diff --git a/README.md b/README.md index 4c782933..b56cb4dc 100644 --- a/README.md +++ b/README.md @@ -88,7 +88,7 @@ CER unless stated). | model | task | size | headline metric | |---|---|---|---| | `ara-diac-2.0` | Arabic diacritization | 2.78 GB fp32 | SadeedDiac-25 DER **2.29%** (beats GLM-5.2 2.69); WikiNews-2024 17.38 WER | -| `ara-diac-news-1.1` | Arabic news register | 2.16 GB fp32 | WikiNews multiref 8.85 WER | +| ~~`ara-diac-news-1.1`~~ | Arabic news register | — | **WITHDRAWN** — artifact lost (never shipped; WO87). `ara-diac-2.0` covers the register | | `ara-diac-small-2.1` | Arabic, fast tier | 1.42 GB fp32 | fullset DER 4.57 (teacher 2.29) | | `ara-diac-plane-large-1.0` | Arabic, int8 fast tier | 0.87 GB int8 | DER 2.74 / WER 8.61 | | `urd-diac-base-1.0` | Urdu diacritization | 1.35 GB fp16 | CER **3.57%** | diff --git a/docs/naming/README.md b/docs/naming/README.md new file mode 100644 index 00000000..caa67ac5 --- /dev/null +++ b/docs/naming/README.md @@ -0,0 +1,29 @@ +# Target documentation — WO87 evidence (2026-10-24) + +`target-documentation.json`: every index id decoded through the +reference loader with a per-language probe — the evidence for +target-accurate naming. + +## Verified target families + +| family | ids | output sample | +|---|---|---| +| Arab-vocalized (diac) | ara-diac-*, urd-diac-* | 'السَّلَامُ عَلَيْكُمْ' | +| Heb-vocalized (nikud) | heb-diac-* (except plane: load ERROR) | 'שָׁלוֹם' | +| Thai-phonemic (toned, dots) | tha-g2p-* , tha-phonemic-1.0 | 'sa˨˩.wat̚˨˩.diː˧' | +| IPA (narrow, stress) | urd-g2p-* | 'ɪst̪ɪqˈbaːl' | +| Latin-letter phonemic | fas-g2p-1.0 | 'salAm' | +| Latin romanization | khm-latn-1.0 | 'suostei' | + +## Defects found by the sweep + +1. **ara-diac-news-1.1: artifact LOST** — index/release 404; bytes + exist nowhere (bucket zip is news-1.0's, sha differs). Withdrawn in + models.yaml; restore = re-export from run-029 best as a new version. +2. **Plane family non-contract** (ara-diac-plane-1.0, + ara-diac-plane-large-1.0, heb-diac-plane-1.0/2.0): zips fail the IMF + v1 file contract ("missing required file: encoder.onnx") — pre- + standard layout. Re-export or mark legacy-incompatible. +3. **ara-diac-small-1.0-int8 output corruption**: spurious 'F' tokens + ('...عَلَيْكُمْFمْF') — quantization-era artifact; superseded by the + 2.x line anyway. diff --git a/docs/naming/target-documentation.json b/docs/naming/target-documentation.json new file mode 100644 index 00000000..074d2818 --- /dev/null +++ b/docs/naming/target-documentation.json @@ -0,0 +1,475 @@ +{ + "ara-diac-1.0": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-2.0": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-2.0-int8": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-layerdrop-1.0": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-layerdrop-1.0-int4": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-layerdrop-1.0-int8": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-news-1.0": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-news-1.1": { + "error": "" + }, + "ara-diac-plane-1.0": { + "error": "ModelFormatError('missing required file: encoder.onnx')" + }, + "ara-diac-plane-large-1.0": { + "error": "ModelFormatError('missing required file: encoder.onnx')" + }, + "ara-diac-small-1.0": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-small-1.0-int8": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْFمْF", + "chars": [ + " ", + "F", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-small-2.0": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-small-2.0-int8": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-small-2.1": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-small-2.1-fp16": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-small-2.1-int8": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "ara-diac-small-int8static-2.1": { + "probe": "السلام عليكم", + "output": "السَّلَامُ عَلَيْكُمْ", + "chars": [ + " ", + "ا", + "س", + "ع", + "ك", + "ل", + "م", + "ي", + "َ", + "ُ", + "ّ", + "ْ" + ] + }, + "fas-diac-base-1.0": { + "probe": "سلام", + "output": "سَلام", + "chars": [ + "ا", + "س", + "ل", + "م", + "َ" + ] + }, + "fas-g2p-1.0": { + "probe": "سلام", + "output": "salAm", + "chars": [ + "A", + "a", + "l", + "m", + "s" + ] + }, + "heb-diac-1.0": { + "probe": "שלום", + "output": "שָׁלוֹם", + "chars": [ + "ָ", + "ֹ", + "ׁ", + "ו", + "ל", + "ם", + "ש" + ] + }, + "heb-diac-1.1": { + "probe": "שלום", + "output": "שָׁלוֹם", + "chars": [ + "ָ", + "ֹ", + "ׁ", + "ו", + "ל", + "ם", + "ש" + ] + }, + "heb-diac-plane-1.0": { + "error": "ModelFormatError('missing required file: encoder.onnx')" + }, + "heb-diac-plane-2.0": { + "error": "ModelFormatError('missing required file: encoder.onnx')" + }, + "heb-diac-small-1.0": { + "probe": "שלום", + "output": "שָׁלוֹם", + "chars": [ + "ָ", + "ֹ", + "ׁ", + "ו", + "ל", + "ם", + "ש" + ] + }, + "khm-latn-1.0": { + "probe": "សួស្តី", + "output": "suostei", + "chars": [ + "e", + "i", + "o", + "s", + "t", + "u" + ] + }, + "tha-g2p-base-1.0": { + "probe": "สวัสดี", + "output": "sa˨˩.wat̚˨˩.diː˧", + "chars": [ + ".", + "a", + "d", + "i", + "s", + "t", + "w", + "ː", + "˧", + "˨", + "˩", + "̚" + ] + }, + "tha-g2p-small-1.0": { + "probe": "สวัสดี", + "output": "sa˨˩.wat̚˨˩.sa˨˩.diː˧", + "chars": [ + ".", + "a", + "d", + "i", + "s", + "t", + "w", + "ː", + "˧", + "˨", + "˩", + "̚" + ] + }, + "tha-g2p-small-1.0-int4": { + "probe": "สวัสดี", + "output": "sa˨˩.wat̚˨˩.diː˧", + "chars": [ + ".", + "a", + "d", + "i", + "s", + "t", + "w", + "ː", + "˧", + "˨", + "˩", + "̚" + ] + }, + "urd-diac-1.0": { + "probe": "استقبال", + "output": "اِسْتْقِبْالْ", + "chars": [ + "ا", + "ب", + "ت", + "س", + "ق", + "ل", + "ِ", + "ْ" + ] + }, + "urd-diac-base-1.0": { + "probe": "استقبال", + "output": "اِسْتْقِبْالْ", + "chars": [ + "ا", + "ب", + "ت", + "س", + "ق", + "ل", + "ِ", + "ْ" + ] + }, + "urd-g2p-1.0": { + "probe": "استقبال", + "output": "ɪst̪ɪqˈbaːl", + "chars": [ + "a", + "b", + "l", + "q", + "s", + "t", + "ɪ", + "ˈ", + "ː", + "̪" + ] + }, + "urd-g2p-base-1.0": { + "probe": "استقبال", + "output": "ɪst̪ɪqˈbaːl", + "chars": [ + "a", + "b", + "l", + "q", + "s", + "t", + "ɪ", + "ˈ", + "ː", + "̪" + ] + } +} \ No newline at end of file diff --git a/models.yaml b/models.yaml index 2cfad307..3688bd66 100644 --- a/models.yaml +++ b/models.yaml @@ -1,4 +1,15 @@ version: 1 + +# WO87 target-accurate renames (evidence: docs/naming/target-documentation.json). +# Old ids stay resolvable via index aliases at the next index tag (zero breakage). +renames: + urd-g2p-1.0: urd-ipa-1.0 # narrow IPA with stress (2026-10-24 decode evidence) + urd-g2p-base-1.0: urd-ipa-base-1.0 # narrow IPA with stress + fas-g2p-1.0: fas-phonemic-1.0 # Latin-letter phonemic (salAm) + tha-g2p-base-1.0: tha-phonemic-base-1.0 # toned phonemic, dot-separated + tha-g2p-small-1.0: tha-phonemic-small-1.0 + tha-g2p-small-1.0-int4: tha-phonemic-small-1.0-int4 + models: tha-phonemic-1.0: task: g2p @@ -722,6 +733,12 @@ models: - Arab precision: fp32 filename: ara-diac-news-1.1.zip + status: withdrawn + withdraw_note: >- + artifact lost — the release was never created and the bytes exist nowhere + (bucket sha 280f6be9 is news-1.0's; WO87 evidence 2026-10-24). Restore + via re-export from run-029-news-specialist/best as a new version if + wanted. url: https://github.com/interscript/interscript-models/releases/download/ara-diac-news-1.1/ara-diac-news-1.1.zip sha256: 2f55e2a3c9e97da3666ef236e8ba31a24399257c9cc448363cc19f96541eb3ba size: 2160091218