This repository contains a LoRA adapter fine-tuned on true/false mental health questionnaire item generation, with a focus on depression and anxiety dimensions.

⚠️ This repository only contains LoRA adapter weights. You must load it together with the base model: Qwen/Qwen2.5-3B-Instruct.

How to Use

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-3B-Instruct",
    device_map="auto",
    torch_dtype="auto",
    trust_remote_code=True
)

model = PeftModel.from_pretrained(
    base_model,
    "Joshua0522/qwen25-3b-mental-health-itemgen-lora"
)

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-3B-Instruct")

prompt = """
Generate exactly 5 true/false depression questionnaire items.
"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=200)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))