CultriX/Qwen2.5-14B-HyperMarck-dl

TEXT GENERATIONConcurrency Cost:1Model Size:14.8BQuant:FP8Ctx Length:32kPublished:Feb 16, 2025License:apache-2.0Architecture:Transformer Open Weights Cold

CultriX/Qwen2.5-14B-HyperMarck-dl is a 14.8 billion parameter language model created by CultriX, based on the Qwen2.5 architecture. This model is a merge of pre-trained language models, specifically utilizing the Linear DELLA merge method with suayptalha/Lamarckvergence-14B as its base. It integrates components from CultriX/MergeStage1v3 and CultriX/MergeStage2v3, offering a combined set of capabilities from its constituent models. With a 32768 token context length, it is designed for general language understanding and generation tasks.

Loading preview...

Model Overview

CultriX/Qwen2.5-14B-HyperMarck-dl is a 14.8 billion parameter language model developed by CultriX, built upon the Qwen2.5 architecture. This model is a product of a sophisticated merge process, combining the strengths of multiple pre-trained language models.

Merge Details

This model was created using the Linear DELLA merge method, a technique designed to effectively combine different model components. The base model for this merge was suayptalha/Lamarckvergence-14B. Key components integrated into this merge include:

The merge configuration utilized specific parameters for epsilon, lambda, and density for each contributing model, aiming to optimize the combined performance. The model supports a substantial context length of 32768 tokens.

Potential Use Cases

Given its foundation in the Qwen2.5 architecture and its merged nature, this model is suitable for a variety of general-purpose natural language processing tasks, including:

  • Text generation
  • Question answering
  • Summarization
  • Conversational AI