Hugging Face Overview
Model Overview
Doctor-Shotgun/L3.3-70B-Magnum-Diamond is a 70 billion parameter language model, fine-tuned from meta-llama/Llama-3.3-70B-Instruct using an rsLoRA adapter. This model is an iteration in the Magnum series, specifically designed to enhance creative writing and roleplay capabilities, aiming to replicate the prose quality of Claude 3 Sonnet/Opus models.
Key Capabilities
- Creative Writing & Roleplay: Optimized for generating high-quality, engaging narrative content and character interactions.
- Llama 3.3-Instruct Base: Leverages the robust foundation of the Llama 3.3-Instruct architecture.
- rsLoRA Fine-tuning: Utilizes an rsLoRA adapter for efficient and effective fine-tuning, with modifications to custom loss masking and pre-tokenization.
- Flexible Prompting: Designed to perform competently with or without prepending character names and prefill in roleplay scenarios.
- Extended Context: Supports a context length of 32768 tokens.
Intended Use Cases
- Creative Story Generation: Ideal for authors, writers, and enthusiasts looking for assistance in generating fictional narratives.
- Interactive Roleplay: Suited for applications requiring dynamic and immersive character-driven interactions.
- Emulating Claude 3 Style: Users seeking outputs with a prose style similar to Claude 3 Sonnet/Opus models will find this model particularly useful.
Usage Notes
This model adheres to the Llama 3 prompt format. Prefill is optional but recommended for roleplay. Quantized GGUF versions are available. Recommended sampler settings include temperature = 1.0 and min_p = 0.1 for optimal creative output.