AshleyQu0311/Qwen3-4B-Structured-Conversion-LoRA-v4

Qwen3-4B-Structured-Conversion-LoRA-v4

This repository provides a LoRA adapter fine-tuned from Qwen/Qwen3-4B-Instruct-2507 using QLoRA (4-bit, Unsloth).

This repository contains LoRA adapter weights only. The base model must be loaded separately.

Training Objective

This adapter is trained to improve structured output accuracy (JSON / YAML / XML / TOML / CSV).

Loss is applied only to the final assistant output. Chain-of-Thought reasoning is masked during training.

Training Configuration

• Base model: Qwen/Qwen3-4B-Instruct-2507
• Method: QLoRA (4-bit)
• Max sequence length: 1024
• Epochs: 1
• Learning rate: 8e-06
• LoRA: r=64, alpha=128

Training Data

This model was trained on a balanced mixture of:

• https://huggingface.co/datasets/u-10bei/structured_data_with_cot_dataset_512_v5
• https://huggingface.co/datasets/daichira/structured-5k-mix-sft
• https://huggingface.co/datasets/daichira/structured-hard-sft-4k

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

base = "Qwen/Qwen3-4B-Instruct-2507"
adapter = "AshleyQu0311/Qwen3-4B-Structured-Conversion-LoRA-v4"

tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(
    base,
    torch_dtype=torch.float16,
    device_map="auto",
)
model = PeftModel.from_pretrained(model, adapter)

License and Terms

Users must comply with:

• The base model license

• The license of each dataset used (MIT)

Dataset License: MIT License. This dataset is used and distributed under the terms of the MIT License. Compliance: Users must comply with the MIT license (including copyright notice) and the base model's original terms of use.