---
library_name: peft
base_model: Qwen/Qwen3-8B
license: apache-2.0
tags:
  - electronics
  - embedded-systems
  - platformio
  - lora
  - sft
  - kiki-tuning
language:
  - en
  - fr
datasets:
  - custom
pipeline_tag: text-generation
---

# KIKI PLATFORMIO SFT — LoRA Adapter

Fine-tuned LoRA adapter for **platformio** domain expertise, based on `Qwen/Qwen3-8B`.

Part of the [KIKI Models Tuning](https://github.com/L-electron-Rare/KIKI-models-tuning) pipeline
for the [FineFab](https://github.com/L-electron-Rare) platform.

## Training Details

| Parameter | Value |
|-----------|-------|
| Base Model | `Qwen/Qwen3-8B` |
| Method | QLoRA (4-bit NF4) |
| LoRA Rank | 16 |
| Epochs | 3 |
| Dataset | 7008 examples |
| Domain | platformio |

## Usage

```python
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B", device_map="auto")
model = PeftModel.from_pretrained(model, "clemsail/kiki-platformio-sft")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-8B")
```

## License

Apache 2.0