Yash345/Qwen2.5-1.5B-Medical-Denoise-Adapter

base_model: Qwen/Qwen2.5-1.5B-Instruct
library_name: peft
tags:
- medical
- asr-correction
- robustness
- qlora

🏥 Edge-Native Medical Phonetic Denoiser (1.5B Adapter)

Author: Yash Sharma
Context: Developed as a reproduction of the paper "Evaluating Robustness in LLM-based Medical Chatbots" (Wadhwani AI).

🎯 Model Description

This is a QLoRA adapter fine-tuned on Qwen/Qwen2.5-1.5B-Instruct. It is designed to run on resource-constrained edge devices (T4 GPU / CPU) to correct severe ASR (Automatic Speech Recognition) phonetic errors in medical queries.

• Recovery: Improves RAG retrieval recall from 34% (Noisy) to 52% (Denoised).
• Training: Trained on 600 samples of "Brutally Noised" HealthCareMagic data.
• Compute: 4-bit Quantization (NF4).

🚀 How to Use

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

# 1. Load Base Model
base_model_id = "Qwen/Qwen2.5-1.5B-Instruct"
model = AutoModelForCausalLM.from_pretrained(
    base_model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(base_model_id)

# 2. Load This Adapter
model = PeftModel.from_pretrained(model, "YOUR_HF_USERNAME/Qwen2.5-1.5B-Medical-Denoise-Adapter")

# 3. Inference
noisy_query = "wat r da simptoms of birus"
prompt = f"User: Fix this medical query: {noisy_query}\nAssistant:"

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=64)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Output: "What are the symptoms of virus"