MythoMax-L2-13b Open-source Model - Free Deployment Empowers Exceptional Role-playing and Story Creation

Mythomax L2 13b

Developed by Gryphe

MythoMax is an improved version fused from MythoLogic-L2 and Huginn models using experimental tensor merging technology, excelling in role-playing and story creation.

Large Language Model

Transformers

EnglishOpen Source License:Other #Tensor Fusion Technology #Role-playing Optimization #Story Creation Enhancement

Downloads 6,150

Release Time : 8/10/2023

Model Overview

This model employs highly experimental tensor-type merging technology, combining the comprehension capabilities of MythoLogic-L2 with the writing abilities of Huginn, delivering outstanding performance in text generation tasks.

Model Features

Tensor Fusion Technology

Utilizes unique tensor-type merging technology, applying different scaling coefficients to each of the 363 tensors to achieve optimized combination of model capabilities.

Dual-Model Advantage Integration

Combines the comprehension capabilities of MythoLogic-L2 with the writing abilities of Huginn, excelling in both input understanding and output quality.

Gradient Fine-tuning

Introduces gradient adjustments during the fusion process to further optimize model behavior.

Model Capabilities

Role-playing dialogue generation

Story creation

Natural language understanding

Coherent text generation

Use Cases

Entertainment

Role-playing Games

Generates high-quality character dialogue content

Provides immersive role-playing experiences

Creative Writing

Assists in story creation and content generation

Produces coherent and creative text content

Education

Writing Assistance

Helps students with creative writing exercises

Enhances writing interest and skills

🚀 MythoMax - A Powerful Model Merge

With the release of Llama 3, it's time for MythoMax to gracefully step aside. But let's do it in style!

MythoMax is an improved, and potentially perfected, variant of MythoMix. It's a merge of my MythoLogic-L2 and Huginn using a highly experimental tensor type merge technique. The key difference from MythoMix is that I allowed more of Huginn to blend with the single tensors at the front and end of the model, enhancing the overall coherency.

The script and the accompanying templates used to create this merge can be found here.

Due to its unique nature, this model excels in both role - playing and story - writing.

Quantized models are available from TheBloke: GGUF - GPTQ - AWQ. (You're the best!)

🚀 Quick Start

This section provides an overview of the model and its capabilities. For more detailed information, refer to the following sections.

✨ Features

Improved Merge: An enhanced version of MythoMix, with better coherency.
Dual Capability: Excels in both role - playing and story - writing.
Quantized Models: Available in GGUF, GPTQ, and AWQ formats from TheBloke.

📚 Documentation

Model details

The concept behind this merge is that each layer consists of multiple tensors, each responsible for specific functions. Using MythoLogic - L2's strong understanding as input and Huginn's extensive writing ability as output has seemingly resulted in a model that performs well in both aspects, validating my theory. (More details will be released later)

This type of merge cannot be visualized because each of its 363 tensors has a unique ratio applied. Similar to my previous merges, gradients were incorporated into these ratios to further fine - tune its behavior.

Prompt Format

This model mainly uses Alpaca formatting. For optimal performance, use the following format:

<System prompt/Character Card>

### Instruction:
Your instruction or question here.
For roleplay purposes, I suggest the following - Write <CHAR NAME>'s next reply in a chat between <YOUR NAME> and <CHAR NAME>. Write a single reply only.

### Response:

📄 License

This project is licensed under the "other" license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご