Aryanne/QwentileSwap

Hugging Face
TEXT GENERATIONConcurrency Cost:2Model Size:32.8BQuant:FP8Ctx Length:32kPublished:Jan 12, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Aryanne/QwentileSwap is a 32.8 billion parameter language model created by Aryanne, merged using a custom task_swapping method based on win10/EVA-QwQ-32B-Preview. This model integrates components from multiple Qwen2.5-32B variants, including ArliAI/Qwen2.5-32B-ArliAI-RPMax-v1.3, maldv/Qwentile2.5-32B-Instruct, and Sao10K/32B-Qwen2.5-Kunou-v1. It is designed to combine the strengths of its constituent models through a unique merging configuration, offering a distinct blend of capabilities for various generative AI tasks.

Loading preview...

Overview

Aryanne/QwentileSwap is a 32.8 billion parameter language model developed by Aryanne, utilizing a custom task_swapping merge method within mergekit. This model is built upon win10/EVA-QwQ-32B-Preview as its base, integrating layers from several other Qwen2.5-32B instruction-tuned models.

Key Capabilities

  • Custom Merge Architecture: Employs a unique task_swapping merge method with specific diagonal_offset, random_mask, and weight parameters applied to different source models across all 64 layers.
  • Blended Expertise: Combines the characteristics of ArliAI/Qwen2.5-32B-ArliAI-RPMax-v1.3, maldv/Qwentile2.5-32B-Instruct, and Sao10K/32B-Qwen2.5-Kunou-v1 to potentially offer a diverse range of generative and instructional capabilities.
  • High Parameter Count: With 32.8 billion parameters, it is suitable for complex language understanding and generation tasks.

Good For

  • Exploratory AI Development: Ideal for developers interested in experimenting with models created through advanced merging techniques.
  • Diverse Generative Tasks: Its merged nature suggests potential for handling a broad spectrum of prompts, from creative writing to instruction following, depending on the strengths inherited from its constituent models.
  • Research into Model Merging: Provides a practical example of a custom task_swapping configuration for those studying or implementing model fusion strategies.