This repository provides a LoRA adapter fine-tuned from Qwen/Qwen3-4B-Instruct-2507 using QLoRA (4-bit, Unsloth). It is optimized for high-precision structured data generation on NVIDIA L4.

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, while intermediate reasoning (Chain-of-Thought) is masked.

Training Configuration

Base model: Qwen/Qwen3-4B-Instruct-2507
Method: QLoRA (4-bit)
Max sequence length: 2048
Epochs: 2
Learning rate: 2e-04
LoRA: r=16, alpha=32
Hardware: NVIDIA L4
Precision: bfloat16

Usage

python

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

base = "Qwen/Qwen3-4B-Instruct-2507"
adapter = "MasaKoma/qwen3-4b-structured-output-lora5"

tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(
    base,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
model = PeftModel.from_pretrained(model, adapter)

Sources & Terms (IMPORTANT)

Training data: u-10bei/structured_data_with_cot_dataset_512_v5

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.

MasaKoma/qwen3-4b-structured-output-lora6

Training Objective

Training Configuration

Usage

Sources & Terms (IMPORTANT)

MasaKoma/
qwen3-4b-structured-output-lora6