models:
  - model: google/gemma-2-2b-it
    parameters:
      weight: 1
merge_method: ties
base_model: google/gemma-2-2b
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16