COMMAND-R-01-ULTRA-NEO-DARK-HORROR V1-V2 35B IMATRIX GGUF

Introduction

COMMAND-R-01-ULTRA-NEO-DARK-HORROR is a text generation model optimized for creating intense horror-themed content. Utilizing two versions, it enhances horror prompts with vivid, psychological elements, avoiding gore, to produce deeply immersive narratives.

Architecture

The model features two versions, V1 and V2, using the DARK HORROR Neo Imatrix datasets. These datasets apply a "dark tint" to creative prompts, increasing the intensity and vividness of prose. The model is designed for horror and general creative writing applications, with a focus on high precision and immersive storytelling.

Training

The DARK HORROR datasets were created using the Grand Horror 16B model with over 90 horror prompts. This process enhances the model's ability to generate intense horror narratives by applying a lighter form of tuning, distinct from traditional fine-tuning.

Guide: Running Locally

1. Download the Model: Choose between V1 and V2 versions based on your use case. Both versions provide different horror effects.
2. Select Quants: For maximum horror effect, use lower quants like IQ1s to IQ4s. Lower quants enhance the Imatrix effect, providing a stronger horror tint.
3. Configure Environment: Use text-generation-webui or KoboldCpp with appropriate settings:
   • Set "Smoothing_factor" to 1.5 to 2.5.
   • If using GGUFs, apply the "llama_HF" configuration.
4. Run the Model: Use a cloud GPU for better performance, such as NVIDIA T4 or V100, to handle large models like this one efficiently.
5. Optimize Performance: Adjust parameters and samplers for optimal results. Refer to the detailed guide on maximizing model performance for specific use cases.

License

This model is licensed under the Apache-2.0 License, allowing for broad use and modification with proper attribution.