|
--- |
|
base_model: unsloth/deepseek-r1-distill-llama-8b-unsloth-bnb-4bit |
|
tags: |
|
- text-generation-inference |
|
- transformers |
|
- unsloth |
|
- llama |
|
- trl |
|
- sft |
|
license: apache-2.0 |
|
language: |
|
- es |
|
- en |
|
--- |
|
|
|
 |
|
|
|
|
|
### **About the Model** |
|
|
|
This model is designed to be a storytelling AI capable of creating fun, engaging, and well-structured narratives. Its purpose is to serve as an interactive tool for generating and experiencing unique stories in real time, tailored to the user's input and preferences. |
|
|
|
### **Key Features** |
|
|
|
- **Interactive Narratives:** Produces coherent and entertaining stories based on user prompts, adapting dynamically to maintain engagement. |
|
- **Consistent World-Building:** Ensures logical progression and consistency in characters, settings, and events across long narratives. |
|
- **Optimized for Efficiency:** Built to perform reliably on limited hardware while delivering high-quality outputs. |
|
|
|
### **Training Overview** |
|
|
|
The model was fine-tuned using datasets focused on narrative construction, character development, and immersive descriptions. Key aspects of the training include: |
|
|
|
- **Adaptability:** Special attention was given to creating a system that responds flexibly to varied user inputs while maintaining coherence. |
|
- **Resource Efficiency:** Techniques like LoRA (Low-Rank Adaptation) and 4-bit quantization were employed to optimize memory usage without compromising output quality. |
|
- **Long-Context Support:** Enhanced with methods to handle extended interactions and complex storylines. |
|
|
|
### **Purpose** |
|
|
|
The primary goal of this model is to create a personal, customizable storytelling AI, allowing users to immerse themselves in unique, AI-driven stories anytime. |
|
|
|
--- |
