---
title: README
emoji: ⚡
colorFrom: blue
colorTo: gray
sdk: static
pinned: false
---

---
license: llama2
---

# Marvinmw/{13b,34b}_reasoner

Welcome to the repository of Marvinmw/{13b,34b}_reasoner, a custom 13-billion and 34-billion parameter model built upon the Llama 2 architecture, tailored for reasoning and code analysis, especially in the domain of smart contract audits.

## Model Description

Marvinmw/{13b,34b}_reasoner is based on the powerful Llama 2 model and has been fine-tuned with a significant dataset from Solodit and Code4rena. This includes over 10,000 findings from smart contract audits, making it uniquely suited for reasoning over complex smart contract code and security vulnerabilities.

### Features

- **Base Model**: Llama 2, known for its robust handling of language and code.
- **Fine-tuning Dataset**: Over 10,000 smart contract audit findings from platforms such as Solodit and Code4rena.
- **Use Case**: Designed primarily for developers, auditors, and researchers engaged in the security analysis of blockchain technologies and smart contracts.

## Getting Started

To use Marvinmw/{13b,34b}_reasoner, follow these steps:

### Prerequisites

- Python 3.8 or newer
- pip or conda

### Installation

Install the necessary packages using pip:

```bash
pip install -r requirements.txt
```

### Usage

You can load and use the model as follows:

```python
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("MetaTrustSig/13b_reasoner")
tokenizer = AutoTokenizer.from_pretrained("MetaTrustSig/13b_reasoner")

# Example usage
text = "Insert your smart contract code or query here"
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)
```

OR

```python
from transformers import LlamaForCausalLM, LlamaTokenizer
import torch

# Path to your model directory
model_path = "MetaTrustSig/13b_reasoner"


# Load the tokenizer
tokenizer = LlamaTokenizer.from_pretrained(model_path)

# Add special tokens if they are missing
if tokenizer.eos_token is None:
    tokenizer.add_special_tokens({
        'eos_token': '</s>',
        'bos_token': '<s>',
        'unk_token': '<unk>',
        'pad_token': '<pad>'
    })

# Load the model with the language modeling head
model = LlamaForCausalLM.from_pretrained(model_path)
model.resize_token_embeddings(len(tokenizer))

# Move model to GPU if available
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model.to(device)

text = "YOUR INPUT"

# Tokenize input and move tensors to the appropriate device
inputs = tokenizer(text, return_tensors="pt").to(device)

# Generate text
# You can change the generation config
generated_outputs = model.generate(
    input_ids=inputs['input_ids'],
    attention_mask=inputs['attention_mask'],
    max_length=1024,
    do_sample=True,
    temperature=0.2,
    top_p=0.9,
    repetition_penalty=1.1,
    eos_token_id=tokenizer.eos_token_id,
)

# Decode the output
generated_text = tokenizer.decode(generated_outputs[0], skip_special_tokens=True)

print(generated_text)

```

## Contributing

Contributions to Marvinmw/{13b,34b}_reasoner are welcome! Here's how you can contribute:

1. **Issues**: For bugs or feature requests, open an issue.
2. **Pull Requests**: Submit a PR to contribute with code changes or documentation updates.

Please see `CONTRIBUTING.md` for more details on our code of conduct and the process for submitting pull requests to us.

## License

This project is licensed under the MIT License - see the `LICENSE` file for details.

## Acknowledgments

- Thanks to the Llama 2 team for the base model.
- Solodit and Code4rena for providing the dataset for fine-tuning.

## Contact

For any further questions or partnership inquiries, please contact us via email at [info@metatrust.io].

## Additional Information

- **Model Performance Metrics**: If available, include details about the model's performance metrics and benchmarks.
- **Updates and Maintenance**: Information about how the model will be updated and maintained.

---

We hope you find Marvinmw/{13b,34b}_reasoner useful for your smart contract security needs. Enjoy using it!