LoRA fine-tune of Qwen/Qwen2.5-0.5B-Instruct for smart home event reasoning and action dispatch. Built for OpenHome — a fully local, open-source smart home security system running on a Raspberry Pi with no cloud dependency.

What it does

Given a sensor event (door opened, motion detected, smoke alarm, etc.) and the current state of all smart home devices, NOVA returns a structured JSON action list:

json

[
  {"action": "notify", "message": "Smoke detected in kitchen — evacuate!", "severity": "critical"},
  {"action": "all_lights", "mode": "alert"}
]

It handles:

Sensor events — panic, smoke/CO, flood, motion, door, doorbell, RFID, climate, air quality
Natural language commands — "goodnight", "movie mode", "I'm cold", "lock everything down"
Routines — away mode, lockdown, morning, party, arrive-home sequences
Constraints — refuses to write read-only sensors, won't invent device IDs

Intended use

Deployed in OpenHome via Ollama on a Raspberry Pi. The hub loads NOVA as the reasoning layer between raw sensor events and actuator commands.

Load it:

bash

# After cloning the OpenHome repo and downloading this adapter to openhome-model/lora/
ollama create openhome-nova -f pi/Modelfile

Or load directly with PEFT:

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
model = PeftModel.from_pretrained(base, "Xx-Vexento-xX/openhome-nova")
tok = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")

Training details


Base model	Qwen/Qwen2.5-0.5B-Instruct
Method	LoRA (PEFT)
LoRA rank	8
LoRA alpha	16
Target modules	q_proj, v_proj
Training examples	1718
Validation examples	191
Epochs	2
Learning rate	2e-4
Batch size	1 (gradient accumulation 4)
Max sequence length	1024 (truncation_side=left)
Hardware	CPU only — AMD Ryzen 7 5800X3D, 34 GB RAM
Training time	~9 hours
Final eval loss	0.044

Training was done entirely on CPU using transformers + peft + trl (no CUDA, no Unsloth). The truncation_side="left" setting was critical to preserve the assistant's response tokens when sequences exceeded max length.

Evaluation

Tested against 10 OpenHome scenarios immediately after training:

Scenario	Result
Panic button pressed	PASS
Smoke alarm triggered	PASS
Lights off command	PASS
Goodnight routine	PASS
Thermostat adjustment	PASS
Flood sensor wet	PASS
No-op (no action needed)	PASS
Door open at night	PASS
Movie mode	PASS
I'm home routine	PASS

10/10 valid JSON, 9/10 action-correct (90%). The one partial-pass was smoke/CO outputting "alert" instead of "notify" as the action type — handled in the OpenHome receiver with an alias rather than requiring retraining.

Training data

The dataset covers 18 categories. Key examples:

All 13 device types with correct sensor → action mappings
Time-of-day logic (door open at night = alert, daytime = log)
Light commands: 200+ variations (color, brightness, mode, all-lights)
Thermostat commands: 200+ (setpoint, mode, hold, schedule)
Household routines: goodnight, morning, movie, party, lockdown, away, arrive
100 constraint examples — teaching what the model must NOT do (can't unlock RFID, can't clear smoke alarm, can't write read-only sensors)
50 no-op examples — so it returns [] instead of inventing actions for benign events

Regenerate or expand: python3 dataset/generate_training_data.py in the OpenHome repo.

Action schema

NOVA outputs arrays of these action types:

json

{"action": "control_device", "device_id": "plug_kitchen", "state": "on|off|toggle"}
{"action": "set_thermostat", "device_id": "thermostat_main", "setpoint": 72, "mode": "heat|cool|auto|off"}
{"action": "set_light_mode", "device_id": "rgb_living", "mode": "alert|rainbow|pulse|fire|solid", "color": "#FF0000"}
{"action": "notify", "message": "...", "severity": "low|medium|high|critical"}
{"action": "all_lights", "mode": "alert|off|on"}
{"action": "log", "message": "..."}

Limitations

Trained for the OpenHome device schema. Device type names and field names must match the OpenHome standard for best results.
At 0.5B parameters, complex multi-step reasoning can occasionally miss edge cases — the RAG memory layer in OpenHome compensates with household context.
Does not handle arbitrary home automation platforms (Home Assistant, SmartThings, etc.) out of the box.

Repository

Full system source, ESP32 firmware, Pi hub, dashboard, and training pipeline: github.com/DaEpickid540/OpenHome

License

MIT