clemsail commited on
Commit
3759008
·
verified ·
1 Parent(s): 7d0573b

Refresh model card: license chain + DISCLOSURE bandeau v2

Browse files
Files changed (1) hide show
  1. README.md +21 -8
README.md CHANGED
@@ -50,20 +50,33 @@ print(generate(model, tokenizer, prompt="..."))
50
  | Optimizer | Adam, LR 1e-5 |
51
  | Hardware | Apple M3 Ultra 512 GB |
52
 
53
- Training data lineage: derived from the **eu-kiki / mascarade** corpus
54
- (Ailiance-fr datasets) curated for the `cpp` domain. See the
55
- [Ailiance-fr dataset catalog](https://huggingface.co/Ailiance-fr) for upstream cards.
 
 
 
 
 
 
 
 
 
 
 
 
56
 
57
  ## EU AI Act compliance
58
 
59
- - **Article 53(1)(c)**: training data licenses preserved (permissive only).
60
- - **Article 53(1)(d)**: training data summary — see source dataset cards on Ailiance-fr.
61
- - **GPAI Code of Practice (July 2025)**: base `mistralai/Devstral-Small-2-24B-Instruct-2512` released under apache-2.0; this LoRA inherits.
62
- - **No web scraping**, **no licensed data**, **no PII**.
 
63
 
64
  ## License
65
 
66
- LoRA weights: **apache-2.0** (inherited from base).
67
 
68
  ## Citation
69
 
 
50
  | Optimizer | Adam, LR 1e-5 |
51
  | Hardware | Apple M3 Ultra 512 GB |
52
 
53
+ ## Training data lineage
54
+
55
+ Derived from the internal **eu-kiki / mascarade** curation. All upstream samples
56
+ are synthetic, permissively-licensed, or generated from Apache-2.0 base resources.
57
+ See the [Ailiance-fr catalog](https://huggingface.co/Ailiance-fr) for related cards.
58
+
59
+ ## License chain
60
+
61
+ | Component | License |
62
+ |-----------------------------------|-------------------|
63
+ | Base model (`mistralai/Devstral-Small-2-24B-Instruct-2512`) | apache-2.0 |
64
+ | Training data (internal Ailiance curation (synthetic + permissive sources)) | apache-2.0 |
65
+ | **LoRA adapter (this repo)** | **apache-2.0**|
66
+
67
+ _All upstream components are Apache 2.0 / MIT — LoRA inherits permissive terms._
68
 
69
  ## EU AI Act compliance
70
 
71
+ - **Article 53(1)(c)**: training data licenses preserved (per-dataset cards declare upstream licenses).
72
+ - **Article 53(1)(d)**: training data summary — see upstream dataset cards on Ailiance-fr.
73
+ - **GPAI Code of Practice (July 2025)**: base `mistralai/Devstral-Small-2-24B-Instruct-2512` released under apache-2.0.
74
+ - **No web scraping by Ailiance**, **no licensed data**, **no PII**.
75
+ - Upstream Stack Exchange content (where applicable) is CC-BY-SA-4.0 and propagates to this adapter.
76
 
77
  ## License
78
 
79
+ LoRA weights: **apache-2.0** see License chain table above for derivation rationale.
80
 
81
  ## Citation
82