base_model: google/gemma-3-4b-it
merge_method: ties
dtype: bfloat16
models:
  - model: google/gemma-3-4b-it
    parameters:
      weight: 0.4
    parameters:
      weight: 0.2
  - model: Qwen/Qwen2.5-3B-Instruct
    parameters:
      weight: 0.2
  - model: Qwen/Qwen2.5-3B-Instruct
    parameters:
      weight: 0.2
parameters:
  normalize: true
  int8_mask: true