README.md

metadata

license: apache-2.0
tags:
  - merge
  - mergekit

StarFuse-7B-DARE

StarFuse-7B-DARE is a merge of the following models using mergekit:

• FuseAI/FuseChat-7B-VaRM
• Nexusflow/Starling-LM-7B-beta
• Weyaxi/Newton-7B

🧩 Configuration

models:
  - model: openchat/openchat-3.5-0106
    # No parameters necessary for base model
  - model: FuseAI/FuseChat-7B-VaRM
    parameters:
      density: 0.5
      weight: 0.5
  - model: Nexusflow/Starling-LM-7B-beta
    parameters:
      density: 0.5
      weight: 0.5
  - model: Weyaxi/Newton-7B
    parameters:
      density: 0.5
      weight: 0.5

merge_method: dare_ties
base_model: openchat/openchat-3.5-0106
parameters:
  int8_mask: true
dtype: bfloat16