Azazelle/Maylin-7b

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Jan 4, 2024License:cc-by-4.0Architecture:Transformer Open Weights Cold

Azazelle/Maylin-7b is a 7 billion parameter language model based on the Mistral-7B-v0.1 architecture, created through a DARE merge. This model is specifically designed to enhance coherence and reduce undesirable biases present in the Argetsu model. It aims to provide a more balanced and focused output for general language generation tasks.

Loading preview...

Maylin-7b Model Overview

Maylin-7b is a 7 billion parameter language model developed by Azazelle, built upon the Mistral-7B-v0.1 base architecture. This model was created using a DARE (DARE_TIES) merge technique, combining several other models to achieve its specific characteristics.

Key Capabilities

  • Enhanced Coherence: The primary goal of Maylin-7b is to improve the overall coherence of generated text.
  • Bias Reduction: It specifically targets and aims to mitigate certain undesirable biases, such as excessive 'horniness', observed in its constituent models like Argetsu.
  • General Language Generation: Suitable for a variety of text generation tasks where balanced and coherent output is desired.

Merge Details

The model was constructed using a DARE_TIES merge method, integrating the following models with specific weights and densities:

  • mistralai/Mistral-7B-v0.1 (base model)
  • SanjiWatsuki/Sonya-7B (weight: 0.45, density: 0.75)
  • Azazelle/Argetsu (weight: 0.39, density: 0.70)
  • Azazelle/Tippy-Toppy-7b (weight: 0.22, density: 0.52)

This specific merge configuration was chosen to refine the output characteristics, making Maylin-7b a more controlled and reliable option for general use cases.