---
license: cc-by-sa-4.0
base_model: lmstudio-community/gemma-4-E4B-it-MLX-4bit
library_name: peft
tags:
- mlx
- lora
- peft
- electronics
- embedded
- kicad
- spice
- ailiance
- gemma
- aggro-test
language:
- en
- fr
pipeline_tag: text-generation
---

# Ailiance — Gemma 4 E4B aggro-test LoRA

LoRA adapter fine-tuned on `lmstudio-community/gemma-4-E4B-it-MLX-4bit` for the **aggro-test** domain (electronics, embedded, KiCad, SPICE).

> Maintained by **Ailiance** — French AI org publishing EU AI Act aligned LoRA adapters and datasets.

## Quick start (MLX)

```python
from mlx_lm import load, generate

model, tokenizer = load(
    "lmstudio-community/gemma-4-E4B-it-MLX-4bit",
    adapter_path="Ailiance-fr/gemma-4-E4B-aggro-test-lora",
)

print(generate(model, tokenizer, prompt="..."))
```

## Benchmark on production tasks

**Gemma aggressive-LR test run** — evaluated through the
[`electron-bench`](https://github.com/ailiance/ailiance-bench/blob/main) functional pipeline
(Phases P1 → P6, base vs LoRA).

| Task | Result |
|---|---|
| status | experimental |

Full base-vs-LoRA matrix (all phases, all adapters):
[`compare_base_vs_lora.md`](https://github.com/ailiance/ailiance-bench/blob/main/bench-results/compare_base_vs_lora.md).

## License chain

| Component                                     | License                       |
|-----------------------------------------------|-------------------------------|
| Base model weights (`lmstudio-community/gemma-4-E4B-it-MLX-4bit`)            | Gemma Terms of Use            |
| Training data ([`Ailiance-fr/kill-life-embedded-qa`](https://huggingface.co/datasets/Ailiance-fr/kill-life-embedded-qa)) | cc-by-sa-4.0                  |
| **LoRA adapter (this repo)**                  | **CC-BY-SA-4.0**              |

**Rationale**: weights of the base model inherit from the Gemma Terms of Use, but
the **LoRA adapter is a derivative of CC-BY-SA-4.0 training data** and is therefore
released under CC-BY-SA-4.0 (share-alike propagates). Downstream users who load
this adapter against the Gemma base must comply with **both** licenses
simultaneously.

## Training data lineage

Primary corpus: [`Ailiance-fr/kill-life-embedded-qa`](https://huggingface.co/datasets/Ailiance-fr/kill-life-embedded-qa) (cc-by-sa-4.0).
See the [Ailiance-fr catalog](https://huggingface.co/Ailiance-fr) for related cards.

## EU AI Act compliance

- **Article 53(1)(c)**: training data licenses preserved upstream.
- **Article 53(1)(d)**: training data summary — see dataset cards on Ailiance-fr.
- **GPAI Code of Practice (July 2025)**: base model Gemma (Google is a signatory).
- **No web scraping by Ailiance**, **no licensed data**, **no PII**.

## License

LoRA weights: **CC-BY-SA-4.0** (training-data share-alike). Base model weights remain under Gemma Terms of Use.

## Citation

```bibtex
@misc{ailiance_gemma_4_E4B_aggro_test_lora_2026,
  author    = {Ailiance},
  title     = {Ailiance — Gemma 4 E4B aggro-test LoRA},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/Ailiance-fr/gemma-4-E4B-aggro-test-lora}
}
```

## Related

See the full [Ailiance-fr LoRA collection](https://huggingface.co/Ailiance-fr).


## Bench comparison (2026-05-11)

### kicad-pcb generation (phase4/5 composite, 5 ref circuits)

| Variant | composite | parse_ok | erc_clean | erc_low_warn |
|---|---:|---:|---:|---:|
| gemma-4-E4B base | 0.060 | 0 | 0 | 0 |
| **This LoRA (tuned)** | **0.060** | 0 | 0 | 0 |
| **Delta** | 0.000 | 0 | 0 | 0 |

Note: all `parse_ok=0` — domain transfer pcb→sch fails on every variant. The
floor reflects the composite formula weights
(0.30·parse + 0.40·no-extra-errs + 0.30·no-extra-warns), not actual valid
generation.

Source: <https://github.com/ailiance/ailiance-bench/blob/main/bench-results/kicad_phase4_lora.json>