B N I M I T A L3 8 B v0.02 G G U F

Introduction

B-NIMITA is an AI model designed to enhance role-playing scenarios through emotional depth and storytelling. It integrates NIHAPPY for narrative foundation, Mythorica for emotional arcs and expressive dialogue, and V-Blackroot for character consistency.

Architecture

B-NIMITA's architecture involves a merge of pre-trained language models using the DARE TIES method. It combines multiple models: NIHAPPY (core narrative), Mythorica (expressiveness), and V-Blackroot (character consistency).

Training

The model was created using a merge of several models with specific weights and densities. The merging process used the DARE TIES method, and configurations were set using YAML parameters to balance narrative flow and emotional expressiveness.

Guide: Running Locally

1. Prerequisites:

   • Install transformers and mergekit libraries.
   • Download the B-NIMITA model from Hugging Face.

2. Setup:

   • Clone the repository containing B-NIMITA.
   • Configure the environment for bfloat16 data type.

3. Execution:

   • Use the model with compatible inference endpoints.
   • For optimal performance, consider using cloud GPUs such as AWS or Google Cloud.

License

The use of B-NIMITA is subject to the licensing terms provided by the creators of the merged models. Ensure compliance with all applicable licenses when using or distributing the model.