app.py

import streamlit as st
import torch
from transformers import AutoModelForCausalLM, LlamaTokenizer
from peft import PeftModel
import gc

@st.cache_resource
def load_model():
    model_name = "peterxyz/detect-llama-34b"
    
    # Load the specific tokenizer type used by the model checkpoint
    tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False)
    
    # Check if CUDA is available
    device = "cuda" if torch.cuda.is_available() else "cpu"
    st.info(f"Using device: {device}")
    
    # Clear memory
    if device == "cuda":
        torch.cuda.empty_cache()
    gc.collect()
    
    # Load model with appropriate settings based on device
    if device == "cuda":
        from transformers import BitsAndBytesConfig
        import bitsandbytes as bnb
        
        nf4_config = BitsAndBytesConfig(
            load_in_4bit=True,
            bnb_4bit_quant_type="nf4",
            bnb_4bit_use_double_quant=True,
            bnb_4bit_compute_dtype=torch.bfloat16
        )
        
        model_nf4 = AutoModelForCausalLM.from_pretrained(
            model_name,
            quantization_config=nf4_config,
            device_map="auto",
            trust_remote_code=True
        )
        model = PeftModel.from_pretrained(model_nf4, model_name)
    else:
        # For CPU, load with reduced precision but without 4-bit quantization
        model = AutoModelForCausalLM.from_pretrained(
            model_name,
            torch_dtype=torch.float32,  # Use float32 for CPU
            device_map={"": device},
            low_cpu_mem_usage=True,
            trust_remote_code=True
        )
    
    return model, tokenizer, device

def analyze_contract(contract_code, model, tokenizer, device):
    prompt = f"{contract_code}\n\nidentify vulnerability of this code given above"
    
    # Add padding token if needed
    if tokenizer.pad_token is None:
        tokenizer.pad_token = tokenizer.eos_token
    
    inputs = tokenizer(
        prompt,
        return_tensors="pt",
        padding=True,
        truncation=True,
        max_length=2048
    ).to(device)
    
    outputs = model.generate(
        **inputs,
        max_length=1024,
        temperature=0.7,
        num_return_sequences=1,
        pad_token_id=tokenizer.pad_token_id,
        eos_token_id=tokenizer.eos_token_id
    )
    
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Set page config
st.set_page_config(
    page_title="Smart Contract Vulnerability Detector",
    page_icon="🔍",
    layout="wide"
)

# Main app
st.title("🔍 Smart Contract Vulnerability Detector")
st.markdown("""
This app analyzes Solidity smart contracts for potential vulnerabilities using a fine-tuned LLaMA model.
Simply paste your smart contract code below and click 'Analyze'.
""")

# Add a loading message while initializing
if 'model_loaded' not in st.session_state:
    st.session_state.model_loaded = False

# Initialize session state for the model
if not st.session_state.model_loaded:
    try:
        with st.spinner('Loading model... This might take a few minutes...'):
            st.session_state.model, st.session_state.tokenizer, st.session_state.device = load_model()
            st.session_state.model_loaded = True
            st.success('Model loaded successfully!')
    except Exception as e:
        st.error(f"Error loading model: {str(e)}")
        st.stop()

# Create the main interface
contract_code = st.text_area(
    "Paste your Solidity contract code here:",
    height=300,
    placeholder="pragma solidity ^0.5.0;\n\ncontract YourContract {\n    // Your code here\n}"
)

col1, col2 = st.columns([1, 4])
with col1:
    analyze_button = st.button("Analyze Contract", type="primary")
with col2:
    load_sample = st.button("Load Sample Contract")

# Sample contract button
if load_sample:
    contract_code = """pragma solidity ^0.5.0;

contract ModifierEntrancy {
    mapping (address => uint) public tokenBalance;
    string constant name = "Nu Token";
    Bank bank;

    constructor() public{
        bank = new Bank();
    }

    function airDrop() hasNoBalance supportsToken public{
        tokenBalance[msg.sender] += 20;
    }

    modifier supportsToken() {
        require(keccak256(abi.encodePacked("Nu Token")) == bank.supportsToken());
        _;
    }

    modifier hasNoBalance {
        require(tokenBalance[msg.sender] == 0);
        _;
    }
}

contract Bank{
    function supportsToken() external returns(bytes32) {
        return keccak256(abi.encodePacked("Nu Token"));
    }
}"""
    st.session_state.contract_code = contract_code
    st.experimental_rerun()

# Analysis section
if analyze_button and contract_code:
    try:
        with st.spinner('Analyzing contract...'):
            analysis = analyze_contract(
                contract_code,
                st.session_state.model,
                st.session_state.tokenizer,
                st.session_state.device
            )
            
            st.subheader("Analysis Results")
            
            # Create an expandable section for the analysis
            with st.expander("View Full Analysis", expanded=True):
                st.markdown(analysis)
            
    except Exception as e:
        st.error(f"An error occurred during analysis: {str(e)}")
        st.markdown("**Debug Information:**")
        st.code(str(e))
            
elif analyze_button:
    st.warning("Please enter some contract code to analyze.")

# Add footer with information
st.markdown("---")
st.markdown("""
<div style='text-align: center'>
    <p>Built with Streamlit and Hugging Face Transformers</p>
    <p>Model: peterxyz/detect-llama-34b</p>
</div>
""", unsafe_allow_html=True)