CoolFace
Modelpublic

shakeleoatmeal/Fraud-call-detection-Qwen-0.5B-Lora

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face
0likes7downloads
Model Card

Fraud Call Detection - Qwen 0.5B LoRA

This is a LoRA fine-tuned model for detecting fraudulent phone calls based on dialogue transcripts.

Base Model

  • —Base: Qwen/Qwen2.5-0.5B-Instruct
  • —Method: LoRA (Low-Rank Adaptation)
  • —Task: Binary Classification (Fraud vs Legitimate)

Training Details

  • —LoRA Rank: 16
  • —LoRA Alpha: 32
  • —Target Modules: qproj, kproj, vproj, oproj
  • —Training: Instruction-tuned format

Usage

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

# Load model and tokenizer
base_model = "Qwen/Qwen2.5-0.5B-Instruct"
model = AutoModelForCausalLM.from_pretrained(
    base_model,
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(base_model)

# Load LoRA adapters
model = PeftModel.from_pretrained(model, "shakeleoatmeal/Fraud-call-detection-Qwen-0.5B-Lora")

# Prepare input
dialogue = "Your phone call dialogue here..."
instruction = f'''Analyze the following phone call dialogue and determine if it is fraudulent or legitimate.
Respond with only "1" if the call is fraudulent/scam, or "0" if it is legitimate.

Phone Call Dialogue:
{dialogue}

Classification:'''

inputs = tokenizer(instruction, return_tensors="pt").to(model.device)

# Generate prediction
outputs = model.generate(**inputs, max_new_tokens=5)
response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
print(response)  # "1" for fraud, "0" for legitimate

Model Output

  • —0: Legitimate call
  • —1: Fraudulent/Scam call

Limitations

  • —Designed for English language phone call dialogues
  • —Performance may vary with dialogue length and format
  • —Should be used as part of a broader fraud detection system

Citation

If you use this model, please cite:

@misc{fraud-call-detection-qwen,
  author = {shakeleoatmeal},
  title = {Fraud Call Detection - Qwen 0.5B LoRA},
  year = {2025},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/shakeleoatmeal/Fraud-call-detection-Qwen-0.5B-Lora}}
}