Gryphe/MythoMax-L2-13b

Warm
Public
13B
FP8
4096
License: other
Hugging Face
Overview

MythoMax-L2-13b: A Merged Model for Creative Text Generation

MythoMax-L2-13b is a 13 billion parameter model developed by Gryphe, representing an advanced variant of the MythoMix concept. It is created through a unique and highly experimental tensor type merge technique, combining Gryphe's own MythoLogic-L2-13b and The-Face-Of-Goonery's Huginn-13b-FP16 models.

Key Capabilities

  • Enhanced Coherency: The merge process allowed for greater intermingling of Huginn's tensors across the model's structure, leading to increased coherency.
  • Proficiency in Roleplaying: The model excels at generating engaging and consistent roleplay scenarios.
  • Strong Storywriting: It demonstrates strong capabilities in creative story generation.
  • Unique Merge Methodology: Each of the model's 363 tensors had a unique ratio applied, incorporating gradients to fine-tune its behavior.

Good For

  • Roleplaying applications: Ideal for interactive narrative experiences and character-driven dialogues.
  • Creative writing tasks: Suitable for generating stories, descriptive passages, and other forms of creative text.
  • Developers experimenting with advanced merge techniques: Offers insights into a novel tensor-level merging approach.

Prompt Format

The model primarily uses Alpaca formatting. For optimal results, structure prompts as follows:

<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: