Gryphe/MythoLogic-13b

TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kPublished:Jul 12, 2023License:otherArchitecture:Transformer0.0K Cold

Gryphe/MythoLogic-13b is a 13 billion parameter language model created by Gryphe, developed through a gradient merge of Chronos, Nous-Hermes, and Wizard-Vicuna Uncensored. This model is designed to enhance story generation and roleplaying capabilities, particularly by augmenting Chronos's verbosity with improved logic from Hermes and Wizard-Vicuna. It features a 4096-token context length and is optimized for all-round performance in creative text generation.

Loading preview...

MythoLogic-13b: Merged for Enhanced Creativity

MythoLogic-13b is a 13 billion parameter model developed by Gryphe, resulting from an experimental gradient merge of several base models. Its primary component is Chronos-13b, known for its verbose output, augmented with Nous-Hermes-13b and Wizard-Vicuna-13B-Uncensored to introduce improved logical coherence, especially in higher layers (10+).

Key Capabilities

  • Improved Story Generation: Designed to produce more coherent and engaging narratives.
  • Enhanced Roleplaying: Optimized for dynamic and responsive roleplay scenarios.
  • All-Round Performance: Aims to balance verbosity with logical consistency for diverse creative tasks.

Prompt Format

This model primarily utilizes Alpaca formatting. For optimal performance, 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:

Quantized versions (GGML and GPTQ) are also available from TheBloke for broader accessibility.