Endikavi commited on
Commit
fcf610f
·
verified ·
1 Parent(s): 074fd92

Ines-1 RC1 (private staging; release commit a421eb5): no pipeline_tag; apodex-mini provenance sentence

Browse files
CHANGELOG_CLAIMS.md CHANGED
@@ -26,3 +26,4 @@ Each is kept here with what replaced it and how it was found.
26
  | 19 | Dockerfile "not built" | Built from scratch and smoke-tested (CPU, GPU bf16: 0 of 400 answers changed vs canonical) | ENVIRONMENT.md |
27
  | 20 | `examples/quickstart.py` and `scripts/serve_jev.py` work from any copy | They failed when run from a `snapshot_download` cache (files are symlinks to blobs and `Path.resolve()` left the snapshot); fixed with `Path.absolute()`, found by the smoke test from the private Hub repo | smoke from the Hub |
28
  | 21 | Model weights licence pending review | Model weights, config and tokenizer: **Apache-2.0** (LICENSE, NOTICE), decided 2026-10-02; code MIT; training data keeps its own terms | LICENSING_NOTES.md §2 |
 
 
26
  | 19 | Dockerfile "not built" | Built from scratch and smoke-tested (CPU, GPU bf16: 0 of 400 answers changed vs canonical) | ENVIRONMENT.md |
27
  | 20 | `examples/quickstart.py` and `scripts/serve_jev.py` work from any copy | They failed when run from a `snapshot_download` cache (files are symlinks to blobs and `Path.resolve()` left the snapshot); fixed with `Path.absolute()`, found by the smoke test from the private Hub repo | smoke from the Hub |
28
  | 21 | Model weights licence pending review | Model weights, config and tokenizer: **Apache-2.0** (LICENSE, NOTICE), decided 2026-10-02; code MIT; training data keeps its own terms | LICENSING_NOTES.md §2 |
29
+ | 22 | `pipeline_tag: text-classification` | No `pipeline_tag`: on the private staging repo the Hub attached a generic text-classification widget that no inference provider can run with this custom interface; descriptive tags kept | Hub metadata of the private repo |
README.md CHANGED
@@ -3,7 +3,6 @@ license: apache-2.0
3
  language:
4
  - en
5
  - es
6
- pipeline_tag: text-classification
7
  tags:
8
  - mixture-of-experts
9
  - decision-model
 
3
  language:
4
  - en
5
  - es
 
6
  tags:
7
  - mixture-of-experts
8
  - decision-model
SHA256SUMS CHANGED
@@ -1,17 +1,17 @@
1
  97ad8982da911d928c3423f164a38f80608b98289ac4f4b3f292d90ae0ef2262 .dockerignore
2
  344fc5bfdbc9c68d218d648034f4a1651ca5bf71e513dedac9b45a8847830913 .gitattributes
3
  862263fa1f46c20f0d1e4dac5ffcc75abd55c08211b2c3864c5f8764b9d87793 .gitignore
4
- 73b5c37b6008d873ca792aadee562b7b99413d1a88fbc7a82e7d8787667fc6ae CHANGELOG_CLAIMS.md
5
  e4a1aeafc1f6c01f8fff69aef8a2518631428f3ec96db4a9a56d7bb296cce95f Dockerfile
6
  5e98ea3d3045927158b597521fcf127ac5569e656e4e665f0c94a35e2a191980 ENVIRONMENT.md
7
  cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSE
8
  6a7ddcad077839a44697c005ea7e6f455caecaa8054412913108d6e21a1c9dbe LICENSE-CODE
9
  2d6f4e21f766eedbfe2a92b3b8058493ad4e907f5f0756049ac1e2c53e3cfd01 LICENSING_NOTES.md
10
  178f429d8400d3f9df9e3b9cac9e64e3cff29cf9cd21bb135c91ede2866cbfe1 NOTICE
11
- cc5f7c9d0a38afbad2aab67708743858e20e34e7849bc98883e33fb0ad6438b1 README.md
12
  6ec8a2bb55f9331d77b9adbeae7fd71eeb5614a988cd21eee18c7bde775df5ba TASKSOURCE_LICENSE_AUDIT.md
13
  058dc5687ddb161d4dbfcf1703ae059e78698714b4ce2ddf1783262c10b95142 THIRD_PARTY_DATA_NOTICE.md
14
- 78a191f876a6927c8686577bc5b7e5ddb2dcf28b0dae13eb9e4745fe61766bb5 config.json
15
  da9bad7608ad2fd759e0e9088fd55a60a22d11ea6b8cb720040594a2b0fc556f eval/fullcase_report.json
16
  60e288cf88dfdf189fa7582daaad8a274911ce47db6cf47f3dfe0367760fa353 eval/fullcase_report.md
17
  a8e81da13d548c4dac0d26ecab081c4640429f1493eb607015bbca8008000d29 eval/items/v3-engram_off__tasksource.jsonl
@@ -73,7 +73,7 @@ d6c695d3cd37cde9907c505876c8c6f105fed46b60e94c2d908a9725d95f8ed9 requirements.t
73
  64566a481519aaa6c543672e7e6fe7096f56fac7e4f38ea501201e02ff34a915 tasksource_license_audit.csv
74
  859e5c63f13924ebed8dd9a3dd6690438fd084771728c1da3a220017a5e74469 tokenizer/tokenizer.json
75
  6bb112f88cf6761e60be73f4ea2d112aa33df6dd5daf8f5b8646e91bb7397da3 tokenizer/tokenizer_identity.json
76
- d61ff5c5e4e5ef36d5b029125d603ff914c53dc796e6cd48d78205ce254c70f3 training/TRANSLATION_PROVENANCE.md
77
  6440a6e61e4ae6761b32ef165beb5f7f3806c3df24e35e60c3d6ab659b8f070c training/code/audit_data.py
78
  90d07f574c1b27b20fcbfa4e4929d2ccd2d80e680640b333837aff49a9411d24 training/code/audit_report.py
79
  c924991b591ea6e96ce50fb5f89f92af3ad7ce5136fcfeeaad07d254b8cf1c20 training/code/decisions.py
 
1
  97ad8982da911d928c3423f164a38f80608b98289ac4f4b3f292d90ae0ef2262 .dockerignore
2
  344fc5bfdbc9c68d218d648034f4a1651ca5bf71e513dedac9b45a8847830913 .gitattributes
3
  862263fa1f46c20f0d1e4dac5ffcc75abd55c08211b2c3864c5f8764b9d87793 .gitignore
4
+ 109741891311a91409d58472eb4daa6d198e1912a8111ad5482d9991bc78c06a CHANGELOG_CLAIMS.md
5
  e4a1aeafc1f6c01f8fff69aef8a2518631428f3ec96db4a9a56d7bb296cce95f Dockerfile
6
  5e98ea3d3045927158b597521fcf127ac5569e656e4e665f0c94a35e2a191980 ENVIRONMENT.md
7
  cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSE
8
  6a7ddcad077839a44697c005ea7e6f455caecaa8054412913108d6e21a1c9dbe LICENSE-CODE
9
  2d6f4e21f766eedbfe2a92b3b8058493ad4e907f5f0756049ac1e2c53e3cfd01 LICENSING_NOTES.md
10
  178f429d8400d3f9df9e3b9cac9e64e3cff29cf9cd21bb135c91ede2866cbfe1 NOTICE
11
+ c6d5fdace486379bf2242d2c98589f946b6cd391057ce240c77304a30742e918 README.md
12
  6ec8a2bb55f9331d77b9adbeae7fd71eeb5614a988cd21eee18c7bde775df5ba TASKSOURCE_LICENSE_AUDIT.md
13
  058dc5687ddb161d4dbfcf1703ae059e78698714b4ce2ddf1783262c10b95142 THIRD_PARTY_DATA_NOTICE.md
14
+ 6935bddba3a216046ba6570cbdbd988cd285d9ee9bf4388c7ca610d2be4a6fb6 config.json
15
  da9bad7608ad2fd759e0e9088fd55a60a22d11ea6b8cb720040594a2b0fc556f eval/fullcase_report.json
16
  60e288cf88dfdf189fa7582daaad8a274911ce47db6cf47f3dfe0367760fa353 eval/fullcase_report.md
17
  a8e81da13d548c4dac0d26ecab081c4640429f1493eb607015bbca8008000d29 eval/items/v3-engram_off__tasksource.jsonl
 
73
  64566a481519aaa6c543672e7e6fe7096f56fac7e4f38ea501201e02ff34a915 tasksource_license_audit.csv
74
  859e5c63f13924ebed8dd9a3dd6690438fd084771728c1da3a220017a5e74469 tokenizer/tokenizer.json
75
  6bb112f88cf6761e60be73f4ea2d112aa33df6dd5daf8f5b8646e91bb7397da3 tokenizer/tokenizer_identity.json
76
+ dc8e89f55b7df1201cc58e3c8f1189951ed3b97f45a89cbb2e0b24ff66c99c46 training/TRANSLATION_PROVENANCE.md
77
  6440a6e61e4ae6761b32ef165beb5f7f3806c3df24e35e60c3d6ab659b8f070c training/code/audit_data.py
78
  90d07f574c1b27b20fcbfa4e4929d2ccd2d80e680640b333837aff49a9411d24 training/code/audit_report.py
79
  c924991b591ea6e96ce50fb5f89f92af3ad7ce5136fcfeeaad07d254b8cf1c20 training/code/decisions.py
config.json CHANGED
@@ -95,7 +95,7 @@
95
  "engram_backend_identity": "token_id:65536:65536:197f7a314b356f70",
96
  "training_code_commit": "0b5e63e3da05c1a51eb9f0ea1758f8d53d200a80-dirty",
97
  "bundle_model_code_commit": "5d1ad97",
98
- "bundle_release_code_commit": "239b54f",
99
  "torch_version_at_training": "2.14.0+cu130"
100
  }
101
  }
 
95
  "engram_backend_identity": "token_id:65536:65536:197f7a314b356f70",
96
  "training_code_commit": "0b5e63e3da05c1a51eb9f0ea1758f8d53d200a80-dirty",
97
  "bundle_model_code_commit": "5d1ad97",
98
+ "bundle_release_code_commit": "a421eb5",
99
  "torch_version_at_training": "2.14.0+cu130"
100
  }
101
  }
training/TRANSLATION_PROVENANCE.md CHANGED
@@ -6,7 +6,7 @@ so that the historical run is documented exactly; it does not claim that a rerun
6
 
7
  | | historical value |
8
  |---|---|
9
- | model name sent in the request (`--translator-model`) | `apodex-mini` |
10
  | what that name is | the served-model alias (`--served-model-name`) of a self-hosted OpenAI-compatible server (SGLang 0.5.18) serving the public model [`apodex/Apodex-1.1-mini-FP8`](https://huggingface.co/apodex/Apodex-1.1-mini-FP8) (Apache-2.0 on its card; base model `apodex/Apodex-1.1-mini`). The alias is not a public model ID; the endpoint was a private server and is not published. The model revision served on that day was not pinned (head of the model repo checked on 2026-10-02: `709359cb`). |
11
  | request | chat completions; `temperature` 0.3, `max_tokens` 4000, `chat_template_kwargs.enable_thinking` false; prompt in `prepare_public.py` (`translate_case`) |
12
  | check | a translation is kept only if keys, option ids, numbers, booleans and structure are identical to the English case (3 tries, else dropped) |
 
6
 
7
  | | historical value |
8
  |---|---|
9
+ | model name sent in the request (`--translator-model`) | `apodex-mini` — **apodex-mini was the historical served-model alias used for the public apodex/Apodex-1.1-mini-FP8 model.** |
10
  | what that name is | the served-model alias (`--served-model-name`) of a self-hosted OpenAI-compatible server (SGLang 0.5.18) serving the public model [`apodex/Apodex-1.1-mini-FP8`](https://huggingface.co/apodex/Apodex-1.1-mini-FP8) (Apache-2.0 on its card; base model `apodex/Apodex-1.1-mini`). The alias is not a public model ID; the endpoint was a private server and is not published. The model revision served on that day was not pinned (head of the model repo checked on 2026-10-02: `709359cb`). |
11
  | request | chat completions; `temperature` 0.3, `max_tokens` 4000, `chat_template_kwargs.enable_thinking` false; prompt in `prepare_public.py` (`translate_case`) |
12
  | check | a translation is kept only if keys, option ids, numbers, booleans and structure are identical to the English case (3 tries, else dropped) |