Llama-3.1-8B-Lexi-Uncensored-V2

Introduction

Llama-3.1-8B-Lexi-Uncensored-V2 is a text generation model developed by Orenguteng, based on Llama-3.1-8b-Instruct. It is designed to provide more compliant and smarter responses, with an uncensored approach. The model is suitable for various text generation tasks and is accessible through Hugging Face's platform.

Architecture

The model is built using the Transformers library and utilizes safetensors for efficient storage. It includes a system prompt feature that allows users to guide its responses. The model's architecture ensures compliance and adaptability for different use cases, although users should implement an alignment layer to prevent unethical outputs.

Training

The model has been evaluated across various datasets with different shot configurations:

IFEval (0-Shot): Achieved a strict accuracy of 77.92.
BBH (3-Shot): Recorded a normalized accuracy of 29.69.
MATH Lvl 5 (4-Shot): Scored an exact match of 16.92.
GPQA (0-Shot): Noted an acc_norm of 4.36.
MuSR (0-Shot): Achieved an acc_norm of 7.77.
MMLU-PRO (5-Shot): Obtained an accuracy of 30.9.

These results are available on the Open LLM Leaderboard.

Guide: Running Locally

To run the model locally, follow these basic steps:

Install the Transformers library:
```
pip install transformers
```

Clone the model repository:

git clone https://huggingface.co/Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2

Load the model in your script:

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2")
model = AutoModelForCausalLM.from_pretrained("Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2")

Run the model with optional GPU acceleration:
- Consider using cloud GPU services like AWS, Google Cloud, or Azure for enhanced performance, especially when running larger models such as this one.

License

The model is licensed under the META LLAMA 3.1 COMMUNITY LICENSE AGREEMENT, which allows for various uses, including commercial applications, provided they comply with Meta's licensing terms. Users are responsible for the content generated using this model and are encouraged to utilize it responsibly.