ajibawa-2023/Uncensored-Jordan-7B
Uncensored-Jordan-7B by ajibawa-2023 is a 7 billion parameter language model, fine-tuned from Llama-1 with a 4096-token context length. It is specifically designed for uncensored conversations, allowing users to discuss a wide range of topics without typical AI guardrails. The model aims to provide unfiltered discussions on politics, religion, and other contentious issues, drawing inspiration from the concept of unconstrained dialogue.
Loading preview...
Uncensored-Jordan-7B: An Unfiltered Conversational Model
Uncensored-Jordan-7B, developed by ajibawa-2023, is a 7 billion parameter language model fine-tuned from Llama-1. Inspired by the concept of unconstrained dialogue, this model is designed to facilitate uncensored conversations across a broad spectrum of topics, including politics, religion, and other subjects often restricted by conventional AI models. It was trained using a dataset of approximately 155,000 conversation sets over three epochs on Azure A100 GPUs, utilizing the DeepSpeed codebase.
Key Characteristics & Capabilities
- Uncensored Dialogue: Engineered to engage in discussions without typical AI guardrails, promoting unfiltered exchanges.
- Llama-1 Base: Built upon the Llama-1 architecture, noted for its utility in uncensored conversational contexts.
- Full Fine-Tune: Represents a full fine-tuned model, with quantized versions (GPTQ, GGUF, AWQ) also available.
- Context Length: Supports a context window of 4096 tokens.
Performance Benchmarks
Evaluated on the Open LLM Leaderboard, Uncensored-Jordan-7B achieved an average score of 46.88. Notable scores include 77.37 on HellaSwag (10-shot) and 71.11 on Winogrande (5-shot), with an MMLU (5-shot) score of 45.69.
Important Considerations
Users are advised that this model has minimal guardrails and are responsible for its outputs. It is intended for candid discussions and should be used with an understanding of its uncensored nature.