Qwen3-4B STAR QLoRA v2 Merged

This repository contains a merged model fine-tuned from Qwen/Qwen3-4B for STAR (Situation, Task, Action, Result) interview answer analysis.

The LoRA adapter was merged into the base model, so the model can be loaded directly without PEFT.

Model Details

• Base model: Qwen/Qwen3-4B
• Model type: Fine-tuned Causal Language Model
• Training method: QLoRA / LoRA
• Model format: Merged model
• Task: STAR interview answer analysis
• Pipeline: Text generation

Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "YOUR_MERGED_MODEL_REPOSITORY"

tokenizer = AutoTokenizer.from_pretrained(model_name)

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

Files

• model-*.safetensors: Merged model weights
• config.json: Model configuration
• Tokenizer configuration files
• generation_config.json: Generation configuration

Framework Versions

• Transformers
• PEFT: 0.19.1
• TRL