sonodd/qwen3-4b-structeval-dpo-v6c

Qwen3-4B StructEval qwen3-4b-structeval-dpo-v6c

This model is a fine-tuned version of sonodd/qwen3-4b-structeval-sft-v6c-yaml-xml-focus-merged 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 align its responses with preferred outputs, focusing on improving structured output quality (JSON, YAML, XML, TOML, CSV).

Training Configuration

• —Base model: sonodd/qwen3-4b-structeval-sft-v6c-yaml-xml-focus-merged
• —SFT Adapter: None (merged SFT used as base)
• —Method: DPO (Direct Preference Optimization)
• —Epochs: 1
• —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 = "sonodd/qwen3-4b-structeval-dpo-v6c"

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

Inference with Standard Code 2

For inference using the competition's standard code 2, set:

MODEL_SOURCE = "merged"
MERGED_MODEL_ID_OR_PATH = "sonodd/qwen3-4b-structeval-dpo-v6c"

Sources & License (IMPORTANT)

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