metadata

license: apache-2.0
tags:
  - merge
  - mergekit
  - lazymergekit
  - bunnycore/Best-Mix-Llama-3.1-8B
  - USTC-KnowledgeComputingLab/Llama3-KALE-LM-Chem-1.5-8B
  - Weyaxi/Einstein-v6.1-Llama3-8B

ZeroXClem/Llama3.1-BestMix-Chem-Einstein-8B

ZeroXClem/Llama3.1-BestMix-Chem-Einstein-8B is a merge of the following models using mergekit:

🧩 Configuration

models:
  - model: bunnycore/Best-Mix-Llama-3.1-8B
    parameters:
      density: [1, 0.7, 0.5] # density gradient, focusing on overall balance
      weight: 1.0

  - model: USTC-KnowledgeComputingLab/Llama3-KALE-LM-Chem-1.5-8B
    parameters:
      density: 0.6 # chemistry focus with significant contribution
      weight: [0.3, 0.7, 1.0] # weight gradient, increasing focus on chemistry tasks

  - model: Weyaxi/Einstein-v6.1-Llama3-8B
    parameters:
      density: 0.4 # focus on conversational adaptability and long-form generation
      weight:
        - filter: mlp
          value: 0.5 # stronger contribution to MLP tasks
        - filter: self_attn
          value: 0.7 # slightly higher weight on attention mechanisms
        - value: 0.5 # balanced contribution for general tasks

merge_method: ties
base_model: bunnycore/Best-Mix-Llama-3.1-8B
parameters:
  normalize: true
  int8_mask: true
dtype: float16