MRockatansky/Gemma-4-31B-storymaxxed
MRockatansky/Gemma-4-31B-storymaxxed is a 31 billion parameter language model, fine-tuned from trohrbaugh/gemma-4-31b-it-heretic-ara. It is specifically optimized for long-form creative storywriting and narrative prose. The model was trained using Direct Preference Optimization (DPO) on a high-quality dataset of narrative preference pairs, enhancing its ability to generate coherent and engaging stories. With a context length of 32768 tokens, it is well-suited for extended creative writing tasks.
Loading preview...
Model Overview
MRockatansky/Gemma-4-31B-storymaxxed is a 31 billion parameter language model, building upon the trohrbaugh/gemma-4-31b-it-heretic-ara base. Its core differentiation lies in its specialized fine-tuning for long-form creative storywriting and narrative prose.
Key Capabilities & Training
- Specialized for Narrative Generation: Optimized to produce coherent and engaging long-form stories.
- Direct Preference Optimization (DPO): Trained using DPO on over 5,000 high-quality narrative preference pairs for 8 hours, enhancing its ability to align with human preferences for storytelling.
- Extended Context: Supports a context length of 32768 tokens, facilitating the generation of longer, more complex narratives.
- Recommended Settings: Optimal inference is achieved with specific sampler settings: Temperature 1.0, Top P 0.95, and Top K 64.
Use Cases
- Creative Writing: Ideal for generating fiction, short stories, novel chapters, and other narrative content.
- Roleplay Scenarios: Can be used to develop detailed character interactions and plotlines.
- Content Generation: Suitable for tasks requiring extensive, descriptive text generation with a focus on storytelling.
This model provides a robust foundation for developers and writers looking to leverage an LLM specifically tailored for the nuances of creative narrative generation.