bitvis_irqc/src/irqc_pif.vhd; bitvis_irqc/src/irqc_pif_pkg.vhd
24
  Dependencies: GHDL v6.0.0 (bitvis_irqc/script/ghdl_compile_all_and_simulate.sh calls ghdl with --std=08 -frelaxed-rules; version not pinned upstream; latest release on or before the unit commit date 2026-04-22, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; UVVM uvvm_util and bitvis_vip_sbi at the unit commit 90d56e93c542bf0d5e2ab9f791cc7395bd1aa896 (the script compiles ../../uvvm_util and ../../bitvis_vip_sbi; not taken, needs the UVVM source tree); open source, not packed
25
  - `bitvis_uart/tb/uart_vvc_demo_tb.vhd`: UVVM bitvis_uart: UART with a self-checking UVVM VVC testbench.
26
  Format: VHDL-2008 DUT with a UVVM self-checking testbench and a GHDL shell script
27
  Origin: self-checking testbench written by the UVVM authors for their own demo IP (tool_fixture): the expected bytes are written in the test
 
 
28
  Check: Build in a sandbox with the versions listed in dependencies and the UVVM tree at the unit commit: from bitvis_uart/script run `sh ghdl_compile_all_and_simulate.sh`, which compiles the UVVM libraries, the VIPs and the DUT and runs `ghdl -r --std=08 -frelaxed-rules --work=bitvis_uart uart_vvc_demo_tb`. The unit passes when the final alert summary reports no ERROR or FAILURE alerts and the log reaches `SIMULATION COMPLETED`.
29
  Companions (not included): bitvis_uart/LICENSE.TXT; bitvis_uart/script/ghdl_compile_all_and_simulate.sh; bitvis_uart/src/uart.vhd; bitvis_uart/src/uart_core.vhd; bitvis_uart/src/uart_pif.vhd; bitvis_uart/src/uart_pif_pkg.vhd; bitvis_uart/src/uart_pkg.vhd; bitvis_uart/tb/uart_vvc_demo_th.vhd
30
  Dependencies: GHDL v6.0.0 (bitvis_uart/script/ghdl_compile_all_and_simulate.sh calls ghdl with --std=08 -frelaxed-rules; version not pinned upstream; latest release on or before the unit commit date 2026-04-22, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; UVVM uvvm_util, uvvm_vvc_framework, bitvis_vip_scoreboard, bitvis_vip_sbi, bitvis_vip_uart and bitvis_vip_clock_generator at the unit commit 90d56e93c542bf0d5e2ab9f791cc7395bd1aa896 (compiled by the script; not taken, needs the UVVM source tree); open source, not packed
campaign_files/2026-10-hw-eda-tests-2/verilog-axi/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 7 of 7 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: self_checking_testbench_pass_fail. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Tools are named in the check written per unit in the notes.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,56 +19,42 @@ Units:
21
  - `tb/axi_fifo/` (2 files): Verilog-axi axi_fifo: parameterised AXI4 FIFO (separate read and write channel FIFOs) with its cocotb testbench.
22
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
23
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
24
- Reference result: self_checking_testbench_pass_fail
25
- Reference source: test_axi_fifo.py writes byte strings (bytes x % 256) of lengths 1 to 2*byte_lanes-1 plus 1024 at offsets near the start and the end of a 4 KiB page through the cocotbext-axi AxiMaster into an AxiRam model and asserts tb.axi_ram.read(addr, length) == test_data with 0xaa guard bytes on both sides (lines 81-110), reads back the same way (lines 116-141), and a stress test runs workers with random.randint lengths and addresses (lines 147-170); write offsets are byte_lanes..2*byte_lanes-1 and 4096-byte_lanes..4095 (lines 96-97); the pytest parametrisation (lines 212-214) varies DATA_WIDTH 8/16/32 and FIFO delay 0/1, the Makefile run uses DATA_WIDTH 32 and WRITE/READ_FIFO_DELAY 0 (Makefile lines 37 and 53-54)
26
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axi_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_fifo.v, rtl/axi_fifo_wr.v, rtl/axi_fifo_rd.v and keep the testbench unchanged.
27
  Companions (not included): rtl/axi_fifo.v; rtl/axi_fifo_wr.v; rtl/axi_fifo_rd.v; COPYING
28
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
29
  - `tb/axil_ram/` (2 files): Verilog-axi axil_ram: AXI4-Lite RAM with its cocotb testbench.
30
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
31
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
32
- Reference result: self_checking_testbench_pass_fail
33
- Reference source: test_axil_ram.py writes byte strings (bytes x % 256) of lengths 1 to 2*byte_lanes-1 at every byte offset within one word (range(byte_lanes), lines 85-86) through the cocotbext-axi AxiLiteMaster between 0xaa guard bytes and reads them back, asserting data.data == b'\xaa'+test_data+b'\xaa' (lines 75-98), a read test asserts read-back equality (lines 104-125), and a stress test runs workers with random.randint lengths and addresses (from line 131)
34
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axil_ram/ (the Makefile's own parameter set; the pytest matrix in test_axil_ram.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axil_ram.v and keep the testbench unchanged.
35
  Companions (not included): rtl/axil_ram.v; COPYING
36
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
37
  - `tb/axi_crossbar/` (2 files): Verilog-axi axi_crossbar: N x M AXI4 crossbar switch (address decode, arbitration, register slices) with its cocotb testbench.
38
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
39
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
40
- Reference result: self_checking_testbench_pass_fail
41
- Reference source: test_axi_crossbar.py connects AxiMaster models on the slave ports and AxiRam models on the master ports; the write test sends byte strings (bytes x % 256) from slave port s to RAM m and asserts tb.axi_ram[m].read(ram_addr, length) == test_data with 0xaa guard bytes (lines 95-125), the read test asserts read-back equality (lines 131-157), and a stress test runs workers with random target RAM, length and address (from line 163)
42
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_crossbar/ (the Makefile's own parameter set; the pytest matrix in test_axi_crossbar.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_crossbar.v, rtl/axi_crossbar_addr.v, rtl/axi_crossbar_rd.v, rtl/axi_crossbar_wr.v, rtl/axi_register_rd.v, rtl/axi_register_wr.v, rtl/arbiter.v, rtl/priority_encoder.v and keep the testbench unchanged. The Makefile generates the top-level wrapper axi_crossbar_wrap_<S>x<M>.v by running rtl/axi_crossbar_wrap.py (a Python generator from the same repo, packed as a companion).
43
  Companions (not included): rtl/axi_crossbar.v; rtl/axi_crossbar_addr.v; rtl/axi_crossbar_rd.v; rtl/axi_crossbar_wr.v; rtl/axi_register_rd.v; rtl/axi_register_wr.v; rtl/arbiter.v; rtl/priority_encoder.v; rtl/axi_crossbar_wrap.py; COPYING
44
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
45
  - `tb/axil_cdc/` (2 files): Verilog-axi axil_cdc: AXI4-Lite clock domain crossing bridge with its cocotb testbench (two fixed clocks, 8 ns and 10 ns).
46
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
47
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
48
- Reference result: self_checking_testbench_pass_fail
49
- Reference source: test_axil_cdc.py runs the cocotbext-axi AxiLiteMaster on s_clk against an AxiLiteRam on m_clk; the two clocks are fixed at 8 ns and 10 ns (lines 48-49), not random; the write test asserts tb.axil_ram.read(addr, length) == test_data with 0xaa guard bytes (lines 86-111), the read test asserts read-back equality (lines 117-138), and a stress test runs workers with random.randint lengths and addresses (from line 144)
50
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axil_cdc/ (the Makefile's own parameter set; the pytest matrix in test_axil_cdc.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axil_cdc.v, rtl/axil_cdc_wr.v, rtl/axil_cdc_rd.v and keep the testbench unchanged.
51
  Companions (not included): rtl/axil_cdc.v; rtl/axil_cdc_wr.v; rtl/axil_cdc_rd.v; COPYING
52
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
53
  - `tb/axi_ram/` (2 files): Verilog-axi axi_ram: AXI4 RAM with its cocotb testbench.
54
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
55
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
56
- Reference result: self_checking_testbench_pass_fail
57
- Reference source: test_axi_ram.py writes byte strings of varying length at varying offsets and burst sizes through the cocotbext-axi AxiMaster between guard bytes and reads them back, asserting equality (asserts at lines 102, 133 and 161: write test, read test, stress workers); the TestFactory at lines 180-194 crosses idle and backpressure inserters with the burst sizes
58
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_ram/ (the Makefile's own parameter set; the pytest matrix in test_axi_ram.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_ram.v and keep the testbench unchanged.
59
  Companions (not included): COPYING; rtl/axi_ram.v
60
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
61
  - `tb/axi_dma/` (2 files): Verilog-axi axi_dma: AXI4 DMA engine with AXI stream interfaces and its cocotb testbench.
62
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
63
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
64
- Reference result: self_checking_testbench_pass_fail
65
- Reference source: test_axi_dma.py issues read and write descriptors to the DMA, moves byte strings between an AxiRam model and AXI stream source and sink models, and asserts the received data and the RAM contents equal the sent data (asserts at lines 144-154 and 200-203); the TestFactory at lines 214-222 crosses the two tests with idle and backpressure inserters
66
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_dma/ (the Makefile's own parameter set; the pytest matrix in test_axi_dma.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_dma.v, rtl/axi_dma_wr.v, rtl/axi_dma_rd.v and keep the testbench unchanged.
67
  Companions (not included): COPYING; rtl/axi_dma.v; rtl/axi_dma_wr.v; rtl/axi_dma_rd.v
68
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
69
  - `tb/axi_adapter/` (2 files): Verilog-axi axi_adapter: AXI4 width adapter with its cocotb testbench.
70
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
71
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
72
- Reference result: self_checking_testbench_pass_fail
73
- Reference source: test_axi_adapter.py writes and reads byte strings through the adapter from a cocotbext-axi AxiMaster into an AxiRam model and asserts the read-back data and the RAM contents equal the written data (asserts at lines 108-110, 141 and 169); the TestFactory at lines 188-202 crosses idle and backpressure inserters with burst sizes
74
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_adapter/ (the Makefile's own parameter set; the pytest matrix in test_axi_adapter.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_adapter.v, rtl/axi_adapter_wr.v, rtl/axi_adapter_rd.v and keep the testbench unchanged.
75
  Companions (not included): rtl/axi_adapter.v; rtl/axi_adapter_wr.v; rtl/axi_adapter_rd.v; COPYING
76
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
 
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 7 of 7 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Tools are named in the check written per unit in the notes.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `tb/axi_fifo/` (2 files): Verilog-axi axi_fifo: parameterised AXI4 FIFO (separate read and write channel FIFOs) with its cocotb testbench.
20
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
21
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
22
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axi_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_fifo.v, rtl/axi_fifo_wr.v, rtl/axi_fifo_rd.v and keep the testbench unchanged.
23
  Companions (not included): rtl/axi_fifo.v; rtl/axi_fifo_wr.v; rtl/axi_fifo_rd.v; COPYING
24
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
25
  - `tb/axil_ram/` (2 files): Verilog-axi axil_ram: AXI4-Lite RAM with its cocotb testbench.
26
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
27
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
28
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axil_ram/ (the Makefile's own parameter set; the pytest matrix in test_axil_ram.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axil_ram.v and keep the testbench unchanged.
29
  Companions (not included): rtl/axil_ram.v; COPYING
30
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
31
  - `tb/axi_crossbar/` (2 files): Verilog-axi axi_crossbar: N x M AXI4 crossbar switch (address decode, arbitration, register slices) with its cocotb testbench.
32
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
33
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
34
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_crossbar/ (the Makefile's own parameter set; the pytest matrix in test_axi_crossbar.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_crossbar.v, rtl/axi_crossbar_addr.v, rtl/axi_crossbar_rd.v, rtl/axi_crossbar_wr.v, rtl/axi_register_rd.v, rtl/axi_register_wr.v, rtl/arbiter.v, rtl/priority_encoder.v and keep the testbench unchanged. The Makefile generates the top-level wrapper axi_crossbar_wrap_<S>x<M>.v by running rtl/axi_crossbar_wrap.py (a Python generator from the same repo, packed as a companion).
35
  Companions (not included): rtl/axi_crossbar.v; rtl/axi_crossbar_addr.v; rtl/axi_crossbar_rd.v; rtl/axi_crossbar_wr.v; rtl/axi_register_rd.v; rtl/axi_register_wr.v; rtl/arbiter.v; rtl/priority_encoder.v; rtl/axi_crossbar_wrap.py; COPYING
36
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
37
  - `tb/axil_cdc/` (2 files): Verilog-axi axil_cdc: AXI4-Lite clock domain crossing bridge with its cocotb testbench (two fixed clocks, 8 ns and 10 ns).
38
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
39
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
40
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axil_cdc/ (the Makefile's own parameter set; the pytest matrix in test_axil_cdc.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axil_cdc.v, rtl/axil_cdc_wr.v, rtl/axil_cdc_rd.v and keep the testbench unchanged.
41
  Companions (not included): rtl/axil_cdc.v; rtl/axil_cdc_wr.v; rtl/axil_cdc_rd.v; COPYING
42
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
43
  - `tb/axi_ram/` (2 files): Verilog-axi axi_ram: AXI4 RAM with its cocotb testbench.
44
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
45
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
46
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_ram/ (the Makefile's own parameter set; the pytest matrix in test_axi_ram.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_ram.v and keep the testbench unchanged.
47
  Companions (not included): COPYING; rtl/axi_ram.v
48
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
49
  - `tb/axi_dma/` (2 files): Verilog-axi axi_dma: AXI4 DMA engine with AXI stream interfaces and its cocotb testbench.
50
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
51
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
52
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_dma/ (the Makefile's own parameter set; the pytest matrix in test_axi_dma.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_dma.v, rtl/axi_dma_wr.v, rtl/axi_dma_rd.v and keep the testbench unchanged.
53
  Companions (not included): COPYING; rtl/axi_dma.v; rtl/axi_dma_wr.v; rtl/axi_dma_rd.v
54
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
55
  - `tb/axi_adapter/` (2 files): Verilog-axi axi_adapter: AXI4 width adapter with its cocotb testbench.
56
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
57
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
58
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axi_adapter/ (the Makefile's own parameter set; the pytest matrix in test_axi_adapter.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axi_adapter.v, rtl/axi_adapter_wr.v, rtl/axi_adapter_rd.v and keep the testbench unchanged.
59
  Companions (not included): rtl/axi_adapter.v; rtl/axi_adapter_wr.v; rtl/axi_adapter_rd.v; COPYING
60
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 516bd5da); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 516bd5da); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 516bd5da); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 516bd5da); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 516bd5da); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 516bd5da); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 516bd5da); open source, not packed; cocotbext-axi==0.1.24 (pinned upstream in tox.ini line 20 at commit 516bd5da); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 516bd5da); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
campaign_files/2026-10-hw-eda-tests-2/verilog-axis/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 3 of 3 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: self_checking_testbench_pass_fail. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_async_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_async_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_async_fifo.v and keep the testbench unchanged. Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_cobs_encode/ (the Makefile's own parameter set; the pytest matrix in test_axis_cobs_encode.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_cobs_encode.v, rtl/axis_fifo.v and keep the testbench unchanged. The Makefile reads rtl/axis_cobs_encode.f (lists axis_cobs_encode.v and axis_fifo.v); all three are packed. Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_fifo.v and keep the testbench unchanged.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,24 +19,18 @@ Units:
21
  - `tb/axis_async_fifo/` (2 files): Verilog-axis axis_async_fifo: AXI4-Stream asynchronous (dual-clock) FIFO with frame mode and drop logic, with its cocotb testbench.
22
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
23
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
24
- Reference result: self_checking_testbench_pass_fail
25
- Reference source: test_axis_async_fifo.py sends AxiStreamFrame objects from AxiStreamSource on s_clk and asserts rx_frame.tdata == test_frame.tdata, tid and tdest equal and tuser clear at the sink on m_clk (lines 106-139), plus single tests for tuser propagation, sink pause at start, reset of either domain while a frame is in flight, shift-in and shift-out resets, pause request, overflow and oversize frames (lines 145-546), and a stress test with random lengths (lines 549-566)
26
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_async_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_async_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_async_fifo.v and keep the testbench unchanged.
27
  Companions (not included): rtl/axis_async_fifo.v; COPYING
28
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 48ff7a7e); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 48ff7a7e); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 48ff7a7e); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 48ff7a7e); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 48ff7a7e); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 48ff7a7e); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 48ff7a7e); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 48ff7a7e); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 48ff7a7e); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
29
  - `tb/axis_cobs_encode/` (2 files): Verilog-axis axis_cobs_encode: AXI4-Stream Consistent Overhead Byte Stuffing (COBS) encoder with its cocotb testbench.
30
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
31
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
32
- Reference result: self_checking_testbench_pass_fail
33
- Reference source: test_axis_cobs_encode.py has a reference Python COBS encoder and decoder; with APPEND_ZERO 1 it asserts rx_frame.tdata == cobs_encode(test_data)+b'\x00' and cobs_decode(rx_frame.tdata[:-1]) == test_data, with APPEND_ZERO 0 (the Makefile value) rx_frame.tdata == cobs_encode(test_data) and cobs_decode(rx_frame.tdata) == test_data (lines 160-166); frame lengths 1-32, 253-258, 512 and 64 one-byte frames cover the 254-byte block boundary, payloads are all zero, nonzero incrementing, zero-framed and PRBS31
34
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_cobs_encode/ (the Makefile's own parameter set; the pytest matrix in test_axis_cobs_encode.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_cobs_encode.v, rtl/axis_fifo.v and keep the testbench unchanged. The Makefile reads rtl/axis_cobs_encode.f (lists axis_cobs_encode.v and axis_fifo.v); all three are packed.
35
  Companions (not included): rtl/axis_cobs_encode.f; rtl/axis_cobs_encode.v; rtl/axis_fifo.v; COPYING
36
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 48ff7a7e); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 48ff7a7e); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 48ff7a7e); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 48ff7a7e); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 48ff7a7e); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 48ff7a7e); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 48ff7a7e); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 48ff7a7e); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 48ff7a7e); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
37
  - `tb/axis_fifo/` (2 files): Verilog-axis axis_fifo: AXI stream FIFO with its cocotb testbench.
38
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
39
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
40
- Reference result: self_checking_testbench_pass_fail
41
- Reference source: test_axis_fifo.py sends AXI stream frames through the FIFO with cocotbext-axi source and sink models and asserts the received frames equal the sent frames, including tuser, overflow and frame-drop cases (single tests and stress tests); the TestFactory at lines 419-441 crosses run_test with payload lengths and idle and backpressure inserters
42
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_fifo.v and keep the testbench unchanged.
43
  Companions (not included): COPYING; rtl/axis_fifo.v
44
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 48ff7a7e); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 48ff7a7e); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 48ff7a7e); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 48ff7a7e); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 48ff7a7e); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 48ff7a7e); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 48ff7a7e); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 48ff7a7e); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 48ff7a7e); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
 
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 3 of 3 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_async_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_async_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_async_fifo.v and keep the testbench unchanged. Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_cobs_encode/ (the Makefile's own parameter set; the pytest matrix in test_axis_cobs_encode.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_cobs_encode.v, rtl/axis_fifo.v and keep the testbench unchanged. The Makefile reads rtl/axis_cobs_encode.f (lists axis_cobs_encode.v and axis_fifo.v); all three are packed. Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_fifo.v and keep the testbench unchanged.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `tb/axis_async_fifo/` (2 files): Verilog-axis axis_async_fifo: AXI4-Stream asynchronous (dual-clock) FIFO with frame mode and drop logic, with its cocotb testbench.
20
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
21
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
22
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_async_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_async_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_async_fifo.v and keep the testbench unchanged.
23
  Companions (not included): rtl/axis_async_fifo.v; COPYING
24
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 48ff7a7e); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 48ff7a7e); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 48ff7a7e); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 48ff7a7e); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 48ff7a7e); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 48ff7a7e); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 48ff7a7e); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 48ff7a7e); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 48ff7a7e); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
25
  - `tb/axis_cobs_encode/` (2 files): Verilog-axis axis_cobs_encode: AXI4-Stream Consistent Overhead Byte Stuffing (COBS) encoder with its cocotb testbench.
26
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
27
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
28
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_cobs_encode/ (the Makefile's own parameter set; the pytest matrix in test_axis_cobs_encode.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_cobs_encode.v, rtl/axis_fifo.v and keep the testbench unchanged. The Makefile reads rtl/axis_cobs_encode.f (lists axis_cobs_encode.v and axis_fifo.v); all three are packed.
29
  Companions (not included): rtl/axis_cobs_encode.f; rtl/axis_cobs_encode.v; rtl/axis_fifo.v; COPYING
30
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 48ff7a7e); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 48ff7a7e); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 48ff7a7e); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 48ff7a7e); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 48ff7a7e); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 48ff7a7e); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 48ff7a7e); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 48ff7a7e); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 48ff7a7e); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
31
  - `tb/axis_fifo/` (2 files): Verilog-axis axis_fifo: AXI stream FIFO with its cocotb testbench.
32
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
33
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
34
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/axis_fifo/ (the Makefile's own parameter set; the pytest matrix in test_axis_fifo.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/axis_fifo.v and keep the testbench unchanged.
35
  Companions (not included): COPYING; rtl/axis_fifo.v
36
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 48ff7a7e); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 48ff7a7e); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 48ff7a7e); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 48ff7a7e); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 48ff7a7e); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 48ff7a7e); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 48ff7a7e); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 48ff7a7e); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 21 at commit 48ff7a7e); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
campaign_files/2026-10-hw-eda-tests-2/verilog-ethernet/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 2 of 2 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: self_checking_testbench_pass_fail. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/eth_mac_1g/ (the Makefile's own parameter set; the pytest matrix in test_eth_mac_1g.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/eth_mac_1g.v, rtl/axis_gmii_rx.v, rtl/axis_gmii_tx.v, rtl/mac_ctrl_rx.v, rtl/mac_ctrl_tx.v, rtl/mac_pause_ctrl_rx.v, rtl/mac_pause_ctrl_tx.v, rtl/lfsr.v and keep the testbench unchanged. Build in a sandbox with the versions listed in dependencies and run make inside tb/arp_cache/ (the Makefile's own parameter set; the pytest matrix in test_arp_cache.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/arp_cache.v, rtl/lfsr.v and keep the testbench unchanged.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,16 +19,12 @@ Units:
21
  - `tb/eth_mac_1g/` (2 files): Verilog-ethernet eth_mac_1g: 1G Ethernet MAC (GMII side, FCS generation and check, PTP timestamps, PFC and pause frames) with its cocotb testbench.
22
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
23
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
24
- Reference result: self_checking_testbench_pass_fail
25
- Reference source: test_eth_mac_1g.py sends frames through cocotbext-eth GmiiSource and AxiStreamSource and asserts that each received or transmitted frame carries the sent packet (for example bytes(tx_pkt).find(bytes(test_pkt)) == 0, line 509) with the right tuser error flags, PTP timestamps against PtpClockSimTime (abs difference < 0.01 ns, lines 223 and 267), TX underrun and error marking, and link-level and priority pause frames (scapy builds the reference control frames; LFC count == 4 at line 511, PFC count in 3..9 at line 650)
26
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/eth_mac_1g/ (the Makefile's own parameter set; the pytest matrix in test_eth_mac_1g.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/eth_mac_1g.v, rtl/axis_gmii_rx.v, rtl/axis_gmii_tx.v, rtl/mac_ctrl_rx.v, rtl/mac_ctrl_tx.v, rtl/mac_pause_ctrl_rx.v, rtl/mac_pause_ctrl_tx.v, rtl/lfsr.v and keep the testbench unchanged.
27
  Companions (not included): rtl/eth_mac_1g.v; rtl/axis_gmii_rx.v; rtl/axis_gmii_tx.v; rtl/mac_ctrl_rx.v; rtl/mac_ctrl_tx.v; rtl/mac_pause_ctrl_rx.v; rtl/mac_pause_ctrl_tx.v; rtl/lfsr.v; COPYING
28
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 77320a94); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 77320a94); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 77320a94); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 77320a94); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 77320a94); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 77320a94); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 77320a94); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 77320a94); open source, not packed; cocotbext-eth==0.1.22 (pinned upstream in tox.ini line 21 at commit 77320a94); open source, not packed; scapy==2.5.0 (pinned upstream in tox.ini line 22 at commit 77320a94); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 23 at commit 77320a94); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
29
  - `tb/arp_cache/` (2 files): Verilog-ethernet arp_cache: hashed ARP cache (IP to MAC lookup, write, clear) with its cocotb testbench.
30
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
31
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
32
- Reference result: self_checking_testbench_pass_fail
33
- Reference source: test_arp_cache.py writes IP to MAC entries, queries them and asserts the exact MAC returned (for example resp.mac == 0x0000c0a80111, line 96) and the error flag for misses, including an entry overwritten after a hash collision (the comment 'overwrites 0xc0a80112', line 112); there is no cache-clear test (clear_cache is only held at 0, line 59)
34
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/arp_cache/ (the Makefile's own parameter set; the pytest matrix in test_arp_cache.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/arp_cache.v, rtl/lfsr.v and keep the testbench unchanged.
35
  Companions (not included): rtl/arp_cache.v; rtl/lfsr.v; COPYING
36
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 77320a94); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 77320a94); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 77320a94); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 77320a94); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 77320a94); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 77320a94); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 77320a94); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 77320a94); open source, not packed; cocotbext-eth==0.1.22 (pinned upstream in tox.ini line 21 at commit 77320a94); open source, not packed; scapy==2.5.0 (pinned upstream in tox.ini line 22 at commit 77320a94); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 23 at commit 77320a94); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
 
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 2 of 2 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/eth_mac_1g/ (the Makefile's own parameter set; the pytest matrix in test_eth_mac_1g.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/eth_mac_1g.v, rtl/axis_gmii_rx.v, rtl/axis_gmii_tx.v, rtl/mac_ctrl_rx.v, rtl/mac_ctrl_tx.v, rtl/mac_pause_ctrl_rx.v, rtl/mac_pause_ctrl_tx.v, rtl/lfsr.v and keep the testbench unchanged. Build in a sandbox with the versions listed in dependencies and run make inside tb/arp_cache/ (the Makefile's own parameter set; the pytest matrix in test_arp_cache.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/arp_cache.v, rtl/lfsr.v and keep the testbench unchanged.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `tb/eth_mac_1g/` (2 files): Verilog-ethernet eth_mac_1g: 1G Ethernet MAC (GMII side, FCS generation and check, PTP timestamps, PFC and pause frames) with its cocotb testbench.
20
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
21
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
22
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/eth_mac_1g/ (the Makefile's own parameter set; the pytest matrix in test_eth_mac_1g.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/eth_mac_1g.v, rtl/axis_gmii_rx.v, rtl/axis_gmii_tx.v, rtl/mac_ctrl_rx.v, rtl/mac_ctrl_tx.v, rtl/mac_pause_ctrl_rx.v, rtl/mac_pause_ctrl_tx.v, rtl/lfsr.v and keep the testbench unchanged.
23
  Companions (not included): rtl/eth_mac_1g.v; rtl/axis_gmii_rx.v; rtl/axis_gmii_tx.v; rtl/mac_ctrl_rx.v; rtl/mac_ctrl_tx.v; rtl/mac_pause_ctrl_rx.v; rtl/mac_pause_ctrl_tx.v; rtl/lfsr.v; COPYING
24
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 77320a94); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 77320a94); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 77320a94); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 77320a94); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 77320a94); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 77320a94); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 77320a94); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 77320a94); open source, not packed; cocotbext-eth==0.1.22 (pinned upstream in tox.ini line 21 at commit 77320a94); open source, not packed; scapy==2.5.0 (pinned upstream in tox.ini line 22 at commit 77320a94); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 23 at commit 77320a94); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
25
  - `tb/arp_cache/` (2 files): Verilog-ethernet arp_cache: hashed ARP cache (IP to MAC lookup, write, clear) with its cocotb testbench.
26
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
27
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
28
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/arp_cache/ (the Makefile's own parameter set; the pytest matrix in test_arp_cache.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/arp_cache.v, rtl/lfsr.v and keep the testbench unchanged.
29
  Companions (not included): rtl/arp_cache.v; rtl/lfsr.v; COPYING
30
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 77320a94); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 77320a94); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 77320a94); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit 77320a94); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 17 at commit 77320a94); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 18 at commit 77320a94); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 19 at commit 77320a94); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 20 at commit 77320a94); open source, not packed; cocotbext-eth==0.1.22 (pinned upstream in tox.ini line 21 at commit 77320a94); open source, not packed; scapy==2.5.0 (pinned upstream in tox.ini line 22 at commit 77320a94); open source, not packed; jinja2==3.1.2 (pinned upstream in tox.ini line 23 at commit 77320a94); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
campaign_files/2026-10-hw-eda-tests-2/verilog-lfsr/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: self_checking_testbench_pass_fail. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/lfsr_crc/ (the Makefile's own parameter set; the pytest matrix in test_lfsr_crc.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/lfsr_crc.v, rtl/lfsr.v and keep the testbench unchanged.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,8 +19,6 @@ Units:
21
  - `tb/lfsr_crc/` (2 files): Verilog-lfsr lfsr_crc: parameterised LFSR-based CRC computation module (CRC-32 configuration) with its cocotb testbench.
22
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
23
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
24
- Reference result: self_checking_testbench_pass_fail
25
- Reference source: test_lfsr_crc.py feeds one block and then 1024 bytes into the CRC module and asserts val == ref (lines 105 and 123), where ref is Python zlib.crc32 for polynomial 0x4c11db7 or the testbench's own CRC-32C function for 0x1edc6f41 (lines 68-80, selected at lines 131-139); the Makefile run uses 0x4c11db7, so the reference there is zlib.crc32 (independent implementation)
26
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/lfsr_crc/ (the Makefile's own parameter set; the pytest matrix in test_lfsr_crc.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/lfsr_crc.v, rtl/lfsr.v and keep the testbench unchanged.
27
  Companions (not included): rtl/lfsr_crc.v; rtl/lfsr.v; COPYING
28
  Dependencies: Python 3.9 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.9: py3` at commit c1f86d03); open source, not packed; pytest==7.1.3 (pinned upstream in tox.ini line 14 at commit c1f86d03); open source, not packed; pytest-xdist==2.5.0 (pinned upstream in tox.ini line 15 at commit c1f86d03); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit c1f86d03); open source, not packed; cocotb==1.7.0 (pinned upstream in tox.ini line 17 at commit c1f86d03); open source, not packed; cocotb-test==0.2.2 (pinned upstream in tox.ini line 18 at commit c1f86d03); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
 
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/lfsr_crc/ (the Makefile's own parameter set; the pytest matrix in test_lfsr_crc.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/lfsr_crc.v, rtl/lfsr.v and keep the testbench unchanged.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `tb/lfsr_crc/` (2 files): Verilog-lfsr lfsr_crc: parameterised LFSR-based CRC computation module (CRC-32 configuration) with its cocotb testbench.
20
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
21
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
22
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/lfsr_crc/ (the Makefile's own parameter set; the pytest matrix in test_lfsr_crc.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/lfsr_crc.v, rtl/lfsr.v and keep the testbench unchanged.
23
  Companions (not included): rtl/lfsr_crc.v; rtl/lfsr.v; COPYING
24
  Dependencies: Python 3.9 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.9: py3` at commit c1f86d03); open source, not packed; pytest==7.1.3 (pinned upstream in tox.ini line 14 at commit c1f86d03); open source, not packed; pytest-xdist==2.5.0 (pinned upstream in tox.ini line 15 at commit c1f86d03); open source, not packed; pytest-split==0.8.0 (pinned upstream in tox.ini line 16 at commit c1f86d03); open source, not packed; cocotb==1.7.0 (pinned upstream in tox.ini line 17 at commit c1f86d03); open source, not packed; cocotb-test==0.2.2 (pinned upstream in tox.ini line 18 at commit c1f86d03); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
campaign_files/2026-10-hw-eda-tests-2/verilog-uart/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: self_checking_testbench_pass_fail. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/uart_rx/ (the Makefile's own parameter set; the pytest matrix in test_uart_rx.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/uart_rx.v and keep the testbench unchanged. Note: the testbench does not compare received byte values with the sent ones (test_uart_rx.py:68-85), so a receiver that outputs the right number of wrong bytes would pass.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,8 +19,6 @@ Units:
21
  - `tb/uart_rx/` (2 files): Verilog-uart uart_rx: UART receiver with AXI4-Stream output and its cocotb testbench.
22
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
23
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
24
- Reference result: self_checking_testbench_pass_fail
25
- Reference source: test_uart_rx.py sends incrementing and PRBS31 payloads through cocotbext-uart UartSource (prescale int(1/8e-9/baud/8), line 53), reads from the AxiStreamSink until as many bytes have arrived as were sent, and asserts only tb.sink.empty() (line 85); the received bytes are logged (line 83) but not compared with the sent bytes, so the testbench checks byte count and framing, not byte values
26
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/uart_rx/ (the Makefile's own parameter set; the pytest matrix in test_uart_rx.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/uart_rx.v and keep the testbench unchanged. Note: the testbench does not compare received byte values with the sent ones (test_uart_rx.py:68-85), so a receiver that outputs the right number of wrong bytes would pass.
27
  Companions (not included): rtl/uart_rx.v; COPYING
28
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 1b867e53); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 1b867e53); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 1b867e53); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 16 at commit 1b867e53); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 17 at commit 1b867e53); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 18 at commit 1b867e53); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 19 at commit 1b867e53); open source, not packed; cocotbext-uart==0.1.2 (pinned upstream in tox.ini line 20 at commit 1b867e53); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
 
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/uart_rx/ (the Makefile's own parameter set; the pytest matrix in test_uart_rx.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/uart_rx.v and keep the testbench unchanged. Note: the testbench does not compare received byte values with the sent ones (test_uart_rx.py:68-85), so a receiver that outputs the right number of wrong bytes would pass.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `tb/uart_rx/` (2 files): Verilog-uart uart_rx: UART receiver with AXI4-Stream output and its cocotb testbench.
20
  Format: cocotb testbench directory: Makefile (cocotb makefile flow, TOPLEVEL, VERILOG_SOURCES, PARAM_* overrides) and a pytest-runnable cocotb Python test module; DUT in synthesizable Verilog-2001
21
  Origin: self-checking cocotb testbench written by the IP author for his own RTL module (tool_fixture); upstream repo carries a deprecation notice pointing to fpganinja/taxi, the pinned commit is the final state
 
 
22
  Check: Build in a sandbox with the versions listed in dependencies and run make inside tb/uart_rx/ (the Makefile's own parameter set; the pytest matrix in test_uart_rx.py is not part of the judge); the unit passes when every cocotb test reports PASS and the results.xml has no failure. For a write-the-RTL task, hide or replace the DUT file(s) rtl/uart_rx.v and keep the testbench unchanged. Note: the testbench does not compare received byte values with the sent ones (test_uart_rx.py:68-85), so a receiver that outputs the right number of wrong bytes would pass.
23
  Companions (not included): rtl/uart_rx.v; COPYING
24
  Dependencies: Python 3.10 (pinned upstream in tox.ini lines 9-10 `[gh-actions] python = 3.10: py3` at commit 1b867e53); open source, not packed; pytest==7.2.1 (pinned upstream in tox.ini line 14 at commit 1b867e53); open source, not packed; pytest-xdist==3.1.0 (pinned upstream in tox.ini line 15 at commit 1b867e53); open source, not packed; cocotb==1.7.2 (pinned upstream in tox.ini line 16 at commit 1b867e53); open source, not packed; cocotb-bus==0.2.1 (pinned upstream in tox.ini line 17 at commit 1b867e53); open source, not packed; cocotb-test==0.2.4 (pinned upstream in tox.ini line 18 at commit 1b867e53); open source, not packed; cocotbext-axi==0.1.20 (pinned upstream in tox.ini line 19 at commit 1b867e53); open source, not packed; cocotbext-uart==0.1.2 (pinned upstream in tox.ini line 20 at commit 1b867e53); open source, not packed; Icarus Verilog v12_0 (iverilog and vvp; upstream does not pin it; latest release on or before the unit commit date 2025-02-27, GitHub release published 2023-06-11, the next one v13_0 came 2026-03-02; https://github.com/steveicarus/iverilog/releases/tag/v12_0); open source, not packed
campaign_files/2026-10-hw-eda-tests-2/vermaete-ipxact2systemverilog-example-directory/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 2 of 2 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: register_map_header_values. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Task: from test.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality. Task: from test2.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality.
14
 
15
  **Licence (SPDX):** GPL-2.0-only
@@ -21,16 +19,12 @@ Files:
21
  - `example/input/test.xml`: IP-XACT register map test.xml: generate the C header, SystemVerilog package and VHDL package with matching offsets and field bits.
22
  Format: IP-XACT 1.5 (SPIRIT) XML; C header; SystemVerilog package; VHDL package
23
  Origin: example register map shipped with the converter and its committed outputs
24
- Reference result: register_map_header_values
25
- Reference source: example/output/example.h lines 2-3 `Automatically generated with the command 'ipxact2c --srcFile example/input/test.xml --destDir example/output'`; e.g. example.h line 29 `#define EXAMPLE_REG4_OFFSET 0x04` against test.xml line 221
26
  Check: Task: from test.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality.
27
  Companions (not included): LICENSE; README.md; example/output/example.h; example/output/example_sv_pkg.sv; example/output/example_vhd_pkg.vhd
28
  Dependencies: ipxact2systemverilog 1.0.28 (pyproject.toml at the pinned commit, GPL-2.0-only, not packed): only needed to regenerate; Python 3.9 or later standard library for the comparison judge
29
  - `example/input/test2.xml`: IP-XACT register map test2.xml: generate the C header, SystemVerilog package and VHDL package with matching offsets and field bits.
30
  Format: IP-XACT 1.5 (SPIRIT) XML; C header; SystemVerilog package; VHDL package
31
  Origin: example register map shipped with the converter and its committed outputs
32
- Reference result: register_map_header_values
33
- Reference source: example/output/example2.h lines 2-3 `Automatically generated with the command 'ipxact2c --srcFile example/input/test2.xml --destDir example/output'`; e.g. example2.h line 27 `#define EXAMPLE2_SAMENAME_OFFSET 0x1D` against test2.xml line 56
34
  Check: Task: from test2.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality.
35
  Companions (not included): LICENSE; README.md; example/output/example2.h; example/output/example2_sv_pkg.sv; example/output/example2_vhd_pkg.vhd
36
  Dependencies: ipxact2systemverilog 1.0.28 (pyproject.toml at the pinned commit, GPL-2.0-only, not packed): only needed to regenerate; Python 3.9 or later standard library for the comparison judge
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 2 of 2 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Task: from test.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality. Task: from test2.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality.
12
 
13
  **Licence (SPDX):** GPL-2.0-only
 
19
  - `example/input/test.xml`: IP-XACT register map test.xml: generate the C header, SystemVerilog package and VHDL package with matching offsets and field bits.
20
  Format: IP-XACT 1.5 (SPIRIT) XML; C header; SystemVerilog package; VHDL package
21
  Origin: example register map shipped with the converter and its committed outputs
 
 
22
  Check: Task: from test.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality.
23
  Companions (not included): LICENSE; README.md; example/output/example.h; example/output/example_sv_pkg.sv; example/output/example_vhd_pkg.vhd
24
  Dependencies: ipxact2systemverilog 1.0.28 (pyproject.toml at the pinned commit, GPL-2.0-only, not packed): only needed to regenerate; Python 3.9 or later standard library for the comparison judge
25
  - `example/input/test2.xml`: IP-XACT register map test2.xml: generate the C header, SystemVerilog package and VHDL package with matching offsets and field bits.
26
  Format: IP-XACT 1.5 (SPIRIT) XML; C header; SystemVerilog package; VHDL package
27
  Origin: example register map shipped with the converter and its committed outputs
 
 
28
  Check: Task: from test2.xml produce the C header and the SystemVerilog and VHDL packages. Judge: compare every register offset and field bit position and width in the outputs against the IP-XACT input by exact integer equality.
29
  Companions (not included): LICENSE; README.md; example/output/example2.h; example/output/example2_sv_pkg.sv; example/output/example2_vhd_pkg.vhd
30
  Dependencies: ipxact2systemverilog 1.0.28 (pyproject.toml at the pinned commit, GPL-2.0-only, not packed): only needed to regenerate; Python 3.9 or later standard library for the comparison judge
campaign_files/2026-10-hw-eda-tests-2/vunit/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 3 of 3 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: self_checking_testbench_pass_fail. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/uart run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/uart_tx.vhd and src/uart_rx.vhd and keep the testbenches unchanged. Build in a sandbox with the versions listed in dependencies: in examples/vhdl/axi_dma run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/axi_dma.vhd (and src/axi_burst_gen.vhd or src/axi_dma_regs.vhd if wanted) and keep the testbench unchanged. Build in a sandbox with the versions listed in dependencies: in examples/vhdl/array run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/sobel_x.vhd and keep the testbench and the CSV files unchanged.
14
 
15
  **Licence (SPDX):** MPL-2.0
@@ -21,24 +19,18 @@ Units:
21
  - `examples/vhdl/uart/` (5 files): VUnit vhdl/uart: UART transmitter and receiver in VHDL with VUnit testbenches using UART and AXI stream verification components.
22
  Format: VHDL-2008 DUT with a VUnit testbench and a VUnit run.py
23
  Origin: self-checking VUnit testbench written by the VUnit maintainers for an example DUT (tool_fixture)
24
- Reference result: self_checking_testbench_pass_fail
25
- Reference source: the VUnit testbenches drive the UART through verification components (UART master and slave, AXI stream) and check every byte against the expected value with check_stream and check_equal
26
  Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/uart run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/uart_tx.vhd and src/uart_rx.vhd and keep the testbenches unchanged.
27
  Companions (not included): LICENSE.rst
28
  Dependencies: vunit_hdl at the unit commit 3600e096f40bfa530aa8cd08389815e81206ee88 with its OSVVM submodule (run.py calls VU.add_vhdl_builtins() and VU.add_osvvm(); OSVVM is Apache-2.0 per LICENSE.rst); open source, not packed; GHDL v6.0.0 (VUnit CI uses ghdl/setup-ghdl@main without a version, .github/workflows/push.yml lines 157-160; latest release on or before the unit commit date 2026-10-03, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; Python 3.12
29
  - `examples/vhdl/axi_dma/` (11 files): VUnit vhdl/axi_dma: AXI DMA engine in VHDL with a VUnit testbench using AXI slave memory models and burst statistics.
30
  Format: VHDL-2008 DUT with a VUnit testbench and a VUnit run.py
31
  Origin: self-checking VUnit testbench written by the VUnit maintainers for an example DUT (tool_fixture)
32
- Reference result: self_checking_testbench_pass_fail
33
- Reference source: the testbench fills a source buffer with random bytes, sets the same bytes as expected values on the destination buffer of the AXI memory model, runs the DMA and checks every byte was written as expected; burst split rules (4 KiB boundary, max burst length) are checked against fixed counts
34
  Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/axi_dma run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/axi_dma.vhd (and src/axi_burst_gen.vhd or src/axi_dma_regs.vhd if wanted) and keep the testbench unchanged.
35
  Companions (not included): LICENSE.rst
36
  Dependencies: vunit_hdl at the unit commit 3600e096f40bfa530aa8cd08389815e81206ee88 with its OSVVM submodule (run.py calls VU.add_vhdl_builtins() and VU.add_osvvm(); OSVVM is Apache-2.0 per LICENSE.rst); open source, not packed; GHDL v6.0.0 (VUnit CI uses ghdl/setup-ghdl@main without a version, .github/workflows/push.yml lines 157-160; latest release on or before the unit commit date 2026-10-03, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; Python 3.12
37
  - `examples/vhdl/array/` (5 files): VUnit vhdl/array: Sobel x filter in VHDL with a VUnit testbench checking against a model and a reference output CSV.
38
  Format: VHDL-2008 DUT with a VUnit testbench and a VUnit run.py
39
  Origin: self-checking VUnit testbench written by the VUnit maintainers for an example DUT (tool_fixture)
40
- Reference result: self_checking_testbench_pass_fail
41
- Reference source: test_random_data_against_model compares every output pixel with a VHDL model of the Sobel x kernel on random images; test_input_file_against_output_file compares with the fixed reference output.csv
42
  Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/array run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/sobel_x.vhd and keep the testbench and the CSV files unchanged.
43
  Companions (not included): LICENSE.rst
44
  Dependencies: vunit_hdl at the unit commit 3600e096f40bfa530aa8cd08389815e81206ee88 with its OSVVM submodule (run.py calls VU.add_vhdl_builtins() and VU.add_osvvm(); OSVVM is Apache-2.0 per LICENSE.rst); open source, not packed; GHDL v6.0.0 (VUnit CI uses ghdl/setup-ghdl@main without a version, .github/workflows/push.yml lines 157-160; latest release on or before the unit commit date 2026-10-03, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; Python 3.12
 
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 3 of 3 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/uart run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/uart_tx.vhd and src/uart_rx.vhd and keep the testbenches unchanged. Build in a sandbox with the versions listed in dependencies: in examples/vhdl/axi_dma run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/axi_dma.vhd (and src/axi_burst_gen.vhd or src/axi_dma_regs.vhd if wanted) and keep the testbench unchanged. Build in a sandbox with the versions listed in dependencies: in examples/vhdl/array run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/sobel_x.vhd and keep the testbench and the CSV files unchanged.
12
 
13
  **Licence (SPDX):** MPL-2.0
 
19
  - `examples/vhdl/uart/` (5 files): VUnit vhdl/uart: UART transmitter and receiver in VHDL with VUnit testbenches using UART and AXI stream verification components.
20
  Format: VHDL-2008 DUT with a VUnit testbench and a VUnit run.py
21
  Origin: self-checking VUnit testbench written by the VUnit maintainers for an example DUT (tool_fixture)
 
 
22
  Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/uart run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/uart_tx.vhd and src/uart_rx.vhd and keep the testbenches unchanged.
23
  Companions (not included): LICENSE.rst
24
  Dependencies: vunit_hdl at the unit commit 3600e096f40bfa530aa8cd08389815e81206ee88 with its OSVVM submodule (run.py calls VU.add_vhdl_builtins() and VU.add_osvvm(); OSVVM is Apache-2.0 per LICENSE.rst); open source, not packed; GHDL v6.0.0 (VUnit CI uses ghdl/setup-ghdl@main without a version, .github/workflows/push.yml lines 157-160; latest release on or before the unit commit date 2026-10-03, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; Python 3.12
25
  - `examples/vhdl/axi_dma/` (11 files): VUnit vhdl/axi_dma: AXI DMA engine in VHDL with a VUnit testbench using AXI slave memory models and burst statistics.
26
  Format: VHDL-2008 DUT with a VUnit testbench and a VUnit run.py
27
  Origin: self-checking VUnit testbench written by the VUnit maintainers for an example DUT (tool_fixture)
 
 
28
  Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/axi_dma run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/axi_dma.vhd (and src/axi_burst_gen.vhd or src/axi_dma_regs.vhd if wanted) and keep the testbench unchanged.
29
  Companions (not included): LICENSE.rst
30
  Dependencies: vunit_hdl at the unit commit 3600e096f40bfa530aa8cd08389815e81206ee88 with its OSVVM submodule (run.py calls VU.add_vhdl_builtins() and VU.add_osvvm(); OSVVM is Apache-2.0 per LICENSE.rst); open source, not packed; GHDL v6.0.0 (VUnit CI uses ghdl/setup-ghdl@main without a version, .github/workflows/push.yml lines 157-160; latest release on or before the unit commit date 2026-10-03, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; Python 3.12
31
  - `examples/vhdl/array/` (5 files): VUnit vhdl/array: Sobel x filter in VHDL with a VUnit testbench checking against a model and a reference output CSV.
32
  Format: VHDL-2008 DUT with a VUnit testbench and a VUnit run.py
33
  Origin: self-checking VUnit testbench written by the VUnit maintainers for an example DUT (tool_fixture)
 
 
34
  Check: Build in a sandbox with the versions listed in dependencies: in examples/vhdl/array run `python run.py --clean` with VUNIT_VHDL_STANDARD=2008 and VUNIT_SIMULATOR=ghdl, as the VUnit acceptance test does. The unit passes when run.py exits 0 (all VUnit tests pass). For a write-the-RTL task, hide or replace src/sobel_x.vhd and keep the testbench and the CSV files unchanged.
35
  Companions (not included): LICENSE.rst
36
  Dependencies: vunit_hdl at the unit commit 3600e096f40bfa530aa8cd08389815e81206ee88 with its OSVVM submodule (run.py calls VU.add_vhdl_builtins() and VU.add_osvvm(); OSVVM is Apache-2.0 per LICENSE.rst); open source, not packed; GHDL v6.0.0 (VUnit CI uses ghdl/setup-ghdl@main without a version, .github/workflows/push.yml lines 157-160; latest release on or before the unit commit date 2026-10-03, release v6.0.0 dated 2026-03-07 on https://github.com/ghdl/ghdl/releases/tag/v6.0.0); open source, not packed; Python 3.12
campaign_files/2026-10-m-issue-trackers-2/unilink-passenger-gtfs/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the service calendar itself (active services on a date, removed or added dates, day counts). Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
14
 
15
  **Licence (SPDX):** OGL-UK-3.0
@@ -22,8 +20,6 @@ Units:
22
  Read out of the archive: https://s3-eu-west-1.amazonaws.com/passenger-sources/unilink/gtfs/unilink_1789121305.zip
23
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt)
24
  Origin: Live static GTFS feed published by Unilink (published through Passenger, data.discoverpassenger.com) at https://s3-eu-west-1.amazonaws.com/passenger-sources/unilink/gtfs/unilink_1789121305.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
25
- Reference result: facts computed from the service calendar itself (active services on a date, removed or added dates, day counts)
26
- Reference source: calendar.txt weekday flags with start_date and end_date, calendar_dates.txt exception rows, agency_timezone of agency.txt, all from the pinned members.
27
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
28
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
29
  Companions (not included): calendar_dates.txt; agency.txt
 
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
12
 
13
  **Licence (SPDX):** OGL-UK-3.0
 
20
  Read out of the archive: https://s3-eu-west-1.amazonaws.com/passenger-sources/unilink/gtfs/unilink_1789121305.zip
21
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt)
22
  Origin: Live static GTFS feed published by Unilink (published through Passenger, data.discoverpassenger.com) at https://s3-eu-west-1.amazonaws.com/passenger-sources/unilink/gtfs/unilink_1789121305.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
 
 
23
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
24
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
25
  Companions (not included): calendar_dates.txt; agency.txt
campaign_files/2026-10-m-issue-trackers-2/victoria-clipper-gtfs/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the service calendar itself (active services on a date, removed or added dates, day counts). Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -22,8 +20,6 @@ Units:
22
  Read out of the archive: https://addtransit.com/gtfsfile/42017/VictoriaClipper.zip
23
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt, feed_info.txt)
24
  Origin: Live static GTFS feed published by Victoria Clipper, feed produced and hosted by AddTransit at https://addtransit.com/gtfsfile/42017/VictoriaClipper.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
25
- Reference result: facts computed from the service calendar itself (active services on a date, removed or added dates, day counts)
26
- Reference source: calendar.txt weekday flags with start_date and end_date, calendar_dates.txt exception rows, agency_timezone of agency.txt, all from the pinned members.
27
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
28
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
29
  Companions (not included): calendar_dates.txt; agency.txt; feed_info.txt
 
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
12
 
13
  **Licence (SPDX):** NOASSERTION
 
20
  Read out of the archive: https://addtransit.com/gtfsfile/42017/VictoriaClipper.zip
21
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt, feed_info.txt)
22
  Origin: Live static GTFS feed published by Victoria Clipper, feed produced and hosted by AddTransit at https://addtransit.com/gtfsfile/42017/VictoriaClipper.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
 
 
23
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
24
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
25
  Companions (not included): calendar_dates.txt; agency.txt; feed_info.txt
campaign_files/2026-10-m-issue-trackers-2/vinyl-misc/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the thread headers. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `m329-03/2020-March.txt`: Varnish not respecting pass to backend for specified http hosts (5 messages, 2020-03-31 to 2020-03-31).
22
  Format: Pipermail month archive, mbox text (plain, not compressed): full headers (From with obfuscated address and name, Date, Subject, Message-ID, In-Reply-To, References) and plain-text bodies. Cut the thread by root Message-ID through the In-Reply-To closure (last References entry when In-Reply-To is missing).
23
  Origin: Real list traffic in a closed past month of the list's archive. The unit is one thread, cut out of the month file by its root Message-ID; the month file is the pinned pointer and is delivered whole. The archive was requested at the old varnish-cache.org address and redirected to vinyl-cache.org (the project renamed itself); The month file is plain text, not gzip.
24
- Reference result: facts computed from the thread headers
25
- Reference source: Message-ID, In-Reply-To, References, From and Date headers of the messages in the thread, read from the pinned month file.
26
  Check: Exact match against values computed from the pinned month file: parse it as mbox, drop repeated copies of a Message-ID already seen (first copy kept), take the root Message-ID, collect every message whose In-Reply-To (or last References entry when In-Reply-To is missing) chains back to it, then count, take dates in UTC (a Date header without zone is read as UTC) and compare sender names. Message-IDs are compared after removing angle brackets and double quotes. People are counted by person: From entries with the same address, or the same name after removing accents and letter case, are one person (the names in scale are the first spelling of each). The first-reply interval is the exact difference in whole seconds between the Date headers of the root and the earliest reply, converted to UTC; Date headers have one-second resolution, so no rounding is involved. Deterministic. The program was checked on 2 threads of this source by a second, independent parser (splits the raw text on mbox From lines and reads headers with regular expressions, no mailbox or email module): the chosen thread (also read message by message) and 'Grace and misbehaving servers' (8 messages, 2 people, root <AM7PR08MB541673FF7DD0C05947E65F17ACF80@AM7PR08MB5416.eurprd08.prod.outlook.com>); message count, Message-ID set, last message time and first-reply seconds agreed on both.
27
  Tools: Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
 
8
 
9
  **Self-contained:** The file is self-contained.
10
 
 
 
11
  **Tools to open and check:** Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `m329-03/2020-March.txt`: Varnish not respecting pass to backend for specified http hosts (5 messages, 2020-03-31 to 2020-03-31).
20
  Format: Pipermail month archive, mbox text (plain, not compressed): full headers (From with obfuscated address and name, Date, Subject, Message-ID, In-Reply-To, References) and plain-text bodies. Cut the thread by root Message-ID through the In-Reply-To closure (last References entry when In-Reply-To is missing).
21
  Origin: Real list traffic in a closed past month of the list's archive. The unit is one thread, cut out of the month file by its root Message-ID; the month file is the pinned pointer and is delivered whole. The archive was requested at the old varnish-cache.org address and redirected to vinyl-cache.org (the project renamed itself); The month file is plain text, not gzip.
 
 
22
  Check: Exact match against values computed from the pinned month file: parse it as mbox, drop repeated copies of a Message-ID already seen (first copy kept), take the root Message-ID, collect every message whose In-Reply-To (or last References entry when In-Reply-To is missing) chains back to it, then count, take dates in UTC (a Date header without zone is read as UTC) and compare sender names. Message-IDs are compared after removing angle brackets and double quotes. People are counted by person: From entries with the same address, or the same name after removing accents and letter case, are one person (the names in scale are the first spelling of each). The first-reply interval is the exact difference in whole seconds between the Date headers of the root and the earliest reply, converted to UTC; Date headers have one-second resolution, so no rounding is involved. Deterministic. The program was checked on 2 threads of this source by a second, independent parser (splits the raw text on mbox From lines and reads headers with regular expressions, no mailbox or email module): the chosen thread (also read message by message) and 'Grace and misbehaving servers' (8 messages, 2 people, root <AM7PR08MB541673FF7DD0C05947E65F17ACF80@AM7PR08MB5416.eurprd08.prod.outlook.com>); message count, Message-ID set, last message time and first-reply seconds agreed on both.
23
  Tools: Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
campaign_files/2026-10-m-issue-trackers-2/vscode-github-issues/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the issue's own timeline (status changes, assignments, comments and their authors). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** curl -H 'Accept: application/vnd.github+json' https://api.github.com/repos/microsoft/vscode/issues/118256/events?per_page=100 and .../issues/118256/comments?per_page=100, merge by created_at with jq or Python json; 'gh api' reaches the same endpoints
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,8 +19,6 @@ Files:
21
  - `repos/microsoft/vscode/issues/118256/events.json`: microsoft/vscode issue #118256: tasks.onDidStartTaskProcess gives the wrong processId if there is a "dependsOn" task.
22
  Format: Two GitHub REST API JSON files fetched from the same issue: events.json (GET /repos/{owner}/{repo}/issues/{number}/events, per_page=100: actor, event, created_at, assignee/assigner, requested_reviewer/review_requester, label, commit_id; 36 entries) and the companion comments.json (GET /repos/{owner}/{repo}/issues/{number}/comments, per_page=100: user, created_at, updated_at, author_association, body; 9 entries). Each is one page with fewer than 100 entries, so both are complete. Together they hold the issue timeline without 'cross-referenced' entries, which embed live repository counters and are not byte-stable (see report).
23
  Origin: Live GitHub REST interface of a closed issue, collected as a NOASSERTION pointer row under spec general rule 4: the events and comments JSON were each fetched twice apart and were byte-identical. Unit files kept only outside the repository for checking and then deleted.
24
- Reference result: facts computed from the issue's own timeline (status changes, assignments, comments and their authors)
25
- Reference source: The pinned responses: repos/microsoft/vscode/issues/118256/events.json and its companion repos/microsoft/vscode/issues/118256/comments.json.
26
  Check: Exact match against values read from the pinned JSON: closed/reopened/merged entries (actor, created_at), assigned/unassigned/review_requested entries (assignee or requested_reviewer, assigner or review_requester, created_at), and comments (author login, created_at). A human comment is a 'commented' entry whose author is not a bot. Bot accounts excluded from the comment count: accounts with type 'Bot', logins ending in '[bot]', and the named automation accounts azure-pipelines, bors, codecov, dotnet-bot, dotnet-issue-labeler, dotnet-maestro, dotnet-policy-service, facebook-github-bot, flutter-dashboard, fluttergithubbot, google-cla, grafanabot, homeassistant, k8s-ci-robot, k8s-triage-robot, linux-foundation-easycla, msftbot, netlify, probot-home-assistant, pytorch-bot, pytorch-probot, rust-highfive, rustbot, typescript-bot, vs-code-engineering, vscode-triage-bot, vscodenpm. Bot accounts seen in this record: github-actions[bot], vscode-triage-bot. Deterministic, no model judgement. Before answers are issued, the extraction program must pass on 2 hand-checked units of this source (spec general rule 5); hand-checking is pending for the delivery step.
27
  Tools: curl -H 'Accept: application/vnd.github+json' https://api.github.com/repos/microsoft/vscode/issues/118256/events?per_page=100 and .../issues/118256/comments?per_page=100, merge by created_at with jq or Python json; 'gh api' reaches the same endpoints
28
  Companions (not included): repos/microsoft/vscode/issues/118256/comments.json
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** curl -H 'Accept: application/vnd.github+json' https://api.github.com/repos/microsoft/vscode/issues/118256/events?per_page=100 and .../issues/118256/comments?per_page=100, merge by created_at with jq or Python json; 'gh api' reaches the same endpoints
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `repos/microsoft/vscode/issues/118256/events.json`: microsoft/vscode issue #118256: tasks.onDidStartTaskProcess gives the wrong processId if there is a "dependsOn" task.
20
  Format: Two GitHub REST API JSON files fetched from the same issue: events.json (GET /repos/{owner}/{repo}/issues/{number}/events, per_page=100: actor, event, created_at, assignee/assigner, requested_reviewer/review_requester, label, commit_id; 36 entries) and the companion comments.json (GET /repos/{owner}/{repo}/issues/{number}/comments, per_page=100: user, created_at, updated_at, author_association, body; 9 entries). Each is one page with fewer than 100 entries, so both are complete. Together they hold the issue timeline without 'cross-referenced' entries, which embed live repository counters and are not byte-stable (see report).
21
  Origin: Live GitHub REST interface of a closed issue, collected as a NOASSERTION pointer row under spec general rule 4: the events and comments JSON were each fetched twice apart and were byte-identical. Unit files kept only outside the repository for checking and then deleted.
 
 
22
  Check: Exact match against values read from the pinned JSON: closed/reopened/merged entries (actor, created_at), assigned/unassigned/review_requested entries (assignee or requested_reviewer, assigner or review_requester, created_at), and comments (author login, created_at). A human comment is a 'commented' entry whose author is not a bot. Bot accounts excluded from the comment count: accounts with type 'Bot', logins ending in '[bot]', and the named automation accounts azure-pipelines, bors, codecov, dotnet-bot, dotnet-issue-labeler, dotnet-maestro, dotnet-policy-service, facebook-github-bot, flutter-dashboard, fluttergithubbot, google-cla, grafanabot, homeassistant, k8s-ci-robot, k8s-triage-robot, linux-foundation-easycla, msftbot, netlify, probot-home-assistant, pytorch-bot, pytorch-probot, rust-highfive, rustbot, typescript-bot, vs-code-engineering, vscode-triage-bot, vscodenpm. Bot accounts seen in this record: github-actions[bot], vscode-triage-bot. Deterministic, no model judgement. Before answers are issued, the extraction program must pass on 2 hand-checked units of this source (spec general rule 5); hand-checking is pending for the delivery step.
23
  Tools: curl -H 'Accept: application/vnd.github+json' https://api.github.com/repos/microsoft/vscode/issues/118256/events?per_page=100 and .../issues/118256/comments?per_page=100, merge by created_at with jq or Python json; 'gh api' reaches the same endpoints
24
  Companions (not included): repos/microsoft/vscode/issues/118256/comments.json
campaign_files/2026-10-m-issue-trackers-2/waikato-busit-gtfs/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the service calendar itself (active services on a date, removed or added dates, day counts). Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -22,8 +20,6 @@ Units:
22
  Read out of the archive: https://wrcscheduledata.blob.core.windows.net/wrcgtfs/busit-nz-public.zip
23
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt)
24
  Origin: Live static GTFS feed published by Waikato Regional Council (Busit) at https://wrcscheduledata.blob.core.windows.net/wrcgtfs/busit-nz-public.zip. calendar.txt and its companions were read out of the whole zip (the server ignores Range) (spec general rule 4: by member); kept as a snapshot pinned by sha256.
25
- Reference result: facts computed from the service calendar itself (active services on a date, removed or added dates, day counts)
26
- Reference source: calendar.txt weekday flags with start_date and end_date, calendar_dates.txt exception rows, agency_timezone of agency.txt, all from the pinned members.
27
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
28
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
29
  Companions (not included): calendar_dates.txt; agency.txt
 
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
12
 
13
  **Licence (SPDX):** NOASSERTION
 
20
  Read out of the archive: https://wrcscheduledata.blob.core.windows.net/wrcgtfs/busit-nz-public.zip
21
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt)
22
  Origin: Live static GTFS feed published by Waikato Regional Council (Busit) at https://wrcscheduledata.blob.core.windows.net/wrcgtfs/busit-nz-public.zip. calendar.txt and its companions were read out of the whole zip (the server ignores Range) (spec general rule 4: by member); kept as a snapshot pinned by sha256.
 
 
23
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
24
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
25
  Companions (not included): calendar_dates.txt; agency.txt
campaign_files/2026-10-m-issue-trackers-2/warrington-passenger-gtfs/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the service calendar itself (active services on a date, removed or added dates, day counts). Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
14
 
15
  **Licence (SPDX):** OGL-UK-3.0
@@ -22,8 +20,6 @@ Units:
22
  Read out of the archive: https://s3-eu-west-1.amazonaws.com/passenger-sources/warrington/gtfs/warrington_1790081092.zip
23
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt)
24
  Origin: Live static GTFS feed published by Warrington's Own Buses (published through Passenger, data.discoverpassenger.com) at https://s3-eu-west-1.amazonaws.com/passenger-sources/warrington/gtfs/warrington_1790081092.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
25
- Reference result: facts computed from the service calendar itself (active services on a date, removed or added dates, day counts)
26
- Reference source: calendar.txt weekday flags with start_date and end_date, calendar_dates.txt exception rows, agency_timezone of agency.txt, all from the pinned members.
27
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
28
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
29
  Companions (not included): calendar_dates.txt; agency.txt
 
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
12
 
13
  **Licence (SPDX):** OGL-UK-3.0
 
20
  Read out of the archive: https://s3-eu-west-1.amazonaws.com/passenger-sources/warrington/gtfs/warrington_1790081092.zip
21
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt)
22
  Origin: Live static GTFS feed published by Warrington's Own Buses (published through Passenger, data.discoverpassenger.com) at https://s3-eu-west-1.amazonaws.com/passenger-sources/warrington/gtfs/warrington_1790081092.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
 
 
23
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
24
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
25
  Companions (not included): calendar_dates.txt; agency.txt
campaign_files/2026-10-md-media-and-documents/unstructured/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Expected list of typed text elements (JSON) of a cp1252 HTML page. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** unstructured-ingest local --encoding cp1252 --metadata-exclude ... --input-path fake-html-cp1252.html (command line) or unstructured.partition.html.partition_html in Python; element JSON compared with the expected file
14
 
15
  **Licence (SPDX):** Apache-2.0
@@ -21,8 +19,6 @@ Files:
21
  - `example-docs/fake-html-cp1252.html`: Unstructured test: Windows-1252 HTML page partitioned into typed elements.
22
  Format: HTML input and one expected JSON file of 4 elements
23
  Origin: a small cp1252 HTML page and the expected element JSON committed by the Unstructured maintainers; the ingest test script local-single-file-with-encoding.sh with check-diff-expected-output.sh is the judge
24
- Reference result: expected list of typed text elements (JSON) of a cp1252 HTML page
25
- Reference source: test_unstructured_ingest/expected-structured-output/local-single-file-with-encoding/fake-html-cp1252.html.json
26
  Check: give the agent the HTML file (decoded as cp1252) and ask for its elements as JSON with type and text; the upstream check is an exact recursive diff after the listed metadata fields are excluded; because metadata.data_source.record_locator.path holds the CI runner path, judge on the ordered list of (type, text) pairs plus languages and filetype
27
  Tools: unstructured-ingest local --encoding cp1252 --metadata-exclude ... --input-path fake-html-cp1252.html (command line) or unstructured.partition.html.partition_html in Python; element JSON compared with the expected file
28
  Companions (not included): test_unstructured_ingest/expected-structured-output/local-single-file-with-encoding/fake-html-cp1252.html.json; test_unstructured_ingest/src/local-single-file-with-encoding.sh; test_unstructured_ingest/check-diff-expected-output.sh; LICENSE.md
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** unstructured-ingest local --encoding cp1252 --metadata-exclude ... --input-path fake-html-cp1252.html (command line) or unstructured.partition.html.partition_html in Python; element JSON compared with the expected file
12
 
13
  **Licence (SPDX):** Apache-2.0
 
19
  - `example-docs/fake-html-cp1252.html`: Unstructured test: Windows-1252 HTML page partitioned into typed elements.
20
  Format: HTML input and one expected JSON file of 4 elements
21
  Origin: a small cp1252 HTML page and the expected element JSON committed by the Unstructured maintainers; the ingest test script local-single-file-with-encoding.sh with check-diff-expected-output.sh is the judge
 
 
22
  Check: give the agent the HTML file (decoded as cp1252) and ask for its elements as JSON with type and text; the upstream check is an exact recursive diff after the listed metadata fields are excluded; because metadata.data_source.record_locator.path holds the CI runner path, judge on the ordered list of (type, text) pairs plus languages and filetype
23
  Tools: unstructured-ingest local --encoding cp1252 --metadata-exclude ... --input-path fake-html-cp1252.html (command line) or unstructured.partition.html.partition_html in Python; element JSON compared with the expected file
24
  Companions (not included): test_unstructured_ingest/expected-structured-output/local-single-file-with-encoding/fake-html-cp1252.html.json; test_unstructured_ingest/src/local-single-file-with-encoding.sh; test_unstructured_ingest/check-diff-expected-output.sh; LICENSE.md
campaign_files/2026-10-md-media-and-documents/voa-learning-english-3160704/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2015/03/16/8a728485-a26d-4efd-b74e-507e6091f9a3.mp3`: VOA Learning English American Stories "The Celebrated Jumping Frog of Calaveras County" by Mark Twain (adapted) program audio with its published story text.
22
  Format: MP3, 6,489,581 bytes, about 13.5 min, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script; read literary story (American Stories), a VOA adaptation of Mark Twain's 1865 story
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: Here is Shep O'Neal with the story." up to the sign-off "I just left."; the credit line ("This story was written by Mark Twain and adapted for Learning English by Karen Leggett."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed; the editor's note at the top and the Angels Camp photo paragraph are not read as story text and are checked against the audio before use; the underlying 1865 Twain text is public domain and the adaptation is VOA's own
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/celebrated-jumping-frog-calaveras-county-mark-twain/3160704.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2015/03/16/8a728485-a26d-4efd-b74e-507e6091f9a3.mp3`: VOA Learning English American Stories "The Celebrated Jumping Frog of Calaveras County" by Mark Twain (adapted) program audio with its published story text.
20
  Format: MP3, 6,489,581 bytes, about 13.5 min, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script; read literary story (American Stories), a VOA adaptation of Mark Twain's 1865 story
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/celebrated-jumping-frog-calaveras-county-mark-twain/3160704.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-7932530/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/01/08/1f0091c2-edc1-4d94-9fbd-cc9bdab3a832.mp3`: VOA Learning English Words and Their Stories "The Story of Jack Frost" (2025-01-11) program audio with its published story text.
22
  Format: MP3, 2,668,113 bytes, about 5.5 min, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script; read story with short acted dialogue skits (Words and Their Stories); more than one voice
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/the-story-of-jack-frost-/7932530.html from "Now, from VOA Learning English, it's time for Words and Their Stories." up to the sign-off "I'm Anna Matteo."; the credit line ("Anna Matteo wrote this story for VOA Learning English."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/the-story-of-jack-frost-/7932530.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/01/08/1f0091c2-edc1-4d94-9fbd-cc9bdab3a832.mp3`: VOA Learning English Words and Their Stories "The Story of Jack Frost" (2025-01-11) program audio with its published story text.
20
  Format: MP3, 2,668,113 bytes, about 5.5 min, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script; read story with short acted dialogue skits (Words and Their Stories); more than one voice
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/the-story-of-jack-frost-/7932530.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-7966519/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/02/07/2ac8b9f9-2d19-4d57-b449-ccbab1eac13b.mp3`: VOA Learning English American Inventions "The Television" (2025-02-11) program audio with its published story text.
22
  Format: MP3, 2,303,208 bytes, about 4.8 min, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script; read feature story (Science and Technology)
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/american-inventions-the-television-/7966519.html from "Television has made it possible for people around the world to share in many different experiences" up to the sign-off "I'm John Russell."; the credit line ("Paul Thompson wrote this story for VOA Learning English. John Russell adapted it."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/american-inventions-the-television-/7966519.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/02/07/2ac8b9f9-2d19-4d57-b449-ccbab1eac13b.mp3`: VOA Learning English American Inventions "The Television" (2025-02-11) program audio with its published story text.
20
  Format: MP3, 2,303,208 bytes, about 4.8 min, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script; read feature story (Science and Technology)
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/american-inventions-the-television-/7966519.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-7979814/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/02/20/57f1d743-7f93-4e15-fd51-08dd4a817620.mp3`: VOA Learning English Words and Their Stories "The Importance of Being 'Dialed In'" (2025-02-22) program audio with its published story text.
22
  Format: MP3, 1,930,634 bytes, about 4 min, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script; read story with idiom explanations (Words and Their Stories)
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/the-importance-of-being-dialed-in-/7979814.html from "And now, Words and Their Stories, from VOA Learning English." up to the sign-off "Until next time, I'm Anna Matteo."; the credit line ("Anna Matteo wrote this story for VOA Learning English."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/the-importance-of-being-dialed-in-/7979814.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/02/20/57f1d743-7f93-4e15-fd51-08dd4a817620.mp3`: VOA Learning English Words and Their Stories "The Importance of Being 'Dialed In'" (2025-02-22) program audio with its published story text.
20
  Format: MP3, 1,930,634 bytes, about 4 min, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script; read story with idiom explanations (Words and Their Stories)
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/the-importance-of-being-dialed-in-/7979814.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-7980848/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/02/27/88d03fae-92b2-4f86-902b-08dd4a843460.mp3`: VOA Learning English Words and Their Stories "How to 'Dish Up' Something Good" (2025-03-01) program audio with its published story text.
22
  Format: MP3, 1,771,355 bytes, about 3.7 min, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script; read story with idiom explanations (Words and Their Stories)
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/how-to-dish-up-something-good/7980848.html from "And now, Words and Their Stories, from VOA Learning English." up to the sign-off "Until next time, I'm Anna Matteo."; the credit line ("Anna Matteo wrote this story for VOA Learning English."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/how-to-dish-up-something-good/7980848.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/02/27/88d03fae-92b2-4f86-902b-08dd4a843460.mp3`: VOA Learning English Words and Their Stories "How to 'Dish Up' Something Good" (2025-03-01) program audio with its published story text.
20
  Format: MP3, 1,771,355 bytes, about 3.7 min, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script; read story with idiom explanations (Words and Their Stories)
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/how-to-dish-up-something-good/7980848.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-a-short-history-of-spacecraft-landings-on-the-mo/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/03/06/283ea8f7-a260-494e-2d69-08dd5c897904.mp3`: VOA Learning English "A Short History of Spacecraft Landings on the Moon" program audio with its published story text.
22
  Format: MP3, 2,658,550 bytes, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/a-short-history-of-spacecraft-landings-on-the-moon/8001341.html up to the sign-off "I'm Bryan Lynn."; the credit line ("The Associated Press reported this story. Bryan Lynn adapted the report for VOA Learning English with additional information from NASA."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/a-short-history-of-spacecraft-landings-on-the-moon/8001341.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `vle/2025/03/06/283ea8f7-a260-494e-2d69-08dd5c897904.mp3`: VOA Learning English "A Short History of Spacecraft Landings on the Moon" program audio with its published story text.
20
  Format: MP3, 2,658,550 bytes, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/a-short-history-of-spacecraft-landings-on-the-moon/8001341.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-american-inventors-rocket-scientist-robert-godda/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/02/20/03286c2d-5139-466c-e52a-08dd4a81f35f.mp3`: VOA Learning English "American Inventors: Rocket Scientist Robert Goddard" program audio with its published story text.
22
  Format: MP3, 1,633,219 bytes, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/american-inventors-rocket-scientist-robert-goddard/7979682.html up to the sign-off "I'm Bryan Lynn."; the credit line ("Shirley Griffith reported on this story for VOA Learning English. Bryan Lynn adapted it."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/american-inventors-rocket-scientist-robert-goddard/7979682.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/02/20/03286c2d-5139-466c-e52a-08dd4a81f35f.mp3`: VOA Learning English "American Inventors: Rocket Scientist Robert Goddard" program audio with its published story text.
20
  Format: MP3, 1,633,219 bytes, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/american-inventors-rocket-scientist-robert-goddard/7979682.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-american-places/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/02/06/4f3c93f1-910e-4a8e-83a9-5a3394b41f0c.mp3`: VOA Learning English "American Places: Gettysburg, Pennsylvania" program audio with its published story text.
22
  Format: MP3, mono 44.1 kHz, 64 kbit/s, about 239 s, slow-paced read story
23
  Origin: VOA Learning English program read from its own published script
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/american-places-gettysburg-pennsylvania/7965779.html, from "The Gettysburg National Military Park is in the small town of Gettysburg ..." to the sign-off "I'm Mario Ritter, Jr."; the "Words in This Story" glossary after the rule line is not read
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides
27
  Companions (not included): a/american-places-gettysburg-pennsylvania/7965779.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/02/06/4f3c93f1-910e-4a8e-83a9-5a3394b41f0c.mp3`: VOA Learning English "American Places: Gettysburg, Pennsylvania" program audio with its published story text.
20
  Format: MP3, mono 44.1 kHz, 64 kbit/s, about 239 s, slow-paced read story
21
  Origin: VOA Learning English program read from its own published script
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides
23
  Companions (not included): a/american-places-gettysburg-pennsylvania/7965779.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-scientists-discover-stonehenge-like-circle-in-de/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/02/27/f80178cd-74b6-4a3a-0343-08dd4a817621.mp3`: VOA Learning English "Scientists Discover Stonehenge-Like Circle in Denmark" program audio with its published story text.
22
  Format: MP3, 1,575,959 bytes, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/scientists-discover-stonehenge-like-circle-in-denmark/7990025.html up to the sign-off "I'm John Russell."; the credit line ("John Russell adapted this story based on an AFP report."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/scientists-discover-stonehenge-like-circle-in-denmark/7990025.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `vle/2025/02/27/f80178cd-74b6-4a3a-0343-08dd4a817621.mp3`: VOA Learning English "Scientists Discover Stonehenge-Like Circle in Denmark" program audio with its published story text.
20
  Format: MP3, 1,575,959 bytes, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/scientists-discover-stonehenge-like-circle-in-denmark/7990025.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-watching-the-grass-grow-is-not-fun/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/03/11/08bbcb88-2304-42bd-0531-08dd5c8b1668.mp3`: VOA Learning English Words and Their Stories "Watching the Grass Grow Is Not Fun" program audio with its published story text.
22
  Format: MP3, 1,599,192 bytes, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/watching-the-grass-grow-is-not-fun/8003108.html up to the sign-off "Until next time, I'm Anna Matteo."; the credit line ("Anna Matteo wrote this story for VOA Learning English."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/watching-the-grass-grow-is-not-fun/8003108.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/03/11/08bbcb88-2304-42bd-0531-08dd5c8b1668.mp3`: VOA Learning English Words and Their Stories "Watching the Grass Grow Is Not Fun" program audio with its published story text.
20
  Format: MP3, 1,599,192 bytes, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/watching-the-grass-grow-is-not-fun/8003108.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-wilbur-and-orville-wright-the-first-airplane/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/03/05/e897fcf6-3ce1-4da9-a5cf-08dd5afe0892.mp3`: VOA Learning English "Wilbur and Orville Wright: The First Airplane" program audio with its published story text.
22
  Format: MP3, 2,132,889 bytes, slow-paced read story (Learning English pace)
23
  Origin: VOA Learning English program read from its own published script
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of a/wilbur-and-orville-wright-the-first-airplane/7998765.html up to the sign-off "I'm John Russell."; the credit line ("Marilyn Rice Christiano wrote this story for VOA Learning English. John Russell adapted it."), quiz box, photo captions and the "Words in This Story" glossary after the rule line are not read and are removed
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
27
  Companions (not included): a/wilbur-and-orville-wright-the-first-airplane/7998765.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/03/05/e897fcf6-3ce1-4da9-a5cf-08dd5afe0892.mp3`: VOA Learning English "Wilbur and Orville Wright: The First Airplane" program audio with its published story text.
20
  Format: MP3, 2,132,889 bytes, slow-paced read story (Learning English pace)
21
  Origin: VOA Learning English program read from its own published script
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions, quiz and page furniture removed); case and punctuation folded on both sides
23
  Companions (not included): a/wilbur-and-orville-wright-the-first-airplane/7998765.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voa-learning-english-yellowstone/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript (the published script of the read story). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides.
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,7 +19,5 @@ Files:
21
  - `vle/2025/02/13/776add5c-bbc5-4a14-83a0-08dd4a843460.mp3`: VOA Learning English "Yellowstone: The World's First National Park" program audio with its published story text.
22
  Format: MP3, mono 44.1 kHz, 64 kbit/s, about 240 s, slow-paced read story
23
  Origin: VOA Learning English program read from its own published script
24
- Reference result: orthographic transcript (the published script of the read story)
25
- Reference source: the story body paragraphs of the story page, from "Yellowstone National Park in the western United States was the world's first national park." to "I'm Andrew Smith."; the credit line and the glossary after the rule line are not read
26
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides
27
  Companions (not included): a/yellowstone-the-world-s-first-national-park/7974034.html; p/6021.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides.
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `vle/2025/02/13/776add5c-bbc5-4a14-83a0-08dd4a843460.mp3`: VOA Learning English "Yellowstone: The World's First National Park" program audio with its published story text.
20
  Format: MP3, mono 44.1 kHz, 64 kbit/s, about 240 s, slow-paced read story
21
  Origin: VOA Learning English program read from its own published script
 
 
22
  Check: give the agent the mp3 only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; reference = the story body paragraphs (glossary, captions and page furniture removed), case and punctuation folded on both sides
23
  Companions (not included): a/yellowstone-the-world-s-first-national-park/7974034.html; p/6021.html
campaign_files/2026-10-md-media-and-documents/voxforge-16khz-16bit-main/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcript per utterance. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the 10 wav files only (remove etc/prompts-original, etc/PROMPTS and the HTK logs); it writes one transcript per file.
14
 
15
  **Licence (SPDX):** GPL-3.0-or-later
@@ -21,12 +19,8 @@ Files:
21
  - `md053-01/Bitbrit-20090528-yit.tgz`: VoxForge submission Bitbrit-20090528-yit: 10 read English sentences (b0176 to b0185) with prompts.
22
  Format: gzip tar archive (789467 bytes): Bitbrit-20090528-yit/wav/b0176.wav to b0185.wav (16 kHz 16-bit), etc/prompts-original, etc/PROMPTS, etc/README, LICENSE, etc/GPL_license.txt, alignment logs
23
  Origin: volunteer read speech submitted through the VoxForge web applet; prompts are CMU ARCTIC b-set sentences
24
- Reference result: orthographic transcript per utterance
25
- Reference source: Bitbrit-20090528-yit/etc/prompts-original lines 1 to 10 (line 1: "b0176 I'll see to poor Hughie."); the prompts were read aloud by the speaker; etc/PROMPTS holds the same text upper-cased
26
  Check: give the agent the 10 wav files only (remove etc/prompts-original, etc/PROMPTS and the HTK logs); it writes one transcript per file.
27
  - `md053-02/Blenderkitty-20090615-qcv.tgz`: VoxForge submission Blenderkitty-20090615-qcv: 10 read English sentences (b0419 to b0428) with prompts.
28
  Format: gzip tar archive (1169027 bytes): Blenderkitty-20090615-qcv/wav/b0419.wav to b0428.wav (16 kHz 16-bit), etc/prompts-original, etc/PROMPTS, etc/README, LICENSE, etc/GPL_license.txt, alignment logs
29
  Origin: volunteer read speech submitted through the VoxForge web applet; prompts are CMU ARCTIC b-set sentences
30
- Reference result: orthographic transcript per utterance
31
- Reference source: Blenderkitty-20090615-qcv/etc/prompts-original lines 1 to 10 (line 1: "b0419 Your father's fifth command, he nodded."); the prompts were read aloud by the speaker; etc/PROMPTS holds the same text upper-cased
32
  Check: give the agent the 10 wav files only (remove etc/prompts-original, etc/PROMPTS and the HTK logs); it writes one transcript per file.
 
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the 10 wav files only (remove etc/prompts-original, etc/PROMPTS and the HTK logs); it writes one transcript per file.
12
 
13
  **Licence (SPDX):** GPL-3.0-or-later
 
19
  - `md053-01/Bitbrit-20090528-yit.tgz`: VoxForge submission Bitbrit-20090528-yit: 10 read English sentences (b0176 to b0185) with prompts.
20
  Format: gzip tar archive (789467 bytes): Bitbrit-20090528-yit/wav/b0176.wav to b0185.wav (16 kHz 16-bit), etc/prompts-original, etc/PROMPTS, etc/README, LICENSE, etc/GPL_license.txt, alignment logs
21
  Origin: volunteer read speech submitted through the VoxForge web applet; prompts are CMU ARCTIC b-set sentences
 
 
22
  Check: give the agent the 10 wav files only (remove etc/prompts-original, etc/PROMPTS and the HTK logs); it writes one transcript per file.
23
  - `md053-02/Blenderkitty-20090615-qcv.tgz`: VoxForge submission Blenderkitty-20090615-qcv: 10 read English sentences (b0419 to b0428) with prompts.
24
  Format: gzip tar archive (1169027 bytes): Blenderkitty-20090615-qcv/wav/b0419.wav to b0428.wav (16 kHz 16-bit), etc/prompts-original, etc/PROMPTS, etc/README, LICENSE, etc/GPL_license.txt, alignment logs
25
  Origin: volunteer read speech submitted through the VoxForge web applet; prompts are CMU ARCTIC b-set sentences
 
 
26
  Check: give the agent the 10 wav files only (remove etc/prompts-original, etc/PROMPTS and the HTK logs); it writes one transcript per file.
campaign_files/2026-10-ml-serving-configs-2/vllm-gsm8k-server-gate/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_aggregate_metric. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Serve the model with vllm serve plus server_args, run gsm8k_eval.py against it for 1319 questions 5-shot, pass when measured accuracy >= accuracy_threshold - tolerance, tolerance default 0.08 absolute (test_gsm8k_correctness.py lines 184 to 196).
14
 
15
  **Licence (SPDX):** Apache-2.0
@@ -20,8 +18,6 @@ Files:
20
 
21
  - `tests/evals/gsm8k/configs/Llama-3.2-1B-Instruct-INT8-CT.yaml`: vLLM server gsm8k accuracy gate: RedHatAI/Llama-3.2-1B-Instruct-quantized.w8a8, 1319 questions 5-shot, accuracy_threshold 0.31, served with --enforce-eager --max-model-len 4096.
22
  Format: YAML: model_name, accuracy_threshold, num_questions, num_fewshot, server_args
23
- Reference result: expected_aggregate_metric
24
- Reference source: accuracy_threshold 0.31 in the YAML (line 2)
25
  Check: Serve the model with vllm serve plus server_args, run gsm8k_eval.py against it for 1319 questions 5-shot, pass when measured accuracy >= accuracy_threshold - tolerance, tolerance default 0.08 absolute (test_gsm8k_correctness.py lines 184 to 196)
26
  Companions (not included): tests/evals/gsm8k/test_gsm8k_correctness.py; tests/evals/gsm8k/gsm8k_eval.py; tests/evals/gsm8k/README.md; LICENSE
27
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu; started as an OpenAI-compatible server (vllm serve) with the server_args of the config (configs/Llama-3.2-1B-Instruct-INT8-CT.yaml line 5); test_gsm8k_correctness.py imports tests.utils.RemoteOpenAIServer from the same commit; RedHatAI/Llama-3.2-1B-Instruct-quantized.w8a8 (Llama 3.2 Community License; gated: licence acceptance on Hugging Face). gsm8k test set as downloaded by gsm8k_eval.py (openai/gsm8k, MIT).
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Serve the model with vllm serve plus server_args, run gsm8k_eval.py against it for 1319 questions 5-shot, pass when measured accuracy >= accuracy_threshold - tolerance, tolerance default 0.08 absolute (test_gsm8k_correctness.py lines 184 to 196).
12
 
13
  **Licence (SPDX):** Apache-2.0
 
18
 
19
  - `tests/evals/gsm8k/configs/Llama-3.2-1B-Instruct-INT8-CT.yaml`: vLLM server gsm8k accuracy gate: RedHatAI/Llama-3.2-1B-Instruct-quantized.w8a8, 1319 questions 5-shot, accuracy_threshold 0.31, served with --enforce-eager --max-model-len 4096.
20
  Format: YAML: model_name, accuracy_threshold, num_questions, num_fewshot, server_args
 
 
21
  Check: Serve the model with vllm serve plus server_args, run gsm8k_eval.py against it for 1319 questions 5-shot, pass when measured accuracy >= accuracy_threshold - tolerance, tolerance default 0.08 absolute (test_gsm8k_correctness.py lines 184 to 196)
22
  Companions (not included): tests/evals/gsm8k/test_gsm8k_correctness.py; tests/evals/gsm8k/gsm8k_eval.py; tests/evals/gsm8k/README.md; LICENSE
23
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu; started as an OpenAI-compatible server (vllm serve) with the server_args of the config (configs/Llama-3.2-1B-Instruct-INT8-CT.yaml line 5); test_gsm8k_correctness.py imports tests.utils.RemoteOpenAIServer from the same commit; RedHatAI/Llama-3.2-1B-Instruct-quantized.w8a8 (Llama 3.2 Community License; gated: licence acceptance on Hugging Face). gsm8k test set as downloaded by gsm8k_eval.py (openai/gsm8k, MIT).
campaign_files/2026-10-ml-serving-configs-2/vllm-lm-eval-gate/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 5 of 5 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_aggregate_metric. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1319 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does; Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1000 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does; Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 250 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does.
14
 
15
  **Licence (SPDX):** Apache-2.0
@@ -20,36 +18,26 @@ Files:
20
 
21
  - `dot-buildkite/lm-eval-harness/configs/Qwen2.5-1.5B-Instruct.yaml`: vLLM CI lm-eval accuracy gate config: Qwen/Qwen2.5-1.5B-Instruct on gsm8k 5-shot, limit 1319, expected strict-match 0.54, flexible-extract 0.59.
22
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
23
- Reference result: expected_aggregate_metric
24
- Reference source: metrics values in the YAML (strict-match 0.54, flexible-extract 0.59), produced by the baseline command in the header comment
25
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1319 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
26
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
27
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). Qwen/Qwen2.5-1.5B-Instruct (Apache-2.0 weights). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
28
  - `dot-buildkite/lm-eval-harness/configs/Meta-Llama-3.2-1B-Instruct-INT8-compressed-tensors.yaml`: vLLM CI lm-eval accuracy gate config: neuralmagic/Llama-3.2-1B-Instruct-quantized.w8a8 on gsm8k 5-shot, limit 1000, expected strict-match 0.356, flexible-extract 0.358.
29
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
30
- Reference result: expected_aggregate_metric
31
- Reference source: metrics values in the YAML (strict-match 0.356, flexible-extract 0.358), produced by the baseline command in the header comment
32
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1000 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
33
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
34
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). neuralmagic/Llama-3.2-1B-Instruct-quantized.w8a8 (Llama 3.2 Community License; gated: licence acceptance on Hugging Face). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
35
  - `dot-buildkite/lm-eval-harness/configs/Qwen2-1.5B-Instruct-INT8-compressed-tensors.yaml`: vLLM CI lm-eval accuracy gate config: neuralmagic/Qwen2-1.5B-Instruct-quantized.w8a8 on gsm8k 5-shot, limit 1000, expected strict-match 0.593, flexible-extract 0.588.
36
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
37
- Reference result: expected_aggregate_metric
38
- Reference source: metrics values in the YAML (strict-match 0.593, flexible-extract 0.588), produced by the baseline command in the header comment
39
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1000 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
40
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
41
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). neuralmagic/Qwen2-1.5B-Instruct-quantized.w8a8 (Apache-2.0 weights). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
42
  - `dot-buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct.yaml`: vLLM CI lm-eval accuracy gate config: meta-llama/Meta-Llama-3-8B-Instruct on gsm8k 5-shot, limit 250, expected strict-match 0.756, flexible-extract 0.752.
43
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
44
- Reference result: expected_aggregate_metric
45
- Reference source: metrics values in the YAML (strict-match 0.756, flexible-extract 0.752), produced by the baseline command in the header comment
46
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 250 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
47
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; .buildkite/lm-eval-harness/run-lm-eval-gsm-hf-baseline.sh; LICENSE
48
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). meta-llama/Meta-Llama-3-8B-Instruct (Meta Llama 3 Community License; gated download on Hugging Face); Hugging Face model repo meta-llama/Meta-Llama-3-8B-Instruct. gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval. transformers backend for the HF baseline (run-lm-eval-gsm-hf-baseline.sh, now a companion) only if the baseline itself is re-derived.
49
  - `dot-buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct-INT8-compressed-tensors.yaml`: vLLM CI lm-eval accuracy gate config: nm-testing/Meta-Llama-3-8B-Instruct-W8-Channel-A8-Dynamic-Per-Token-Test on gsm8k 5-shot, limit 250, expected strict-match 0.728, flexible-extract 0.728.
50
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
51
- Reference result: expected_aggregate_metric
52
- Reference source: metrics values in the YAML (strict-match 0.728, flexible-extract 0.728), produced by the baseline command in the header comment
53
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 250 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
54
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
55
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). nm-testing/Meta-Llama-3-8B-Instruct-W8-Channel-A8-Dynamic-Per-Token-Test (Meta Llama 3 Community License; gated terms of the base model). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 5 of 5 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1319 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does; Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1000 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does; Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 250 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does.
12
 
13
  **Licence (SPDX):** Apache-2.0
 
18
 
19
  - `dot-buildkite/lm-eval-harness/configs/Qwen2.5-1.5B-Instruct.yaml`: vLLM CI lm-eval accuracy gate config: Qwen/Qwen2.5-1.5B-Instruct on gsm8k 5-shot, limit 1319, expected strict-match 0.54, flexible-extract 0.59.
20
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
 
 
21
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1319 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
22
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
23
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). Qwen/Qwen2.5-1.5B-Instruct (Apache-2.0 weights). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
24
  - `dot-buildkite/lm-eval-harness/configs/Meta-Llama-3.2-1B-Instruct-INT8-compressed-tensors.yaml`: vLLM CI lm-eval accuracy gate config: neuralmagic/Llama-3.2-1B-Instruct-quantized.w8a8 on gsm8k 5-shot, limit 1000, expected strict-match 0.356, flexible-extract 0.358.
25
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
 
 
26
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1000 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
27
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
28
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). neuralmagic/Llama-3.2-1B-Instruct-quantized.w8a8 (Llama 3.2 Community License; gated: licence acceptance on Hugging Face). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
29
  - `dot-buildkite/lm-eval-harness/configs/Qwen2-1.5B-Instruct-INT8-compressed-tensors.yaml`: vLLM CI lm-eval accuracy gate config: neuralmagic/Qwen2-1.5B-Instruct-quantized.w8a8 on gsm8k 5-shot, limit 1000, expected strict-match 0.593, flexible-extract 0.588.
30
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
 
 
31
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 1000 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
32
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
33
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). neuralmagic/Qwen2-1.5B-Instruct-quantized.w8a8 (Apache-2.0 weights). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
34
  - `dot-buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct.yaml`: vLLM CI lm-eval accuracy gate config: meta-llama/Meta-Llama-3-8B-Instruct on gsm8k 5-shot, limit 250, expected strict-match 0.756, flexible-extract 0.752.
35
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
 
 
36
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 250 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
37
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; .buildkite/lm-eval-harness/run-lm-eval-gsm-hf-baseline.sh; LICENSE
38
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). meta-llama/Meta-Llama-3-8B-Instruct (Meta Llama 3 Community License; gated download on Hugging Face); Hugging Face model repo meta-llama/Meta-Llama-3-8B-Instruct. gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval. transformers backend for the HF baseline (run-lm-eval-gsm-hf-baseline.sh, now a companion) only if the baseline itself is re-derived.
39
  - `dot-buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct-INT8-compressed-tensors.yaml`: vLLM CI lm-eval accuracy gate config: nm-testing/Meta-Llama-3-8B-Instruct-W8-Channel-A8-Dynamic-Per-Token-Test on gsm8k 5-shot, limit 250, expected strict-match 0.728, flexible-extract 0.728.
40
  Format: YAML: model_name, tasks[gsm8k].metrics[name,value], limit, num_fewshot; header comment gives the exact baseline command
 
 
41
  Check: Run lm-eval gsm8k with the vLLM backend (or HF backend where the header says so) with num_fewshot 5 and limit 250 on CPU; pass when each measured metric >= expected * (1 - rtol), rtol default 0.08, as test_lm_eval_correctness.py does
42
  Companions (not included): .buildkite/lm-eval-harness/test_lm_eval_correctness.py; .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh; LICENSE
43
  Dependencies: vLLM at commit 049507aa76200090f8bc6ca4e48ab0e09f2f0d29 (committed 2026-10-06T13:01:04Z), CPU build VLLM_TARGET_DEVICE=cpu. lm-eval[api] 0.4.13 (latest PyPI release on or before the vLLM commit date 2026-10-06, released 2026-08-31; run-lm-eval-gsm-vllm-baseline.sh line 6 asks for >=0.4.12). nm-testing/Meta-Llama-3-8B-Instruct-W8-Channel-A8-Dynamic-Per-Token-Test (Meta Llama 3 Community License; gated terms of the base model). gsm8k test split (openai/gsm8k, MIT) as downloaded by lm-eval.
campaign_files/2026-10-ml-serving-configs-2/vllm-tool-and-reasoning-parser-test-expectations/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 10 of 10 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_test_outputs. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Tools are named in the check written per file in the notes.
14
 
15
  **Licence (SPDX):** Apache-2.0
@@ -21,80 +19,60 @@ Files:
21
  - `tests/tool_parsers/test_minimax_m2_tool_parser.py`: vLLM MiniMax-M2 tool-call parser: 21 tests feed raw MiniMax-M2 output text (whole and streamed in deltas) and state the exact tool calls, arguments and content the parser must extract.
22
  Format: pytest Python module holding model output strings, tool schemas and the expected parsed tool calls inline, with an in-file fake tokenizer
23
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
24
- Reference result: expected_test_outputs
25
- Reference source: the expected tool names, argument dicts, content strings and delta sequences asserted in each test, written by the maintainers
26
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_minimax_m2_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. The file defines its own FakeTokenizer, so nothing is downloaded. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
27
  Companions (not included): LICENSE
28
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled
29
  - `tests/tool_parsers/test_hermes_tool_parser.py`: vLLM Hermes and Granite4 tool-call parsers: 13 tests, each run for both parsers, feed Hermes-format <tool_call> output (whole and streamed) and state the exact tool calls and content to extract.
30
  Format: pytest Python module holding model output strings and expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
31
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
32
- Reference result: expected_test_outputs
33
- Reference source: the expected tool names, argument JSON and content strings asserted in each test, written by the maintainers
34
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_hermes_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. One test is skipped upstream for the granite4 parser (line 361, the parser enforces a complete response). Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
35
  Companions (not included): LICENSE
36
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer Qwen/Qwen3-32B from the Hugging Face hub (test file line 28), licence apache-2.0, not gated; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-32B fetched 2026-10-10T19:34:49Z (sha256 5723ebdcbf86aa74..., robots allowed)
37
  - `tests/tool_parsers/test_qwen3coder_tool_parser.py`: vLLM Qwen3-Coder XML tool-call parser: 23 tests, run for the chat-completion and responses API paths, feed <function=...><parameter=...> output and state the exact typed arguments and content to extract.
38
  Format: pytest Python module holding model output strings, tool schemas and the expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
39
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
40
- Reference result: expected_test_outputs
41
- Reference source: the expected tool names, typed argument values and content strings asserted in each test, written by the maintainers
42
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_qwen3coder_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
43
  Companions (not included): LICENSE
44
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 from the Hugging Face hub (test file line 31), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 fetched 2026-10-10T19:35:07Z (sha256 c7e272aa634ac64b..., robots allowed)
45
  - `tests/tool_parsers/test_jamba_tool_parser.py`: vLLM Jamba tool-call parser: 4 parametrized tests feed Jamba <tool_calls> JSON output (whole and streamed token by token) and state the exact tool calls and content to extract.
46
  Format: pytest Python module holding model output strings and expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
47
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
48
- Reference result: expected_test_outputs
49
- Reference source: the expected tool calls and content strings in each parametrize case, written by the maintainers
50
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_jamba_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
51
  Companions (not included): LICENSE
52
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer ai21labs/Jamba-tiny-dev from the Hugging Face hub (test file line 16), licence apache-2.0, not gated; model card metadata https://huggingface.co/api/models/ai21labs/Jamba-tiny-dev fetched 2026-10-10T19:34:53Z (sha256 c4a4b391c83b9e8c..., robots allowed)
53
  - `tests/tool_parsers/test_llama3_json_tool_parser.py`: vLLM Llama 3 JSON tool-call parser: 15 tests feed Llama 3.x JSON tool-call output (single, multiple, with surrounding text, malformed) and state the exact tool calls and content to extract.
54
  Format: pytest Python module holding model output strings and expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
55
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
56
- Reference result: expected_test_outputs
57
- Reference source: the expected tool names, argument JSON and content strings asserted in each test, written by the maintainers
58
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_llama3_json_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
59
  Companions (not included): LICENSE
60
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer meta-llama/Llama-3.2-1B-Instruct from the Hugging Face hub (test file line 12), licence llama3.2 (Llama 3.2 Community License), GATED: the hub asks each user to accept the licence before download (gated: manual); the test needs an HF token of an account that accepted it; model card metadata https://huggingface.co/api/models/meta-llama/Llama-3.2-1B-Instruct fetched 2026-10-10T19:35:03Z (sha256 4c82059d48b4413c..., robots allowed)
61
  - `tests/tool_parsers/test_deepseekv32_tool_parser.py`: vLLM DeepSeek-V3.2 DSML tool-call parser: 48 tests feed DSML-delimited output (whole and streamed, delimiters split across deltas) and state the exact tool calls, typed arguments and content to extract.
62
  Format: pytest Python module holding model output strings and expected parsed tool calls inline, with a MagicMock tokenizer and the shared streaming helper module
63
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
64
- Reference result: expected_test_outputs
65
- Reference source: the expected tool names, argument values and content strings asserted in each test, written by the maintainers
66
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_deepseekv32_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. The tokenizer is a MagicMock (MOCK_TOKENIZER, file line 32), so nothing is downloaded. All comparisons are exact equality except two float arguments compared with pytest.approx(3.14) at lines 435 and 1126, which use the pytest default tolerance. The file sets pytestmark = pytest.mark.skip_global_cleanup (line 24), a marker vLLM registers in pyproject.toml line 123; without that registration pytest only warns.
67
  Companions (not included): tests/__init__.py; tests/tool_parsers/__init__.py; tests/tool_parsers/utils.py; LICENSE
68
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled
69
  - `tests/tool_parsers/test_pythonic_tool_parser.py`: vLLM pythonic tool-call parser: 4 tests over 14 parametrized cases feed Python-call-list output ([f(a=1), g(b="x")]) whole and streamed and state the exact tool calls to extract.
70
  Format: pytest Python module holding model output strings and expected parsed tool calls inline, with the shared helper module and the tokenizer fixture
71
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
72
- Reference result: expected_test_outputs
73
- Reference source: the expected FunctionCall objects and content in each pytest.param case, written by the maintainers
74
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_pythonic_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
75
  Companions (not included): tests/__init__.py; tests/tool_parsers/__init__.py; tests/tool_parsers/utils.py; tests/tool_parsers/conftest.py; LICENSE
76
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer gpt2 (openai-community/gpt2) from the Hugging Face hub via tests/tool_parsers/conftest.py, licence mit, not gated; model card metadata https://huggingface.co/api/models/openai-community/gpt2 fetched 2026-10-10T19:35:31Z (sha256 ef3a19a264185920..., robots allowed)
77
  - `tests/tool_parsers/test_kimi_k2_tool_parser.py`: vLLM Kimi-K2 tool-call parser: 26 tests feed Kimi-K2 section-token output (whole, streamed and at varying stream intervals) and state the exact tool-call ids, names, arguments and content to extract.
78
  Format: pytest Python module holding model output strings and expected parsed tool calls inline, with the shared helper module; tokenizer fetched from the Hugging Face hub at run time
79
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
80
- Reference result: expected_test_outputs
81
- Reference source: the expected tool-call ids, names, argument JSON and content strings asserted in each test, written by the maintainers
82
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_kimi_k2_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
83
  Companions (not included): tests/__init__.py; tests/tool_parsers/__init__.py; tests/tool_parsers/utils.py; LICENSE
84
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer moonshotai/Kimi-K2-Instruct from the Hugging Face hub, loaded with trust_remote_code=True (test file line 25), so the test executes the tokenizer code published in that model repository; licence modified-mit (cardData license other, license_name modified-mit), not gated; model card metadata https://huggingface.co/api/models/moonshotai/Kimi-K2-Instruct fetched 2026-10-10T19:34:56Z (sha256 045f36fa880a85ed..., robots allowed)
85
  - `tests/reasoning/test_deepseekr1_reasoning_parser.py`: vLLM DeepSeek-R1 reasoning parser: 24 parametrized cases feed <think>...</think> output (whole and streamed, with and without the start tag, truncated) and state the exact reasoning and content strings to split out.
86
  Format: pytest Python module holding model output strings and the expected reasoning and content split inline, with the shared reasoning helper module
87
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
88
- Reference result: expected_test_outputs
89
- Reference source: the reasoning and content values of each pytest.param case, written by the maintainers
90
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/reasoning/test_deepseekr1_reasoning_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
91
  Companions (not included): tests/__init__.py; tests/reasoning/__init__.py; tests/reasoning/utils.py; LICENSE
92
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B from the Hugging Face hub (test file line 14, loaded with transformers AutoTokenizer at line 19), licence mit, not gated; model card metadata https://huggingface.co/api/models/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B fetched 2026-10-10T19:35:17Z (sha256 6bc6a9009545d581..., robots allowed)
93
  - `tests/reasoning/test_qwen3_reasoning_parser.py`: vLLM Qwen3 reasoning parser: 28 parametrized cases, each run under three Qwen3 tokenizers, feed <think> output and state the exact reasoning and content strings to split out, whole and streamed.
94
  Format: pytest Python module holding model output strings and the expected reasoning and content split inline, with the shared reasoning helper module
95
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
96
- Reference result: expected_test_outputs
97
- Reference source: the reasoning and content values of each pytest.param case, written by the maintainers
98
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/reasoning/test_qwen3_reasoning_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
99
  Companions (not included): tests/__init__.py; tests/reasoning/__init__.py; tests/reasoning/utils.py; LICENSE
100
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer Qwen/Qwen3-0.6B from the Hugging Face hub (test file lines 19 to 23, module fixture at line 26), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-0.6B fetched 2026-10-10T19:35:21Z (sha256 ed015eb68f3bd652..., robots allowed); tokenizer Qwen/Qwen3.5-397B-A17B from the Hugging Face hub (test file lines 19 to 23, module fixture at line 26), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3.5-397B-A17B fetched 2026-10-10T19:35:24Z (sha256 82d6bdfeff3c0607..., robots allowed); tokenizer Qwen/Qwen3-4B-Thinking-2507 from the Hugging Face hub (test file lines 19 to 23, module fixture at line 26), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-4B-Thinking-2507 fetched 2026-10-10T19:35:28Z (sha256 e6a702deb94119fa..., robots allowed)
 
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 10 of 10 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Tools are named in the check written per file in the notes.
12
 
13
  **Licence (SPDX):** Apache-2.0
 
19
  - `tests/tool_parsers/test_minimax_m2_tool_parser.py`: vLLM MiniMax-M2 tool-call parser: 21 tests feed raw MiniMax-M2 output text (whole and streamed in deltas) and state the exact tool calls, arguments and content the parser must extract.
20
  Format: pytest Python module holding model output strings, tool schemas and the expected parsed tool calls inline, with an in-file fake tokenizer
21
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
22
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_minimax_m2_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. The file defines its own FakeTokenizer, so nothing is downloaded. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
23
  Companions (not included): LICENSE
24
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled
25
  - `tests/tool_parsers/test_hermes_tool_parser.py`: vLLM Hermes and Granite4 tool-call parsers: 13 tests, each run for both parsers, feed Hermes-format <tool_call> output (whole and streamed) and state the exact tool calls and content to extract.
26
  Format: pytest Python module holding model output strings and expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
27
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
28
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_hermes_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. One test is skipped upstream for the granite4 parser (line 361, the parser enforces a complete response). Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
29
  Companions (not included): LICENSE
30
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer Qwen/Qwen3-32B from the Hugging Face hub (test file line 28), licence apache-2.0, not gated; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-32B fetched 2026-10-10T19:34:49Z (sha256 5723ebdcbf86aa74..., robots allowed)
31
  - `tests/tool_parsers/test_qwen3coder_tool_parser.py`: vLLM Qwen3-Coder XML tool-call parser: 23 tests, run for the chat-completion and responses API paths, feed <function=...><parameter=...> output and state the exact typed arguments and content to extract.
32
  Format: pytest Python module holding model output strings, tool schemas and the expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
33
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
34
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_qwen3coder_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
35
  Companions (not included): LICENSE
36
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 from the Hugging Face hub (test file line 31), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 fetched 2026-10-10T19:35:07Z (sha256 c7e272aa634ac64b..., robots allowed)
37
  - `tests/tool_parsers/test_jamba_tool_parser.py`: vLLM Jamba tool-call parser: 4 parametrized tests feed Jamba <tool_calls> JSON output (whole and streamed token by token) and state the exact tool calls and content to extract.
38
  Format: pytest Python module holding model output strings and expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
39
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
40
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_jamba_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
41
  Companions (not included): LICENSE
42
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer ai21labs/Jamba-tiny-dev from the Hugging Face hub (test file line 16), licence apache-2.0, not gated; model card metadata https://huggingface.co/api/models/ai21labs/Jamba-tiny-dev fetched 2026-10-10T19:34:53Z (sha256 c4a4b391c83b9e8c..., robots allowed)
43
  - `tests/tool_parsers/test_llama3_json_tool_parser.py`: vLLM Llama 3 JSON tool-call parser: 15 tests feed Llama 3.x JSON tool-call output (single, multiple, with surrounding text, malformed) and state the exact tool calls and content to extract.
44
  Format: pytest Python module holding model output strings and expected parsed tool calls inline; tokenizer fetched from the Hugging Face hub at run time
45
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
46
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_llama3_json_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
47
  Companions (not included): LICENSE
48
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer meta-llama/Llama-3.2-1B-Instruct from the Hugging Face hub (test file line 12), licence llama3.2 (Llama 3.2 Community License), GATED: the hub asks each user to accept the licence before download (gated: manual); the test needs an HF token of an account that accepted it; model card metadata https://huggingface.co/api/models/meta-llama/Llama-3.2-1B-Instruct fetched 2026-10-10T19:35:03Z (sha256 4c82059d48b4413c..., robots allowed)
49
  - `tests/tool_parsers/test_deepseekv32_tool_parser.py`: vLLM DeepSeek-V3.2 DSML tool-call parser: 48 tests feed DSML-delimited output (whole and streamed, delimiters split across deltas) and state the exact tool calls, typed arguments and content to extract.
50
  Format: pytest Python module holding model output strings and expected parsed tool calls inline, with a MagicMock tokenizer and the shared streaming helper module
51
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
52
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_deepseekv32_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. The tokenizer is a MagicMock (MOCK_TOKENIZER, file line 32), so nothing is downloaded. All comparisons are exact equality except two float arguments compared with pytest.approx(3.14) at lines 435 and 1126, which use the pytest default tolerance. The file sets pytestmark = pytest.mark.skip_global_cleanup (line 24), a marker vLLM registers in pyproject.toml line 123; without that registration pytest only warns.
53
  Companions (not included): tests/__init__.py; tests/tool_parsers/__init__.py; tests/tool_parsers/utils.py; LICENSE
54
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled
55
  - `tests/tool_parsers/test_pythonic_tool_parser.py`: vLLM pythonic tool-call parser: 4 tests over 14 parametrized cases feed Python-call-list output ([f(a=1), g(b="x")]) whole and streamed and state the exact tool calls to extract.
56
  Format: pytest Python module holding model output strings and expected parsed tool calls inline, with the shared helper module and the tokenizer fixture
57
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
58
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_pythonic_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
59
  Companions (not included): tests/__init__.py; tests/tool_parsers/__init__.py; tests/tool_parsers/utils.py; tests/tool_parsers/conftest.py; LICENSE
60
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer gpt2 (openai-community/gpt2) from the Hugging Face hub via tests/tool_parsers/conftest.py, licence mit, not gated; model card metadata https://huggingface.co/api/models/openai-community/gpt2 fetched 2026-10-10T19:35:31Z (sha256 ef3a19a264185920..., robots allowed)
61
  - `tests/tool_parsers/test_kimi_k2_tool_parser.py`: vLLM Kimi-K2 tool-call parser: 26 tests feed Kimi-K2 section-token output (whole, streamed and at varying stream intervals) and state the exact tool-call ids, names, arguments and content to extract.
62
  Format: pytest Python module holding model output strings and expected parsed tool calls inline, with the shared helper module; tokenizer fetched from the Hugging Face hub at run time
63
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
64
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/tool_parsers/test_kimi_k2_tool_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
65
  Companions (not included): tests/__init__.py; tests/tool_parsers/__init__.py; tests/tool_parsers/utils.py; LICENSE
66
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer moonshotai/Kimi-K2-Instruct from the Hugging Face hub, loaded with trust_remote_code=True (test file line 25), so the test executes the tokenizer code published in that model repository; licence modified-mit (cardData license other, license_name modified-mit), not gated; model card metadata https://huggingface.co/api/models/moonshotai/Kimi-K2-Instruct fetched 2026-10-10T19:34:56Z (sha256 045f36fa880a85ed..., robots allowed)
67
  - `tests/reasoning/test_deepseekr1_reasoning_parser.py`: vLLM DeepSeek-R1 reasoning parser: 24 parametrized cases feed <think>...</think> output (whole and streamed, with and without the start tag, truncated) and state the exact reasoning and content strings to split out.
68
  Format: pytest Python module holding model output strings and the expected reasoning and content split inline, with the shared reasoning helper module
69
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
70
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/reasoning/test_deepseekr1_reasoning_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
71
  Companions (not included): tests/__init__.py; tests/reasoning/__init__.py; tests/reasoning/utils.py; LICENSE
72
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B from the Hugging Face hub (test file line 14, loaded with transformers AutoTokenizer at line 19), licence mit, not gated; model card metadata https://huggingface.co/api/models/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B fetched 2026-10-10T19:35:17Z (sha256 6bc6a9009545d581..., robots allowed)
73
  - `tests/reasoning/test_qwen3_reasoning_parser.py`: vLLM Qwen3 reasoning parser: 28 parametrized cases, each run under three Qwen3 tokenizers, feed <think> output and state the exact reasoning and content strings to split out, whole and streamed.
74
  Format: pytest Python module holding model output strings and the expected reasoning and content split inline, with the shared reasoning helper module
75
  Origin: expectations written by the project maintainers into the test sources; tool_fixture per general rule 6 (tool repository test expectations)
 
 
76
  Check: With vllm 0.31.0 installed (CPU build), lay the unit and its companions out at their upstream paths under tests/ and run pytest tests/reasoning/test_qwen3_reasoning_parser.py from the layout root. Hide the expected literals (tool names, argument JSON strings, content strings, reasoning strings, booleans); a solver writes or fixes the parser, or predicts the parsed output for a given model text, and the hidden literals are compared. Every comparison is exact string, dict, list or boolean equality (assert ==, assert is, json.loads equality); there is no numeric tolerance.
77
  Companions (not included): tests/__init__.py; tests/reasoning/__init__.py; tests/reasoning/utils.py; LICENSE
78
  Dependencies: vllm==0.31.0 (the pinned release) installed as a CPU build (vLLM documents a CPU backend built with VLLM_TARGET_DEVICE=cpu); the tests import the parsers from the installed vllm package; pytest; not bundled; tokenizer Qwen/Qwen3-0.6B from the Hugging Face hub (test file lines 19 to 23, module fixture at line 26), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-0.6B fetched 2026-10-10T19:35:21Z (sha256 ed015eb68f3bd652..., robots allowed); tokenizer Qwen/Qwen3.5-397B-A17B from the Hugging Face hub (test file lines 19 to 23, module fixture at line 26), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3.5-397B-A17B fetched 2026-10-10T19:35:24Z (sha256 82d6bdfeff3c0607..., robots allowed); tokenizer Qwen/Qwen3-4B-Thinking-2507 from the Hugging Face hub (test file lines 19 to 23, module fixture at line 26), licence apache-2.0, not gated; only tokenizer files are read; model card metadata https://huggingface.co/api/models/Qwen/Qwen3-4B-Thinking-2507 fetched 2026-10-10T19:35:28Z (sha256 e6a702deb94119fa..., robots allowed)
campaign_files/2026-10-ml-serving-configs-2/vlosing-driftdatasets/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 3 of 3 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: labelled_class_stream_with_known_drift. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,24 +19,18 @@ Files:
21
  - `artificial/movingSquares/movingSquares.data`: Moving Squares concept drift stream (Losing et al. 2016): 200000 two-dimensional samples, 4 classes, incremental drift.
22
  Format: plain text, 200000 lines of two space-separated floats; the .labels file holds 200000 lines with one integer class each (4 classes)
23
  Origin: synthetic benchmark stream generated by the dataset authors and published as a drift benchmark (Losing, Hammer and Wersing 2016); the generator is not in the repository
24
- Reference result: labelled_class_stream_with_known_drift
25
- Reference source: the companion file artificial/movingSquares/movingSquares.labels (the class of every sample, in stream order); the drift schedule is fixed by construction: incremental drift: four squares of uniform samples, one per class, move horizontally at constant speed and reverse at a boundary, with a horizon of 120 examples before old and new samples overlap (README.md lines 23 to 26)
26
  Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
27
  Companions (not included): artificial/movingSquares/movingSquares.labels; README.md
28
  Dependencies: none to read the data (space-separated numbers, one sample per line; labels one integer per line).
29
  - `artificial/chess/transientChessboard.data`: Transient Chessboard concept drift stream (Losing et al. 2016): 199999 two-dimensional samples, 8 classes, virtual drift.
30
  Format: plain text, 199999 lines of two space-separated floats; the .labels file holds 199999 lines with one integer class each (8 classes)
31
  Origin: synthetic benchmark stream generated by the dataset authors and published as a drift benchmark (Losing, Hammer and Wersing 2016); the generator is not in the repository
32
- Reference result: labelled_class_stream_with_known_drift
33
- Reference source: the companion file artificial/chess/transientChessboard.labels (the class of every sample, in stream order); the drift schedule is fixed by construction: virtual drift: chessboard squares are revealed one at a time, and after every four squares samples cover the whole board again (README.md lines 28 to 31)
34
  Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
35
  Companions (not included): artificial/chess/transientChessboard.labels; README.md
36
  Dependencies: none to read the data (space-separated numbers, one sample per line; labels one integer per line).
37
  - `artificial/rbf/interchangingRBF.data`: Interchanging RBF concept drift stream (Losing et al. 2016): 199999 two-dimensional samples, 15 classes, abrupt drift.
38
  Format: plain text, 199999 lines of two space-separated floats; the .labels file holds 199999 lines with one integer class each (15 classes)
39
  Origin: synthetic benchmark stream generated by the dataset authors and published as a drift benchmark (Losing, Hammer and Wersing 2016); the generator is not in the repository
40
- Reference result: labelled_class_stream_with_known_drift
41
- Reference source: the companion file artificial/rbf/interchangingRBF.labels (the class of every sample, in stream order); the drift schedule is fixed by construction: abrupt drift: fifteen Gaussians swap positions every 3000 samples, with one more Gaussian moving each time, 67 abrupt drifts in all (README.md lines 19 to 21)
42
  Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
43
  Companions (not included): artificial/rbf/interchangingRBF.labels; README.md
44
  Dependencies: none to read the data (space-separated numbers, one sample per line; labels one integer per line).
 
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 3 of 3 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `artificial/movingSquares/movingSquares.data`: Moving Squares concept drift stream (Losing et al. 2016): 200000 two-dimensional samples, 4 classes, incremental drift.
20
  Format: plain text, 200000 lines of two space-separated floats; the .labels file holds 200000 lines with one integer class each (4 classes)
21
  Origin: synthetic benchmark stream generated by the dataset authors and published as a drift benchmark (Losing, Hammer and Wersing 2016); the generator is not in the repository
 
 
22
  Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
23
  Companions (not included): artificial/movingSquares/movingSquares.labels; README.md
24
  Dependencies: none to read the data (space-separated numbers, one sample per line; labels one integer per line).
25
  - `artificial/chess/transientChessboard.data`: Transient Chessboard concept drift stream (Losing et al. 2016): 199999 two-dimensional samples, 8 classes, virtual drift.
26
  Format: plain text, 199999 lines of two space-separated floats; the .labels file holds 199999 lines with one integer class each (8 classes)
27
  Origin: synthetic benchmark stream generated by the dataset authors and published as a drift benchmark (Losing, Hammer and Wersing 2016); the generator is not in the repository
 
 
28
  Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
29
  Companions (not included): artificial/chess/transientChessboard.labels; README.md
30
  Dependencies: none to read the data (space-separated numbers, one sample per line; labels one integer per line).
31
  - `artificial/rbf/interchangingRBF.data`: Interchanging RBF concept drift stream (Losing et al. 2016): 199999 two-dimensional samples, 15 classes, abrupt drift.
32
  Format: plain text, 199999 lines of two space-separated floats; the .labels file holds 199999 lines with one integer class each (15 classes)
33
  Origin: synthetic benchmark stream generated by the dataset authors and published as a drift benchmark (Losing, Hammer and Wersing 2016); the generator is not in the repository
 
 
34
  Check: Give the agent the samples one at a time in file order with the label revealed after each prediction (the prequential protocol of the source paper).
35
  Companions (not included): artificial/rbf/interchangingRBF.labels; README.md
36
  Dependencies: none to read the data (space-separated numbers, one sample per line; labels one integer per line).
campaign_files/2026-10-op-rule-calculations-2/uscis-aao-non-precedent-decisions/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated.
10
 
11
- **Answers:** No answer files are included. Reference result: published_adjudication_labels. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
14
 
15
  **Licence (SPDX):** LicenseRef-US-Government-Work
@@ -21,28 +19,20 @@ Files:
21
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG242026_05B2203.pdf`: AAO non-precedent decision of AUG. 24, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: an investment banker.
22
  Format: PDF, 5 pages, 665985 bytes; text layer extracts with pypdf
23
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
24
- Reference result: published_adjudication_labels
25
- Reference source: the ORDER line at the end of the decision: "The appeal is dismissed."
26
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: dismissed). Exact label match, no tolerance
27
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
28
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG312026_01B2203.pdf`: AAO non-precedent decision of AUG. 31, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: a petitioner who intends to work as a principal in the field of North American strategy and public diplomacy.
29
  Format: PDF, 7 pages, 712826 bytes; text layer extracts with pypdf
30
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
31
- Reference result: published_adjudication_labels
32
- Reference source: the ORDER line at the end of the decision: "The appeal is dismissed."
33
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: dismissed). Exact label match, no tolerance
34
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
35
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG272026_01B2203.pdf`: AAO non-precedent decision of AUG. 27, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: an information technology service delivery manager.
36
  Format: PDF, 3 pages, 578554 bytes; text layer extracts with pypdf
37
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
38
- Reference result: published_adjudication_labels
39
- Reference source: the ORDER line at the end of the decision: "The appeal is dismissed."
40
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: dismissed). Exact label match, no tolerance
41
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
42
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG252026_01B2203.pdf`: AAO non-precedent decision of AUG. 25, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: a financial professional.
43
  Format: PDF, 5 pages, 536028 bytes; text layer extracts with pypdf
44
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
45
- Reference result: published_adjudication_labels
46
- Reference source: the ORDER line at the end of the decision: "The decision is withdrawn. The matter is remanded for the entry of a new decision"
47
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: remanded). Exact label match, no tolerance
48
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
 
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated.
10
 
 
 
11
  **Tools to open and check:** extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
12
 
13
  **Licence (SPDX):** LicenseRef-US-Government-Work
 
19
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG242026_05B2203.pdf`: AAO non-precedent decision of AUG. 24, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: an investment banker.
20
  Format: PDF, 5 pages, 665985 bytes; text layer extracts with pypdf
21
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
 
 
22
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: dismissed). Exact label match, no tolerance
23
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
24
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG312026_01B2203.pdf`: AAO non-precedent decision of AUG. 31, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: a petitioner who intends to work as a principal in the field of North American strategy and public diplomacy.
25
  Format: PDF, 7 pages, 712826 bytes; text layer extracts with pypdf
26
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
 
 
27
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: dismissed). Exact label match, no tolerance
28
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
29
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG272026_01B2203.pdf`: AAO non-precedent decision of AUG. 27, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: an information technology service delivery manager.
30
  Format: PDF, 3 pages, 578554 bytes; text layer extracts with pypdf
31
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
 
 
32
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: dismissed). Exact label match, no tolerance
33
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
34
  - `sites/default/files/err/B2 - Aliens with Extraordinary Ability/Decisions_Issued_in_2026/AUG252026_01B2203.pdf`: AAO non-precedent decision of AUG. 25, 2026 on Form I-140 extraordinary ability classification (EB-1A, INA 203(b)(1)(A)); petitioner: a financial professional.
35
  Format: PDF, 5 pages, 536028 bytes; text layer extracts with pypdf
36
  Origin: published non-precedent appellate decision on an immigrant visa eligibility petition, a real adjudication with a stated order
 
 
37
  Check: give the agent the caption, the introduction up to the statement of the burden of proof, the section I. LAW, and the evidence and arguments described in II. ANALYSIS with the AAO's own findings removed; hide every sentence where the AAO concludes, agrees, withdraws or finds a criterion met or not met, the section III. CONCLUSION and the ORDER line. The agent answers one label from {sustained, dismissed, remanded}; pass when it equals the order (here: remanded). Exact label match, no tolerance
38
  Tools: extract the text with pypdf (or pdftotext) and split it at the headings I. LAW, II. ANALYSIS and III. CONCLUSION; the agent writes the order label to a JSON file
campaign_files/2026-10-s-security-detection-2/wazuh-ruleset-tests/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 7 of 7 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_rule_and_level. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order; Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order; rule 5403 uses if_fts (first time seen), so its expected answer depends on the session's earlier lines: feed the file's lines in order in one fresh wazuh-logtest session; Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order; Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order; rule 20100 uses if_fts (first time seen), so each line must go to a fresh wazuh-logtest call as runtests.py does.
14
 
15
  **Licence (SPDX):** GPL-2.0-only
@@ -21,56 +19,42 @@ Files:
21
  - `ruleset/testing/tests/web_appsec.ini`: Wazuh web_appsec.ini: 31 web access log lines in 17 sections with the expected web attack rule (31501 to 31516, 31530, 31531) and alert level.
22
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
23
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
24
- Reference result: expected_rule_and_level
25
- Reference source: web_appsec.ini: 17 sections, each with rule, alert and decoder (web-accesslog) keys and 'log N pass' lines (31 lines); rules in 0270-web_appsec_rules.xml, parents 31100 and 31108 in 0245-web_rules.xml
26
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order
27
  Companions (not included): ruleset/rules/0270-web_appsec_rules.xml; ruleset/rules/0245-web_rules.xml; ruleset/decoders/0375-web-accesslog_decoders.xml; ruleset/testing/runtests.py; LICENSE
28
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
29
  - `ruleset/testing/tests/proftpd.ini`: Wazuh proftpd.ini: 7 ProFTPD log lines with the expected rule (11204 to 11207, 11222) and alert level.
30
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
31
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
32
- Reference result: expected_rule_and_level
33
- Reference source: proftpd.ini: sections with rule, alert and decoder (proftpd) keys and 'log N pass' lines (7 lines); rules in 0175-proftpd_rules.xml under parent 11200 (decoded_as proftpd, line 10 to 13)
34
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order
35
  Companions (not included): ruleset/rules/0175-proftpd_rules.xml; ruleset/decoders/0230-proftpd_decoders.xml; ruleset/testing/runtests.py; LICENSE
36
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
37
  - `ruleset/testing/tests/pam.ini`: Wazuh pam.ini: 5 PAM log lines with the expected rule (5501 to 5504, 2502) and alert level.
38
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
39
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
40
- Reference result: expected_rule_and_level
41
- Reference source: pam.ini: 5 sections with rule, alert and decoder (pam) keys: 5503/5, 5504/5, 5501/3, 5502/3, 2502/10; rules in 0085-pam_rules.xml (5500 decoded_as pam, line 12) and 0020-syslog_rules.xml
42
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order
43
  Companions (not included): ruleset/rules/0085-pam_rules.xml; ruleset/rules/0020-syslog_rules.xml; ruleset/decoders/0205-pam_decoders.xml; ruleset/testing/runtests.py; LICENSE
44
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
45
  - `ruleset/testing/tests/sudo.ini`: Wazuh sudo.ini: 8 sudo log lines with the expected rule (5401, 5403, 5404, 5405) and alert level.
46
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
47
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
48
- Reference result: expected_rule_and_level
49
- Reference source: sudo.ini: 5 sections with rule, alert and decoder (sudo) keys and 8 'log N pass' lines; rules 5400 to 5405 in 0020-syslog_rules.xml lines 636 to 697 (5400 decoded_as sudo, 5404 if_sid 5401)
50
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order; rule 5403 uses if_fts (first time seen), so its expected answer depends on the session's earlier lines: feed the file's lines in order in one fresh wazuh-logtest session
51
  Companions (not included): ruleset/rules/0020-syslog_rules.xml; ruleset/decoders/0320-sudo_decoders.xml; ruleset/testing/runtests.py; LICENSE
52
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
53
  - `ruleset/testing/tests/dovecot.ini`: Wazuh dovecot.ini: 15 Dovecot log lines with the expected rule (9702 to 9707, 9770, 9771, or 1002 for negative lines), alert level and decoder.
54
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
55
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
56
- Reference result: expected_rule_and_level
57
- Reference source: dovecot.ini: 11 sections with rule, alert and decoder (dovecot or dovecot-info) keys; 12 'log N pass' lines and 3 'log N fail' lines (lines 40, 41 and 69, expected not to give the stated rule); rules in 0155-dovecot_rules.xml (9700 to 9771)
58
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order
59
  Companions (not included): ruleset/rules/0155-dovecot_rules.xml; ruleset/decoders/0085-dovecot_decoders.xml; ruleset/testing/runtests.py; LICENSE
60
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
61
  - `ruleset/testing/tests/apparmor.ini`: Wazuh apparmor.ini: 5 kernel audit lines from AppArmor (ALLOWED, STATUS, DENIED open, mknod and exec) with the expected rule (52001 to 52004), alert level and decoder.
62
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
63
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
64
- Reference result: expected_rule_and_level
65
- Reference source: apparmor.ini: 5 sections with rule, alert and decoder (kernel) keys, one 'log 1 pass' line each; rules in 0315-apparmor_rules.xml (52000 decoded_as kernel at lines 12 to 13, children 52001 to 52004 at lines 18 to 39)
66
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order
67
  Companions (not included): ruleset/rules/0315-apparmor_rules.xml; ruleset/decoders/0140-kernel_decoders.xml; ruleset/testing/runtests.py; LICENSE
68
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
69
  - `ruleset/testing/tests/cisco_ios.ini`: Wazuh cisco_ios.ini: 17 Cisco IOS syslog lines (IPS signatures, access-list denies, SSH and 802.1X messages) with the expected rule (20100, 4700, 4713, 4715, 4716), alert level and decoder.
70
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
71
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
72
- Reference result: expected_rule_and_level
73
- Reference source: cisco_ios.ini: 6 sections with rule, alert and decoder (cisco-ios) keys, 17 'log N pass' lines; rules in 0075-cisco-ios_rules.xml (4700 decoded_as cisco-ios, lines 11 to 12, children to 4724) and 0240-ids_rules.xml (20100 first-time IDS alert, lines 13 to 18)
74
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order; rule 20100 uses if_fts (first time seen), so each line must go to a fresh wazuh-logtest call as runtests.py does
75
  Companions (not included): ruleset/rules/0075-cisco-ios_rules.xml; ruleset/rules/0240-ids_rules.xml; ruleset/decoders/0065-cisco-ios_decoders.xml; ruleset/testing/runtests.py; LICENSE
76
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 7 of 7 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order; Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order; rule 5403 uses if_fts (first time seen), so its expected answer depends on the session's earlier lines: feed the file's lines in order in one fresh wazuh-logtest session; Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order; Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order; rule 20100 uses if_fts (first time seen), so each line must go to a fresh wazuh-logtest call as runtests.py does.
12
 
13
  **Licence (SPDX):** GPL-2.0-only
 
19
  - `ruleset/testing/tests/web_appsec.ini`: Wazuh web_appsec.ini: 31 web access log lines in 17 sections with the expected web attack rule (31501 to 31516, 31530, 31531) and alert level.
20
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
21
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
 
 
22
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order
23
  Companions (not included): ruleset/rules/0270-web_appsec_rules.xml; ruleset/rules/0245-web_rules.xml; ruleset/decoders/0375-web-accesslog_decoders.xml; ruleset/testing/runtests.py; LICENSE
24
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
25
  - `ruleset/testing/tests/proftpd.ini`: Wazuh proftpd.ini: 7 ProFTPD log lines with the expected rule (11204 to 11207, 11222) and alert level.
26
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
27
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
 
 
28
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order
29
  Companions (not included): ruleset/rules/0175-proftpd_rules.xml; ruleset/decoders/0230-proftpd_decoders.xml; ruleset/testing/runtests.py; LICENSE
30
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
31
  - `ruleset/testing/tests/pam.ini`: Wazuh pam.ini: 5 PAM log lines with the expected rule (5501 to 5504, 2502) and alert level.
32
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
33
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
 
 
34
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order
35
  Companions (not included): ruleset/rules/0085-pam_rules.xml; ruleset/rules/0020-syslog_rules.xml; ruleset/decoders/0205-pam_decoders.xml; ruleset/testing/runtests.py; LICENSE
36
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
37
  - `ruleset/testing/tests/sudo.ini`: Wazuh sudo.ini: 8 sudo log lines with the expected rule (5401, 5403, 5404, 5405) and alert level.
38
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
39
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
 
 
40
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160) and a 'log N pass' line passes when the exit code is 0 (line 174); log lines are syslog or access-log text as written, fed one at a time in file order; rule 5403 uses if_fts (first time seen), so its expected answer depends on the session's earlier lines: feed the file's lines in order in one fresh wazuh-logtest session
41
  Companions (not included): ruleset/rules/0020-syslog_rules.xml; ruleset/decoders/0320-sudo_decoders.xml; ruleset/testing/runtests.py; LICENSE
42
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
43
  - `ruleset/testing/tests/dovecot.ini`: Wazuh dovecot.ini: 15 Dovecot log lines with the expected rule (9702 to 9707, 9770, 9771, or 1002 for negative lines), alert level and decoder.
44
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
45
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
 
 
46
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order
47
  Companions (not included): ruleset/rules/0155-dovecot_rules.xml; ruleset/decoders/0085-dovecot_decoders.xml; ruleset/testing/runtests.py; LICENSE
48
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
49
  - `ruleset/testing/tests/apparmor.ini`: Wazuh apparmor.ini: 5 kernel audit lines from AppArmor (ALLOWED, STATUS, DENIED open, mknod and exec) with the expected rule (52001 to 52004), alert level and decoder.
50
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
51
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
 
 
52
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order
53
  Companions (not included): ruleset/rules/0315-apparmor_rules.xml; ruleset/decoders/0140-kernel_decoders.xml; ruleset/testing/runtests.py; LICENSE
54
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
55
  - `ruleset/testing/tests/cisco_ios.ini`: Wazuh cisco_ios.ini: 17 Cisco IOS syslog lines (IPS signatures, access-list denies, SSH and 802.1X messages) with the expected rule (20100, 4700, 4713, 4715, 4716), alert level and decoder.
56
  Format: INI test file (sections of log lines with expected rule, alert level and decoder) + XML rules + XML decoders + runner
57
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test
 
 
58
  Check: Task: for each log line give the decoder, rule id and alert level (or write the rule). Score: exact match with the section's rule, alert and decoder keys, as runtests.py does: it calls wazuh-logtest -U rule:alert:decoder (runtests.py lines 157 to 160); a 'log N pass' line passes when the exit code is 0 and a 'log N fail' line passes when it is not (line 174, negate set at lines 216 to 219); log lines are fed one at a time, one wazuh-logtest call each, in file order; rule 20100 uses if_fts (first time seen), so each line must go to a fresh wazuh-logtest call as runtests.py does
59
  Companions (not included): ruleset/rules/0075-cisco-ios_rules.xml; ruleset/rules/0240-ids_rules.xml; ruleset/decoders/0065-cisco-ios_decoders.xml; ruleset/testing/runtests.py; LICENSE
60
  Dependencies: wazuh-manager v4.12.0 (wazuh-logtest, called by ruleset/testing/runtests.py): needed only to rerun the tests; pinned to the v4.12.0 tag; not packed
campaign_files/2026-10-s-security-detection/velociraptor-sigma-golden-test/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_matched_events. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Run the VQL in the golden test harness (or apply the Sigma rule by any engine); the matched event set must equal the golden output.
14
 
15
  **Licence (SPDX):** AGPL-3.0-only
@@ -21,7 +19,5 @@ Files:
21
  - `artifacts/testdata/server/testcases/sigma.in.yaml`: Velociraptor golden test: Sigma rule whoami-as-SYSTEM over a Sysmon EVTX with the expected matched event.
22
  Format: VQL query + golden output YAML + Sigma rule + EVTX (69,632 B)
23
  Origin: attack telemetry recorded in a test lab and shipped as the detection project's test data (was lab_generated)
24
- Reference result: expected_matched_events
25
- Reference source: sigma.out.yaml: the full matched Sysmon event (EventRecordID 196375) and _Match details
26
  Check: Run the VQL in the golden test harness (or apply the Sigma rule by any engine); the matched event set must equal the golden output.
27
  Companions (not included): artifacts/testdata/server/testcases/sigma.out.yaml; artifacts/testdata/files/evtx/proc_creation_win_whoami_as_system.yml; artifacts/testdata/files/evtx/Microsoft-Windows-Sysmon%4Operational.evtx
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Run the VQL in the golden test harness (or apply the Sigma rule by any engine); the matched event set must equal the golden output.
12
 
13
  **Licence (SPDX):** AGPL-3.0-only
 
19
  - `artifacts/testdata/server/testcases/sigma.in.yaml`: Velociraptor golden test: Sigma rule whoami-as-SYSTEM over a Sysmon EVTX with the expected matched event.
20
  Format: VQL query + golden output YAML + Sigma rule + EVTX (69,632 B)
21
  Origin: attack telemetry recorded in a test lab and shipped as the detection project's test data (was lab_generated)
 
 
22
  Check: Run the VQL in the golden test harness (or apply the Sigma rule by any engine); the matched event set must equal the golden output.
23
  Companions (not included): artifacts/testdata/server/testcases/sigma.out.yaml; artifacts/testdata/files/evtx/proc_creation_win_whoami_as_system.yml; artifacts/testdata/files/evtx/Microsoft-Windows-Sysmon%4Operational.evtx
campaign_files/2026-10-s-security-detection/virustotal-yara-x-tests/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The unit is marked self-contained upstream.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_boolean_match. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Check: Extract (rule, data, expected) triples from the macros; run yara or yara-x (or ask the agent whether the rule matches); compare booleans. jumps.bin is plain data (header 00 00 00 00 61 61 61 61), not an executable.
14
 
15
  **Licence (SPDX):** BSD-3-Clause
@@ -21,6 +19,4 @@ Units:
21
  - `lib/src/tests/` (2 files): YARA-X rule semantics tests: inline rules and byte strings with expected match or no match.
22
  Format: Rust test file with 780 assertion macros (rule_true 86, rule_false 22, pattern_true 100, pattern_false 167, pattern_match 405 with expected matched bytes) + 1,664 B data file
23
  Origin: test cases written by hand for the detection project's test suite (was constructed)
24
- Reference result: expected_boolean_match
25
- Reference source: macro name: rule_true!/pattern_true! must match, rule_false!/pattern_false! must not, pattern_match! also gives the expected matched bytes
26
  Check: Extract (rule, data, expected) triples from the macros; run yara or yara-x (or ask the agent whether the rule matches); compare booleans. jumps.bin is plain data (header 00 00 00 00 61 61 61 61), not an executable.
 
8
 
9
  **Self-contained:** The unit is marked self-contained upstream.
10
 
 
 
11
  **Tools to open and check:** Check: Extract (rule, data, expected) triples from the macros; run yara or yara-x (or ask the agent whether the rule matches); compare booleans. jumps.bin is plain data (header 00 00 00 00 61 61 61 61), not an executable.
12
 
13
  **Licence (SPDX):** BSD-3-Clause
 
19
  - `lib/src/tests/` (2 files): YARA-X rule semantics tests: inline rules and byte strings with expected match or no match.
20
  Format: Rust test file with 780 assertion macros (rule_true 86, rule_false 22, pattern_true 100, pattern_false 167, pattern_match 405 with expected matched bytes) + 1,664 B data file
21
  Origin: test cases written by hand for the detection project's test suite (was constructed)
 
 
22
  Check: Extract (rule, data, expected) triples from the macros; run yara or yara-x (or ask the agent whether the rule matches); compare booleans. jumps.bin is plain data (header 00 00 00 00 61 61 61 61), not an executable.
campaign_files/2026-10-s-security-detection/wazuh-ruleset-logtest/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_rule_and_level. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Feed each log line to wazuh-logtest (or a rule an agent writes); the matched rule id, level and decoder must equal the values in the section.
14
 
15
  **Licence (SPDX):** GPL-2.0-only
@@ -21,7 +19,5 @@ Files:
21
  - `ruleset/testing/tests/sshd.ini`: Wazuh ruleset logtest cases for sshd (33 cases: log line, expected rule id, level, decoder).
22
  Format: INI test file + XML rules + XML decoders
23
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test (was mixed_real_and_constructed)
24
- Reference result: expected_rule_and_level
25
- Reference source: each [section] gives log N pass = <line>, rule = <id>, alert = <level>, decoder = <name>
26
  Check: Feed each log line to wazuh-logtest (or a rule an agent writes); the matched rule id, level and decoder must equal the values in the section.
27
  Companions (not included): ruleset/rules/0095-sshd_rules.xml; ruleset/decoders/0310-ssh_decoders.xml
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Feed each log line to wazuh-logtest (or a rule an agent writes); the matched rule id, level and decoder must equal the values in the section.
12
 
13
  **Licence (SPDX):** GPL-2.0-only
 
19
  - `ruleset/testing/tests/sshd.ini`: Wazuh ruleset logtest cases for sshd (33 cases: log line, expected rule id, level, decoder).
20
  Format: INI test file + XML rules + XML decoders
21
  Origin: test log lines in the detection project's test suite, part real log samples and part written for the test (was mixed_real_and_constructed)
 
 
22
  Check: Feed each log line to wazuh-logtest (or a rule an agent writes); the matched rule id, level and decoder must equal the values in the section.
23
  Companions (not included): ruleset/rules/0095-sshd_rules.xml; ruleset/decoders/0310-ssh_decoders.xml