Use relative links for SECURITY + MODEL_LICENSES (work pre-public)
Browse files
README.md
CHANGED
|
@@ -6,7 +6,6 @@ tags:
|
|
| 6 |
- conference-demo
|
| 7 |
- local-ai
|
| 8 |
- intent-classification
|
| 9 |
-
- logistic-regression
|
| 10 |
---
|
| 11 |
|
| 12 |
> ⚠️ **Conference talk demo — not production weights.**
|
|
@@ -16,7 +15,8 @@ tags:
|
|
| 16 |
> deployable artefact. No security audit, no SLA, pinned to the talk's state.
|
| 17 |
>
|
| 18 |
> - Source repository: [thinktecture-labs/local-multi-model-agent-slm](https://github.com/thinktecture-labs/local-multi-model-agent-slm)
|
| 19 |
-
> - Threat model + out-of-scope: [`SECURITY.md`](
|
|
|
|
| 20 |
> - All five models in the stack: [Collection — Local Multi-Model Agent — nextera fine-tunes](https://huggingface.co/collections/thinktecture/local-multi-model-agent-nextera-fine-tunes-6a04a8ff2a40e5696f3c2f18)
|
| 21 |
|
| 22 |
---
|
|
@@ -26,7 +26,7 @@ tags:
|
|
| 26 |
| | |
|
| 27 |
|---|---|
|
| 28 |
| **Base** | scikit-learn `LogisticRegression`, multinomial, L2 penalty |
|
| 29 |
-
| **License** | Apache-2.0 (this repo) — but inputs are EmbeddingGemma vectors so the [Gemma Terms](
|
| 30 |
| **Training script** | [`training/train_intent_logreg.py`](https://github.com/thinktecture-labs/local-multi-model-agent-slm/blob/main/training/train_intent_logreg.py) |
|
| 31 |
| **Method** | LogReg on FT-EmbeddingGemma's 768-dim output vectors. Held-out 90/10 split. ~2 minutes on CPU. |
|
| 32 |
| **Training data** | Same as Gemma3-1B intent: `data/training-data/gemma3_intent_{scenario}.jsonl` (re-embedded with the FT EmbeddingGemma) |
|
|
|
|
| 6 |
- conference-demo
|
| 7 |
- local-ai
|
| 8 |
- intent-classification
|
|
|
|
| 9 |
---
|
| 10 |
|
| 11 |
> ⚠️ **Conference talk demo — not production weights.**
|
|
|
|
| 15 |
> deployable artefact. No security audit, no SLA, pinned to the talk's state.
|
| 16 |
>
|
| 17 |
> - Source repository: [thinktecture-labs/local-multi-model-agent-slm](https://github.com/thinktecture-labs/local-multi-model-agent-slm)
|
| 18 |
+
> - Threat model + out-of-scope: [`SECURITY.md`](SECURITY.md)
|
| 19 |
+
> - Licensing details: [`MODEL_LICENSES.md`](MODEL_LICENSES.md)
|
| 20 |
> - All five models in the stack: [Collection — Local Multi-Model Agent — nextera fine-tunes](https://huggingface.co/collections/thinktecture/local-multi-model-agent-nextera-fine-tunes-6a04a8ff2a40e5696f3c2f18)
|
| 21 |
|
| 22 |
---
|
|
|
|
| 26 |
| | |
|
| 27 |
|---|---|
|
| 28 |
| **Base** | scikit-learn `LogisticRegression`, multinomial, L2 penalty |
|
| 29 |
+
| **License** | Apache-2.0 (this repo) — but inputs are EmbeddingGemma vectors so the [Gemma Terms](MODEL_LICENSES.md) cover the embedding step |
|
| 30 |
| **Training script** | [`training/train_intent_logreg.py`](https://github.com/thinktecture-labs/local-multi-model-agent-slm/blob/main/training/train_intent_logreg.py) |
|
| 31 |
| **Method** | LogReg on FT-EmbeddingGemma's 768-dim output vectors. Held-out 90/10 split. ~2 minutes on CPU. |
|
| 32 |
| **Training data** | Same as Gemma3-1B intent: `data/training-data/gemma3_intent_{scenario}.jsonl` (re-embedded with the FT EmbeddingGemma) |
|