MRockatansky/Gemma-4-31B-storymaxxed2
MRockatansky/Gemma-4-31B-storymaxxed2 is a 31 billion parameter language model fine-tuned from trohrbaugh/gemma-4-31b-it-heretic-ara. This model is specifically optimized for creative writing and narrative prose, leveraging Direct Preference Optimization (DPO) on a high-quality dataset of narrative preference pairs. It is designed to excel in generating engaging and coherent stories, making it suitable for applications requiring advanced narrative capabilities.
Loading preview...
Model Overview
MRockatansky/Gemma-4-31B-storymaxxed2 is a 31 billion parameter language model, fine-tuned from the trohrbaugh/gemma-4-31b-it-heretic-ara base model. It has been specifically optimized for creative writing and narrative prose through a fine-tuning process using TRL.
Key Capabilities
- Narrative Generation: Excels at producing engaging and coherent stories and creative text.
- Direct Preference Optimization (DPO): Trained using DPO on a high-quality dataset of over 5,000 narrative preference pairs, enhancing its ability to generate preferred narrative styles.
- Vision Support: Includes an
mmprojfile for vision capabilities, allowing for multimodal applications (available separately).
When to Use This Model
This model is particularly well-suited for use cases requiring advanced creative text generation, such as:
- Story writing and plot generation.
- Role-playing scenarios.
- Content creation for narrative-driven applications.
For optimal inference, the recommended sampler settings are Temperature 1.0, Top P 0.95, and Top K 64, as suggested for Gemma-4 models.