PiotrSty commited on
Commit
fba1cd8
·
verified ·
1 Parent(s): de358f4

Register preserved print engine and explicit contemporary-document evidence boundary

Browse files
MANIFEST.jsonl CHANGED
@@ -1,4 +1,4 @@
1
- {"bytes": 1486, "path": "README.md", "sha256": "5b2925878ddac090ed6bd85d036e321e8db5a291861165730349c7598b438853"}
2
  {"bytes": 717, "path": "REGISTRY.json", "sha256": "1c2f1fb668880a3ccf25458b5eb291558d1c7bbeff434e2ef17c43571836a160"}
3
  {"bytes": 8216, "path": "SOURCE-MANIFEST.jsonl", "sha256": "a1616e46b4efe06e418717c8ef7efe39d868e3124caac1af699fe0c1328ffc25"}
4
  {"bytes": 118289709, "path": "archives/slayer-rfdetr-layout-v1-model-20260929.zip", "sha256": "68578a6e008788fd41b6b9f64d31b3a9def82f25c615ead8fc97da5a9962fb0b"}
@@ -16,6 +16,8 @@
16
  {"bytes": 1475, "path": "models/historical-recognizer-v2-model-9c992c67/tokenizer_config.json", "sha256": "520a55b2ad2110f95a6e00df66bb5bc85582304a8fa8fa34abb6f7a8275a4fa2"}
17
  {"bytes": 4484, "path": "models/historical-recognizer-v2-model-9c992c67/trainer_state.json", "sha256": "514ff68352f14c88579db463060524c1146a6bd8a279edd74d68f1e8cdf16553"}
18
  {"bytes": 798293, "path": "models/historical-recognizer-v2-model-9c992c67/vocab.json", "sha256": "ed19656ea1707df69134c4af35c8ceda2cc9860bf2c3495026153a133670ab5e"}
 
 
19
  {"bytes": 1131, "path": "models/slayer-ocr-qwen3vl-hist-0p5-v1/README.md", "sha256": "74f2c5d7d74cdaf2aa6e025d24a3671d0ee6577a380e7ce6b08e5acfe117d356"}
20
  {"bytes": 2377, "path": "models/slayer-ocr-qwen3vl-hist-0p5-v1/release.json", "sha256": "a086cabc0419a06e18b02b5ddd74f74f901437e9e341bef0d351ea3204bd21af"}
21
  {"bytes": 127594673, "path": "models/slayer-rfdetr-layout-v1-model-20260929/checkpoint_best_total.pth", "sha256": "eda8e385179fa9df51baafcfd1accb3591d83ba88a73c620ba3436bcdf13b93c"}
 
1
+ {"bytes": 2601, "path": "README.md", "sha256": "dc3b4aaf11ac8230e1d53fd16e79ca95660b8bc057c42ca1fcc1e53b8583150a"}
2
  {"bytes": 717, "path": "REGISTRY.json", "sha256": "1c2f1fb668880a3ccf25458b5eb291558d1c7bbeff434e2ef17c43571836a160"}
3
  {"bytes": 8216, "path": "SOURCE-MANIFEST.jsonl", "sha256": "a1616e46b4efe06e418717c8ef7efe39d868e3124caac1af699fe0c1328ffc25"}
4
  {"bytes": 118289709, "path": "archives/slayer-rfdetr-layout-v1-model-20260929.zip", "sha256": "68578a6e008788fd41b6b9f64d31b3a9def82f25c615ead8fc97da5a9962fb0b"}
 
16
  {"bytes": 1475, "path": "models/historical-recognizer-v2-model-9c992c67/tokenizer_config.json", "sha256": "520a55b2ad2110f95a6e00df66bb5bc85582304a8fa8fa34abb6f7a8275a4fa2"}
17
  {"bytes": 4484, "path": "models/historical-recognizer-v2-model-9c992c67/trainer_state.json", "sha256": "514ff68352f14c88579db463060524c1146a6bd8a279edd74d68f1e8cdf16553"}
18
  {"bytes": 798293, "path": "models/historical-recognizer-v2-model-9c992c67/vocab.json", "sha256": "ed19656ea1707df69134c4af35c8ceda2cc9860bf2c3495026153a133670ab5e"}
19
+ {"bytes": 2592, "path": "models/slayer-ocr-print-engine-v1/capabilities.json", "sha256": "fe439446b39f54e468c7548b78d7e49198fab77655aec416c0fcaaf4147a50f9"}
20
+ {"bytes": 1787, "path": "models/slayer-ocr-print-engine-v1/release.json", "sha256": "79e7b31476cb2afe070c2cce3d6c3c6db09566365db564dd2e4822a7374d0f72"}
21
  {"bytes": 1131, "path": "models/slayer-ocr-qwen3vl-hist-0p5-v1/README.md", "sha256": "74f2c5d7d74cdaf2aa6e025d24a3671d0ee6577a380e7ce6b08e5acfe117d356"}
22
  {"bytes": 2377, "path": "models/slayer-ocr-qwen3vl-hist-0p5-v1/release.json", "sha256": "a086cabc0419a06e18b02b5ddd74f74f901437e9e341bef0d351ea3204bd21af"}
23
  {"bytes": 127594673, "path": "models/slayer-rfdetr-layout-v1-model-20260929/checkpoint_best_total.pth", "sha256": "eda8e385179fa9df51baafcfd1accb3591d83ba88a73c620ba3436bcdf13b93c"}
MANIFEST.sha256 CHANGED
@@ -1 +1 @@
1
- c81179623b1665776f1f0562a57019d4b154d79d1d67176ba831c477789e6445 MANIFEST.jsonl
 
1
+ 23894cb5e185e623472c3bc1b82f98fcd6437a5125107f664b33cdf1a6318535 MANIFEST.jsonl
README.md CHANGED
@@ -24,3 +24,21 @@ Release: `qwen3vl-hist-0p5-v1-20261013`; [registry receipt](models/slayer-ocr-qw
24
 
25
  This registry remains code-free. Runnable frozen code and original adapter weights are in the dedicated model snapshot above. Existing model artifacts remain unchanged.
26
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
  This registry remains code-free. Runnable frozen code and original adapter weights are in the dedicated model snapshot above. Existing model artifacts remain unchanged.
26
 
27
+ ## Preserved Polish print engine — v1, 2026-10-16
28
+
29
+ [SLAYER OCR print engine, frozen full snapshot](https://huggingface.co/PiotrSty/slayer-ocr-print-engine-frozen/tree/34d78c3c6ebcc42f0f9294e8f4d4b0d1a67491ff):
30
+ OpenCV + TrOCR mixed-v3, full unchanged weights, tokenizer/processor, runnable
31
+ engine code and exact configuration. Tag: `print-engine-v1-20261016`;
32
+ [registry receipt](models/slayer-ocr-print-engine-v1/release.json).
33
+
34
+ **Współczesny druk PL: potwierdzono wykonanie pełnego pipeline'u na jednej
35
+ syntetycznej stronie, nie jakość na reprezentatywnym zbiorze rzeczywistych
36
+ współczesnych dokumentów.** CPU smoke: 5/5 lines, 6 edits / 168 characters,
37
+ CER 3.5714%, WER 24.00%, no LLM correction. The older 75-line 5.3307% CER
38
+ control is historical ordinary print, not contemporary fullpage evidence.
39
+ No production-readiness or SOTA claim; no existing model was overwritten.
40
+
41
+ This is separate from Qwen/EZH historical experiments. This registry entry
42
+ contains only metadata; all runnable code and weights are preserved in the
43
+ dedicated snapshot, with raw inputs/outputs, limitations and checksum proof.
44
+
models/slayer-ocr-print-engine-v1/capabilities.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema": "slayer-print-engine-capabilities-v1",
3
+ "date": "2026-10-16",
4
+ "explicit_polish_status": "Silnik wykonuje OCR współczesnego, wyraźnego polskiego druku; potwierdzenie ogranicza się do jawnego syntetycznego smoke. Jakość na reprezentatywnym zbiorze rzeczywistych współczesnych dokumentów nie została potwierdzona.",
5
+ "engine": "OpenCV page detection + pinned TrOCR mixed-v3",
6
+ "source_model": "PiotrSty/trocr-pl-mixed-v3",
7
+ "source_model_revision": "85d0c91c26f8e088849096dded7c9ba10b4cd9c9",
8
+ "full_weights_preserved": true,
9
+ "modern_synthetic_fullpage_smoke": {
10
+ "schema": "slayer-modern-print-synthetic-smoke-v1",
11
+ "kind": "execution-and-accuracy-smoke-not-representative-benchmark",
12
+ "pages": 1,
13
+ "expected_lines": 5,
14
+ "detected_lines": 5,
15
+ "reference_characters": 168,
16
+ "character_errors": 6,
17
+ "substitutions": 5,
18
+ "deletions": 1,
19
+ "insertions": 0,
20
+ "cer_micro": 0.03571428571428571,
21
+ "wer_micro": 0.24,
22
+ "runtime_seconds": 43.437000000005355,
23
+ "model_revision": "85d0c91c26f8e088849096dded7c9ba10b4cd9c9",
24
+ "full_page_detection_and_recognition": true,
25
+ "external_text_correction": false,
26
+ "network_disabled_for_inference": true,
27
+ "real_document_generalization_validated": false,
28
+ "claims_not_supported": [
29
+ "representative real contemporary document accuracy",
30
+ "tables/KIE accuracy",
31
+ "handwriting",
32
+ "SOTA",
33
+ "production readiness"
34
+ ]
35
+ },
36
+ "historical_ordinary_line_control": {
37
+ "scope": "75 correctly cropped lines from two historical public-domain pages; recognizer-only control, not contemporary fullpage validation",
38
+ "lines": 75,
39
+ "cer_micro": 0.053307284398548704,
40
+ "wer_micro": 0.2288,
41
+ "reference_characters": 3583,
42
+ "character_errors": 191,
43
+ "source_repository": "PiotrSty/slayer-ocr-experiment-evidence",
44
+ "source_revision": "43b749b453f563f9589ab3d6a418e72079cd3214",
45
+ "source_file": "experiments/2026-10-06/recognizer-reviewed-colab-v1-result/line-comparison.jsonl",
46
+ "source_file_sha256": "22a94533d24587fc8ef913f1925c64bbd7c3ba62460693f5970c0f8b12cbf761",
47
+ "source_dataset_manifest_sha256": "987c89227e5539b104b398ee591dda27835def465b7d82fe8c86ea0e329b8b86",
48
+ "new_inference_performed": false
49
+ },
50
+ "modern_real_fullpage_benchmark_available": false,
51
+ "historical_Qwen_results_are_not_modern_document_evidence": true,
52
+ "frozen_original_models_overwritten": false,
53
+ "new_training_performed": false,
54
+ "production_readiness_claim": false,
55
+ "sota_claim": false
56
+ }
models/slayer-ocr-print-engine-v1/release.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "release_id": "print-engine-v1-20261016",
3
+ "status": "preserved-working-print-profile-modern-synthetic-smoke-only",
4
+ "model_repository": "PiotrSty/slayer-ocr-print-engine-frozen",
5
+ "model_commit": "34d78c3c6ebcc42f0f9294e8f4d4b0d1a67491ff",
6
+ "release_tag": "print-engine-v1-20261016",
7
+ "model_url": "https://huggingface.co/PiotrSty/slayer-ocr-print-engine-frozen/tree/34d78c3c6ebcc42f0f9294e8f4d4b0d1a67491ff",
8
+ "source_recognizer": "PiotrSty/trocr-pl-mixed-v3",
9
+ "source_recognizer_revision": "85d0c91c26f8e088849096dded7c9ba10b4cd9c9",
10
+ "weights_sha256": "19437f5769b8df8be5af1fb0c5bbcc3acccae9d889f337f0c4b9e94fa80d8bb7",
11
+ "weights_bytes": 1335747032,
12
+ "engine_recipe_sha256": "775d29534ee8268a334fc3bbd98b34f83ce1f5b91540aed1b708d19f3a5434c9",
13
+ "checksums_sha256": "d651253d12bc554b139834c8b309ebf3cbde576423bcbf0bc343fba3cb39b689",
14
+ "capabilities_sha256": "fe439446b39f54e468c7548b78d7e49198fab77655aec416c0fcaaf4147a50f9",
15
+ "source_runtime_modules": 16,
16
+ "modern_fullpage_smoke": {
17
+ "synthetic": true,
18
+ "pages": 1,
19
+ "detected_lines": 5,
20
+ "expected_lines": 5,
21
+ "character_errors": 6,
22
+ "reference_characters": 168,
23
+ "cer_micro": 0.03571428571428571,
24
+ "wer_micro": 0.24
25
+ },
26
+ "representative_real_contemporary_document_benchmark": false,
27
+ "original_models_unchanged": true,
28
+ "new_training_performed": false,
29
+ "code_included_in_registry_entry": false,
30
+ "code_and_full_weights_in_dedicated_model": true,
31
+ "published_payload_verification": {
32
+ "small_files_downloaded_and_verified": 44,
33
+ "full_weights_LFS_SHA256_and_size_match": true,
34
+ "downloaded_runner_integrity_check_passed": true
35
+ },
36
+ "engine_regression_tests_passed": 43,
37
+ "production_readiness_claim": false,
38
+ "SOTA_claim": false
39
+ }