deepkick/qwen3-4b-struct-dpo-v07-merged

qwen3-4b-structured-dpo-v07-merged

This model is a fine-tuned version of Qwen/Qwen3-4B-Instruct-2507 using Direct Preference Optimization (DPO) via the Unsloth library.

This repository contains the full-merged 16-bit weights. No adapter loading is required.

Training Objective

This model has been optimized using DPO to improve structured response stability and schema adherence based on the provided preference dataset.

Training Configuration

• Base model: Qwen/Qwen3-4B-Instruct-2507
• Method: DPO (Direct Preference Optimization)
• Epochs: 2
• Learning rate: 1e-07
• Beta: 0.1
• Max sequence length: 1024
• LoRA Config: r=8, alpha=16 (merged into base)

Usage

Since this is a merged model, you can use it directly with transformers:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "deepkick/qwen3-4b-struct-dpo-v07-merged"

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

prompt = "Your question here"
inputs = tokenizer.apply_chat_template(
    [{"role": "user", "content": prompt}],
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt"
).to("cuda")

outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0]))

Sources & License (IMPORTANT)

• Training Data: u-10bei/dpo-dataset-qwen-cot
• Dataset License: MIT License (as per dataset terms)
• Compliance: Users must follow the original base model's license terms.