M N 12 B Mag Mell R1

Introduction

MN-12B-MAG-MELL-R1 is a merged pre-trained language model designed for text generation and creative tasks. It integrates multiple models using mergekit to produce a generalized model that excels in narrative and fictional text generation.

Architecture

The architecture of MN-12B-MAG-MELL-R1 is a composite of several pre-trained models. It employs a multi-stage merge inspired by hyper-merges. The merged models include:

• IntervitensInc/Mistral-Nemo-Base-2407-chatml
• nbeerbower/mistral-nemo-bophades-12B
• nbeerbower/mistral-nemo-wissenschaft-12B
• elinas/Chronos-Gold-12B-1.0
• Fizzarolli/MN-12b-Sunrose
• nbeerbower/mistral-nemo-gutenberg-12B-v4
• anthracite-org/magnum-12b-v2.5-kto

Training

The model was created using the DARE-TIES merge method, which combines models by evaluating them in specific domains. This involved the use of layer-weighted SLERP to merge intermediate "specialist" models before integrating them into the base model. The merged model is designed to balance qualities from each component part, aiming for superior narrative capabilities.

Guide: Running Locally

To run MN-12B-MAG-MELL-R1 locally:

1. Environment Setup: Install the transformers library and set up Python with required dependencies.
2. Model Download: Retrieve the model from Hugging Face's model hub.
3. Loading the Model: Load the model using the transformers library.
4. Inference: Utilize the model for text generation tasks by customizing parameters like temperature and MinP based on stability requirements.

For enhanced performance, consider using cloud-based GPUs such as those provided by AWS, Google Cloud, or Azure.

License

The usage, modification, and distribution of MN-12B-MAG-MELL-R1 are subject to the licensing terms specified by the original model contributors on Hugging Face. Ensure compliance with these terms before usage.