Gemma The Writer N Restless Quill 10 B Uncensored G G U F
DavidAUIntroduction
The Gemma-The-Writer-N-Restless-Quill-10B-Uncensored-GGUF is a text generation model designed for creative and fiction writing tasks. It is uncensored, allowing for a wide range of expressions, including swearing and horror. This model is particularly adept at generating stories across various genres such as science fiction and romance, maintaining vivid prose and strong storytelling capabilities.
Architecture
This model is a merge of several top storytelling and writing models, including "gemma2-gutenberg," "Tiger-Gemma," and "Ellaria." It uses a "Brainstorm 5X" adapter developed by DavidAU, adding nearly 1 billion parameters to the model, resulting in a total of 10 billion parameters. These enhancements focus on improving prose, sentence structure, and reducing GPT-like outputs, thus increasing output length and quality.
Training
The model has been trained by modifying and merging multiple models at the layer level, resulting in a high-precision merge. The Brainstorm process alters the model's reasoning centers, enhancing prose quality, detail, and emotional nuance without compromising instruction following. This process introduces subtle changes that improve the model's storytelling capabilities.
Guide: Running Locally
- Setup Environment: Ensure you have Python and necessary libraries installed.
- Download Model: Obtain the model files from the Hugging Face repository.
- Install Dependencies: Use a package manager to install any required libraries, such as transformers and PyTorch.
- Configure Settings: Adjust parameters like temperature, repetition penalty, and context window to optimize performance.
- Run Model: Use a script or notebook to load the model and generate text based on prompts.
Suggested Cloud GPUs
For optimal performance, consider using cloud GPUs from providers like AWS, Google Cloud, or Azure. These platforms offer powerful GPUs that can handle the model's computational demands, especially for high quant settings like Q8_0.
License
The model is licensed under the Apache-2.0 license, which allows for both personal and commercial use with attribution. For more detailed information, refer to the full license text.