mt628754/qwen3-struct-sft

#qwen3-4b-structured-output-lora

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).

Chain-of-Thought reasoning was removed from training data, and loss is applied directly to the final structured output.

Training Configuration

• Base model: Qwen/Qwen3-4B-Instruct-2507
• Method: QLoRA (4-bit)
• Max sequence length: 1024
• Epochs: 2
• Learning rate: 1e-06
• LoRA: r=96, alpha=192

Usage

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

base = "Qwen/Qwen3-4B-Instruct-2507"
adapter = "mt628754/qwen3-struct-sft"

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

Sources & Terms (IMPORTANT)

Training data: u-10bei/structured_data_with_cot_dataset_512_v2, u-10bei/structured_data_with_cot_dataset_512_v4, u-10bei/structured_data_with_cot_dataset_512_v5

Data preprocessing: Combined the above three versions with removal of unparseable outputs, deduplication, and removal of Chain-of-Thought reasoning from assistant responses.

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.