FlareRebellion/WeirdCompound-v1.5-24b

TEXT GENERATIONConcurrency Cost:2Model Size:24BQuant:FP8Ctx Length:32kArchitecture:Transformer0.0K Cold

WeirdCompound-v1.5-24b is a 24 billion parameter language model created by FlareRebellion through a multi-stage merge process using mergekit. This model integrates components from various base models, including TheDrummer/Cydonia-24B-v4, aixonlab/Eurydice-24b-v3.5, and PocketDoc/Dans-PersonalityEngine-V1.3.0-24b, among others. It is designed to combine diverse capabilities, with specific components aimed at storytelling, roleplay, prompt adherence, and adventure generation, making it suitable for creative text generation tasks.

Loading preview...

Overview

WeirdCompound-v1.5-24b is a 24 billion parameter language model developed by FlareRebellion, constructed through a sophisticated multi-stage merging process using mergekit. This iteration, v1.5, refines previous versions by adjusting the blend of its constituent models to achieve a desired output profile. The merge process involved several methods, including Model Stock, SLERP, and NuSLERP, with TheDrummer/Cydonia-24B-v4 serving as a primary base.

Key Capabilities

  • Multi-faceted Integration: Combines elements from models like aixonlab/Eurydice-24b-v3.5 for storytelling/roleplay, PocketDoc/Dans-PersonalityEngine-V1.3.0-24b for prompt adherence, and Delta-Vector/MS3.2-Austral-Winton for adventure narratives.
  • Iterative Refinement: The model's development involved multiple version updates (v1.1 to v1.5), with specific changes to intermediate models and merge recipes to fine-tune its characteristics.
  • Diverse Component Models: Incorporates a wide array of specialized models, suggesting a broad range of potential applications in creative text generation.

Good For

  • Creative Writing: Excellent for generating stories, roleplay scenarios, and adventure narratives due to its integrated components.
  • Prompt Adherence: Benefits from Dans-PersonalityEngine-V1.3.0-24b, which is noted for improving adherence to user prompts.
  • Experimental Merging: Represents an advanced example of model merging techniques, suitable for users interested in exploring complex model compositions.

Finetunes
1 models