schonsense/Llama-3.3-70B-Inst-Ablit-Flammades-SLERP

Model Overview

schonsense/Llama-3.3-70B-Inst-Ablit-Flammades-SLERP is a 70 billion parameter instruction-following language model, created by merging two base models: huihui-ai/Llama-3.3-70B-Instruct-abliterated and flammenai/Llama3.1-Flammades-70B. The merge was performed using the SLERP (Spherical Linear Interpolation) method, aiming to combine their respective strengths.

Key Characteristics

• Enhanced Creativity & Humanity: This model specifically targets more human-like and creative instruction following, building upon the abliterated Llama 3.3 instruct base.
• Diverse Training Influence: The Flammades component was trained on the Gutenberg dataset and a dataset focused on generating human-like replies, contributing to its nuanced conversational style.
• Instruction-Tuned: Designed to excel in instruction-based tasks, providing coherent and contextually relevant responses.
• Context Length: Supports a substantial context window of 32768 tokens.

Intended Use Cases

This model is particularly suited for applications requiring:

• Creative Writing: Generating imaginative narratives, dialogues, and descriptive text.
• Roleplay Scenarios: Producing engaging and character-consistent responses in interactive storytelling.
• Human-like Conversation: Developing chatbots or virtual assistants that aim for more natural and empathetic interactions.
• Instruction Following: Executing complex instructions with a focus on nuanced and creative output.

Users are advised to experiment with sampler settings (e.g., temperature, repetition penalty) to fine-tune the model's output for desired tone and consistency, as highlighted by the creator's provided example chat and settings.