giraffe176/Open_Neural_Monarch_Maidv0.1

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kTool Calling:SupportedPublished:Feb 29, 2024License:cc-by-nc-4.0Architecture:Transformer Open Weights Cold

giraffe176/Open_Neural_Monarch_Maidv0.1 is a 7 billion parameter language model, merged using the DARE TIES method with Mistral-7B-v0.1 as its base. This model integrates components from Intel/neural-chat-7b-v3-1, NeverSleep/Noromaid-7B-0.4-DPO, teknium/OpenHermes-2.5-Mistral-7B, and mlabonne/Monarch-7B. It achieves an average score of 69.28 on the Open LLM Leaderboard, demonstrating balanced performance across various reasoning and language understanding tasks, making it suitable for general-purpose conversational AI and text generation.

Loading preview...

Open_Neural_Monarch_Maidv0.1 Overview

This model, created by giraffe176, is a 7 billion parameter language model built upon the mistralai/Mistral-7B-v0.1 base. It was developed using the DARE TIES merge method from mergekit, combining several specialized models to enhance its capabilities.

Key Components and Merge Details

The Open_Neural_Monarch_Maidv0.1 integrates the strengths of four distinct models:

  • Intel/neural-chat-7b-v3-1
  • NeverSleep/Noromaid-7B-0.4-DPO
  • teknium/OpenHermes-2.5-Mistral-7B
  • mlabonne/Monarch-7B

Each component was weighted and integrated using specific density parameters, with int8_mask enabled and bfloat16 precision for the merge process.

Performance Highlights

Evaluated on the Open LLM Leaderboard, Open_Neural_Monarch_Maidv0.1 demonstrates competitive performance with an average score of 69.28. Notable scores include:

  • AI2 Reasoning Challenge (25-Shot): 67.66
  • HellaSwag (10-Shot): 85.94
  • MMLU (5-Shot): 65.02
  • Winogrande (5-Shot): 79.32
  • GSM8k (5-Shot): 61.33

These results indicate a well-rounded model capable of handling a variety of reasoning, common sense, and language understanding tasks. Detailed evaluation results are available on the Hugging Face Open LLM Leaderboard.