merge_method: mult
models:
  - model: failspy/Llama-3-8B-Instruct-MopeyMule
  - model: NousResearch/DeepHermes-3-Llama-3-8B-Preview
parameters:
  scale: 3 # adjust as needed
  normalize: true