Modelpublic

DDDDD-433/orbitdesk-qwen3-support-lora

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face

0likes5downloads

Model Card

OrbitDesk Qwen3 Support LoRA

Tiny PEFT LoRA adapter fine-tuned from Qwen/Qwen3-0.6B on 150 synthetic OrbitDesk Nova customer-support conversations.

This is a demo of domain-specific supervised fine-tuning, not a real support product.

Training Data

The synthetic dataset was generated inside the Hugging Face Job and covers setup, height presets, app pairing, calibration, error codes, wobble/noise troubleshooting, warranty, returns, safety, and escalation.

Sample Prompts

—My OrbitDesk shows E07. What should I do?
—How do I save a standing preset?
—The app says APP-204 when I pair the desk.

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = "Qwen/Qwen3-0.6B"
adapter = "DDDDD-433/orbitdesk-qwen3-support-lora"

tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base)
model = PeftModel.from_pretrained(model, adapter)

Public comparison demo: https://huggingface.co/spaces/DDDDD-433/orbitdesk-support-compare.