RDson/WomboCombo-R1-Coder-14B-Preview

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:14.8BQuant:FP8Ctx Length:32kPublished:Jan 24, 2025Architecture:Transformer0.0K Warm

RDson/WomboCombo-R1-Coder-14B-Preview is a 14.8 billion parameter language model, merged using the 'sce' method with Qwen/Qwen2.5-Coder-14B as its base. This model integrates components from Qwen/Qwen2.5-Coder-14B-Instruct, deepseek-ai/DeepSeek-R1-Distill-Qwen-14B, and arcee-ai/SuperNova-Medius. It is specifically designed for coding tasks, leveraging its merged architecture to enhance code generation and understanding capabilities.

Loading preview...

WomboCombo-R1-Coder-14B-Preview Overview

RDson/WomboCombo-R1-Coder-14B-Preview is a 14.8 billion parameter language model, created through a merge of several pre-trained models using the mergekit tool and the sce merge method. Its foundation is built upon Qwen/Qwen2.5-Coder-14B, a model known for its coding proficiency.

Key Capabilities

This model integrates the strengths of multiple specialized models to enhance its performance in coding-related tasks. The merge includes:

  • Qwen/Qwen2.5-Coder-14B-Instruct: Likely contributes instruction-following capabilities tailored for coding.
  • deepseek-ai/DeepSeek-R1-Distill-Qwen-14B: Suggests a focus on distilled knowledge, potentially improving efficiency or specific coding domains.
  • arcee-ai/SuperNova-Medius: Implies additional general or specialized language understanding.

The combination aims to create a robust model for code generation, completion, and understanding, leveraging the distinct features of its constituent models.

Good For

  • Code Generation: Creating new code snippets or functions based on natural language prompts.
  • Code Completion: Assisting developers by suggesting relevant code during development.
  • Code Understanding: Analyzing and interpreting existing codebases.
  • Instruction-following in Coding: Executing complex coding instructions effectively.