Hugging Face Overview
Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2 Overview
This model, developed by Orenguteng, is a fine-tuned version of Meta's Llama-3.1-8B-Instruct, featuring 8 billion parameters and a 32768 token context window. It is designed to be uncensored and highly compliant with user prompts, including those that might be considered unethical, providing maximum flexibility for developers. Users are advised to implement their own alignment layers when deploying the model as a service.
Key Characteristics & Usage Notes
- Uncensored & Compliant: Lexi is engineered to be highly responsive to all requests, even those with potentially unethical content. Users are responsible for the content generated.
- Base Model: It serves as a powerful base for developers to build custom applications, requiring external alignment layers for responsible deployment.
- Llama 3.1 License: The model operates under the META LLAMA 3.1 COMMUNITY LICENSE AGREEMENT, permitting commercial use in accordance with its terms.
- Inference Template: Requires the same system tokens and template as the official Llama 3.1 8B instruct model for optimal performance.
- Quantization Note: The developer notes potential refusal issues with Q4 quantization and suggests using F16 or Q8 if possible, with plans to address this in future versions.
Performance Insights
Evaluations on the Open LLM Leaderboard show an average score of 27.93. Specific metrics include:
- IFEval (0-Shot): 77.92
- BBH (3-Shot): 29.69
- MMLU-PRO (5-shot): 30.90
Detailed results are available on the Open LLM Leaderboard.